{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "# Epidemiological models: Introduction\n", "\n", "This tutorial introduces the [pyro.contrib.epidemiology](http://docs.pyro.ai/en/latest/contrib.epidemiology.html) module, an epidemiological modeling language with a number of black box inference algorithms. This tutorial assumes the reader is already familiar with [modeling](http://pyro.ai/examples/intro_long.html), [inference](http://pyro.ai/examples/intro_long.html), and [distribution shapes](http://pyro.ai/examples/tensor_shapes.html).\n", "\n", "See also the following scripts:\n", "\n", "- [Epidemiological models: Univariate](http://pyro.ai/examples/epi_sir.html)\n", "- [Epidemiological models: Regional](http://pyro.ai/examples/epi_regional.html)\n", "- [Epidemiological inference via HMC](http://pyro.ai/examples/sir_hmc.html)\n", "\n", "#### Summary\n", "\n", "- To create a new model, inherit from the [CompartmentalModel](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel) base class.\n", "- Override methods [.global_model()](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel.global_model), [.initialize(params)](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel.initialize), and [.transition(params, state, t)](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel.transition).\n", "- Take care to support broadcasting and vectorized interpretation in those methods.\n", "- For single time series, set `population` to an integer.\n", "- For batched time series, let `population` be a vector, and use [self.region_plate](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel).\n", "- For models with complex inter-compartment flows, override the [.compute_flows()](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel.compute_flows) method. \n", "- Flows with loops (undirected or directed) are not currently supported.\n", "- To perform cheap approximate inference via SVI, call the [.fit_svi()](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel.fit_svi) method.\n", "- To perform more expensive inference via MCMC, call the [.fit_mcmc()](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel.fit_mcmc) method.\n", "- To stochastically predict latent and future variables, call the [.predict()](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel.predict) method.\n", "\n", "#### Table of contents\n", "\n", "- [Basic workflow](#Basic-workflow)\n", " - [Modeling](#Modeling)\n", " - [Generating data](#Generating-data)\n", " - [Inference](#Inference)\n", " - [Prediction](#Prediction)\n", " - [Forecasting](#Forecasting)\n", "- [Advanced modeling](#Advanced-modeling)\n", " - [Regional models](#Regional-models)\n", " - [Phylogenetic likelihoods](#Phylogenetic-likelihoods)\n", " - [Heterogeneous models](#Heterogeneous-models)\n", " - [Complex compartment flow](#Complex-compartment-flow)\n", "- [References](#References)" ] }, { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [], "source": [ "import os\n", "import matplotlib.pyplot as plt\n", "import seaborn as sns\n", "import torch\n", "import pyro\n", "import pyro.distributions as dist\n", "from pyro.contrib.epidemiology import CompartmentalModel, binomial_dist, infection_dist\n", "\n", "%matplotlib inline\n", "assert pyro.__version__.startswith('1.9.0')\n", "torch.set_default_dtype(torch.double) # Required for MCMC inference.\n", "smoke_test = ('CI' in os.environ)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Basic workflow \n", "\n", "The [pyro.contrib.epidemiology](http://docs.pyro.ai/en/latest/contrib.epidemiology.html) module provides a modeling language for a class of stochastic discrete-time discrete-count compartmental models, together with a number of black box inference algorithms to perform joint inference on global parameters and latent variables. This modeling language is more restrictive than the full Pyro probabilistic programming language:\n", "\n", "- control flow must be static;\n", "- compartmental distributions are restricted to [binomial_dist()](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.distributions.binomial_dist), [beta_binomial_dist()](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.distributions.beta_binomial_dist), and [infection_dist()](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.distributions.infection_dist);\n", "- plates are not allowed, except for the single optional [.region_plate](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel.region_plate);\n", "- all random variables must be either global or Markov and sampled at every time step, so e.g. time-windowed random variables are not supported;\n", "- models must support broadcasting and vectorization of time `t`.\n", "\n", "These restrictions allow inference algorithms to vectorize over the time dimension, leading to inference algorithms with per-iteration parallel complexity sublinear in length of the time axis. The restriction on distributions allows inference algorithms to approximate parts of the model as Gaussian via moment matching, further speeding up inference. Finally, because real data is so often overdispersed relative to Binomial idealizations, the three distribution helpers provide an [overdispersion](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.distributions.binomial_dist) parameter calibrated so that in the large-population limit all distribution helpers converge to log-normal.\n", "\n", "Black box inference algorithms currently include: [SVI](http://docs.pyro.ai/en/latest/inference_algos.html#pyro.infer.svi.SVI) with a moment-matching approximation, and [NUTS](http://docs.pyro.ai/en/latest/mcmc.html#pyro.infer.mcmc.NUTS) either with a moment-matched approximation or with an exact auxiliary variable method detailed in the [SIR HMC tutorial](http://pyro.ai/examples/sir_hmc.html). All three algorithms initialize using [SMC](http://docs.pyro.ai/en/latest/inference_algos.html#pyro.infer.smcfilter.SMCFilter) and reparameterize time dependent variables using a fast [Haar wavelet](http://docs.pyro.ai/en/latest/infer.reparam.html#pyro.infer.reparam.haar.HaarReparam) transform. Default inference parameters are set for cheap approximate results; accurate results will require more steps and ideally comparison among different inference algorithms. We recommend that, when running MCMC inference, you use multiple chains, thus making it easier to diagnose mixing issues.\n", "\n", "While MCMC inference can be more accurate for a given model, SVI is much faster and thus allows richer model structure (e.g. incorporating neural networks) and more rapid [model iteration](https://www.annualreviews.org/doi/abs/10.1146/annurev-statistics-022513-115657?journalCode=statistics). We recommend starting model exploration using mean field SVI (via `.fit_svi(guide_rank=0)`), then optionally increasing accuracy using a low-rank multivariate normal guide (via `.fit_svi(guide_rank=None)`). For even more accurate posteriors you could then try moment-matched MCMC (via `.fit_mcmc(num_quant_bins=1)`), or the most accurate and most expensive enumerated MCMC (via `.fit_mcmc(num_quant_bins=4)`). We recommend that, when fitting models with neural networks, you train via `.fit_svi()`, then freeze the network (say by omitting a `pyro.module()` statement) before optionally running MCMC inference." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Modeling \n", "\n", "The [pyro.contrib.epidemiology.models](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#module-pyro.contrib.epidemiology.models) module provides a number of example models. While in principle these are reusable, we recommend forking and modifying these models for your task. Let's take a look at one of the simplest examples, [SimpleSIRModel](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.models.SimpleSIRModel). This model derives from the [CompartmentalModel](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel) base class and overrides the three standard methods using familiar Pyro modeling code in each method.\n", "\n", "- [.global_model()](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel.global_model) samples global parameters and packs them into a single return value (here a tuple, but any structure is allowed). The return value is available as the `params` argument to the other two methods.\n", "- [.initialize(params)](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel.initialize) samples (or deterministically sets) initial values of time series, returning a dictionary mapping time series name to initial value.\n", "- [.transition(params, state, t)](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel.transition) inputs global `params`, the `state` at the previous time step, and the time index `t` (which may be a slice!). It then samples flows and updates the state dict." ] }, { "cell_type": "code", "execution_count": 2, "metadata": {}, "outputs": [], "source": [ "class SimpleSIRModel(CompartmentalModel):\n", " def __init__(self, population, recovery_time, data):\n", " compartments = (\"S\", \"I\") # R is implicit.\n", " duration = len(data)\n", " super().__init__(compartments, duration, population)\n", " assert isinstance(recovery_time, float)\n", " assert recovery_time > 1\n", " self.recovery_time = recovery_time\n", " self.data = data\n", "\n", " def global_model(self):\n", " tau = self.recovery_time\n", " R0 = pyro.sample(\"R0\", dist.LogNormal(0., 1.))\n", " rho = pyro.sample(\"rho\", dist.Beta(100, 100))\n", " return R0, tau, rho\n", "\n", " def initialize(self, params):\n", " # Start with a single infection.\n", " return {\"S\": self.population - 1, \"I\": 1}\n", "\n", " def transition(self, params, state, t):\n", " R0, tau, rho = params\n", "\n", " # Sample flows between compartments.\n", " S2I = pyro.sample(\"S2I_{}\".format(t),\n", " infection_dist(individual_rate=R0 / tau,\n", " num_susceptible=state[\"S\"],\n", " num_infectious=state[\"I\"],\n", " population=self.population))\n", " I2R = pyro.sample(\"I2R_{}\".format(t),\n", " binomial_dist(state[\"I\"], 1 / tau))\n", "\n", " # Update compartments with flows.\n", " state[\"S\"] = state[\"S\"] - S2I\n", " state[\"I\"] = state[\"I\"] + S2I - I2R\n", "\n", " # Condition on observations.\n", " t_is_observed = isinstance(t, slice) or t < self.duration\n", " pyro.sample(\"obs_{}\".format(t),\n", " binomial_dist(S2I, rho),\n", " obs=self.data[t] if t_is_observed else None)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Note that we've stored data in the model. These models have a scikit-learn like interface: we instantiate a model class with data, then call a `.fit_*()` method to train, then call `.predict()` on a trained model.\n", "\n", "Note also that we've taken special care so that `t` can be either an integer or a `slice`. Under the hood, `t` is an integer during SMC initialization, a `slice` during SVI or MCMC inference, and an integer again during prediction." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Generating data \n", "\n", "To check that our model generates plausible data, we can create a model with empty data and call the model's [.generate()](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel.generate) method. This method first calls, `.global_model()`, then calls `.initialize()`, then calls `.transition()` once per time step (based on the length of our empty data." ] }, { "cell_type": "code", "execution_count": 3, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Simulated 4055.0 infections after 6 attempts\n" ] } ], "source": [ "population = 10000\n", "recovery_time = 10.\n", "empty_data = [None] * 90\n", "model = SimpleSIRModel(population, recovery_time, empty_data)\n", "\n", "# We'll repeatedly generate data until a desired number of infections is found.\n", "pyro.set_rng_seed(20200709)\n", "for attempt in range(100):\n", " synth_data = model.generate({\"R0\": 2.0})\n", " total_infections = synth_data[\"S2I\"].sum().item()\n", " if 4000 <= total_infections <= 6000:\n", " break\n", "print(\"Simulated {} infections after {} attempts\".format(total_infections, 1 + attempt))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "The generated data contains both global variables and time series, packed into tensors." ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "I.shape = (90,)\n", "I2R.shape = (90,)\n", "R0.shape = ()\n", "S.shape = (90,)\n", "S2I.shape = (90,)\n", "obs.shape = (90,)\n", "rho.shape = ()\n" ] } ], "source": [ "for key, value in sorted(synth_data.items()):\n", " print(\"{}.shape = {}\".format(key, tuple(value.shape)))" ] }, { "cell_type": "code", "execution_count": 5, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "plt.figure(figsize=(8,4))\n", "for name, value in sorted(synth_data.items()):\n", " if value.dim():\n", " plt.plot(value, label=name)\n", "plt.xlim(0, len(empty_data) - 1)\n", "plt.ylim(0.8, None)\n", "plt.xlabel(\"time step\")\n", "plt.ylabel(\"individuals\")\n", "plt.yscale(\"log\")\n", "plt.legend(loc=\"best\")\n", "plt.title(\"Synthetic time series\")\n", "plt.tight_layout()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Inference \n", "\n", "Next let's recover estimates of the latent variables given only observations `obs`. To do this we'll create a new model instance from the synthetic observations." ] }, { "cell_type": "code", "execution_count": 6, "metadata": {}, "outputs": [], "source": [ "obs = synth_data[\"obs\"]\n", "model = SimpleSIRModel(population, recovery_time, obs)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "The `CompartmentalModel` provides a number of inference algorithms. The cheapest and most scalable algorithm is SVI, avilable via the `.fit_svi()` method. This method returns a list of losses to help us diagnose convergence; the fitted parameters are stored in the model object." ] }, { "cell_type": "code", "execution_count": 7, "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "INFO \t Heuristic init: R0=1.83, rho=0.546\n", "INFO \t Running inference...\n", "INFO \t step 0 loss = 6.808\n", "INFO \t step 200 loss = 9.099\n", "INFO \t step 400 loss = 7.384\n", "INFO \t step 600 loss = 4.401\n", "INFO \t step 800 loss = 3.428\n", "INFO \t step 1000 loss = 3.242\n", "INFO \t step 1200 loss = 3.13\n", "INFO \t step 1400 loss = 3.016\n", "INFO \t step 1600 loss = 3.029\n", "INFO \t step 1800 loss = 3.05\n", "INFO \t step 2000 loss = 3.017\n", "INFO \t SVI took 12.7 seconds, 157.9 step/sec\n" ] }, { "name": "stdout", "output_type": "stream", "text": [ "CPU times: user 12.8 s, sys: 278 ms, total: 13.1 s\n", "Wall time: 13.2 s\n" ] } ], "source": [ "%%time\n", "losses = model.fit_svi(num_steps=101 if smoke_test else 2001,\n", " jit=True)" ] }, { "cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "plt.figure(figsize=(8, 3))\n", "plt.plot(losses)\n", "plt.xlabel(\"SVI step\")\n", "plt.ylabel(\"loss\")\n", "plt.ylim(min(losses), max(losses[50:]));" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "After inference, samples of latent variables are stored in the `.samples` attribute. These are primarily for internal use, and do not contain the full set of latent variables." ] }, { "cell_type": "code", "execution_count": 9, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "R0.shape = (100, 1)\n", "auxiliary.shape = (100, 1, 2, 90)\n", "rho.shape = (100, 1)\n" ] } ], "source": [ "for key, value in sorted(model.samples.items()):\n", " print(\"{}.shape = {}\".format(key, tuple(value.shape)))" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Prediction \n", "\n", "After inference we can both examine latent variables and forecast forward using the [.predict()](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel.predict) method. First let's simply predict latent variables." ] }, { "cell_type": "code", "execution_count": 10, "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "INFO \t Predicting latent variables for 90 time steps...\n" ] }, { "name": "stdout", "output_type": "stream", "text": [ "CPU times: user 113 ms, sys: 2.82 ms, total: 116 ms\n", "Wall time: 115 ms\n" ] } ], "source": [ "%%time\n", "samples = model.predict()" ] }, { "cell_type": "code", "execution_count": 11, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "I.shape = (100, 90)\n", "I2R.shape = (100, 90)\n", "R0.shape = (100, 1)\n", "S.shape = (100, 90)\n", "S2I.shape = (100, 90)\n", "auxiliary.shape = (100, 1, 2, 90)\n", "obs.shape = (100, 90)\n", "rho.shape = (100, 1)\n" ] } ], "source": [ "for key, value in sorted(samples.items()):\n", " print(\"{}.shape = {}\".format(key, tuple(value.shape)))" ] }, { "cell_type": "code", "execution_count": 12, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "names = [\"R0\", \"rho\"]\n", "fig, axes = plt.subplots(2, 1, figsize=(5, 5))\n", "axes[0].set_title(\"Posterior estimates of global parameters\")\n", "for ax, name in zip(axes, names):\n", " truth = synth_data[name]\n", " sns.distplot(samples[name], ax=ax, label=\"posterior\")\n", " ax.axvline(truth, color=\"k\", label=\"truth\")\n", " ax.set_xlabel(name)\n", " ax.set_yticks(())\n", " ax.legend(loc=\"best\")\n", "plt.tight_layout()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Notice that while the inference recovers the basic reproductive number `R0`, it poorly estimates the response rate `rho` and underestimates its uncertainty. While perfect inference would provide better uncertainty estimates, the response rate is known to be difficult to recover from data. Ideally the model can either incorporate a narrower prior, either obtained by testing a random sample of the population, or by more accurate observations, e.g. counting deaths rather than confirmed infections." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Forecasting \n", "\n", "We can forecast forward by passing a `forecast` argument to the `.predict()` method, specifying the number of time steps ahead we'd like to forecast. The returned `sample` will contain time values during both the first observed time interval (here 90 days) and the forecasted window (say 30 days)." ] }, { "cell_type": "code", "execution_count": 13, "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "INFO \t Predicting latent variables for 90 time steps...\n", "INFO \t Forecasting 30 steps ahead...\n" ] }, { "name": "stdout", "output_type": "stream", "text": [ "CPU times: user 2 µs, sys: 1 µs, total: 3 µs\n", "Wall time: 5.96 µs\n" ] } ], "source": [ "%time\n", "samples = model.predict(forecast=30)" ] }, { "cell_type": "code", "execution_count": 14, "metadata": {}, "outputs": [ { "data": { "image/png": "iVBORw0KGgoAAAANSUhEUgAAAxYAAAGGCAYAAADmRxfNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAAPYQAAD2EBqD+naQAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAgAElEQVR4nOzdd3gUVffA8e+md0IgkEIgIZRQQqiC1NCLggjKiyJKVwQRFVCwUCMggooIGF+kifIqiPwUMAgSmkgPvcaElkaAEEhIn98f112yaWwaoZzP8+QhOzN7584EyJy999yj0zRNQwghhBBCCCGKwaysOyCEEEIIIYR4+ElgIYQQQgghhCg2CSyEEEIIIYQQxSaBhRBCCCGEEKLYJLAQQgghhBBCFJsEFkIIIYQQQohik8BCCCGEEEIIUWwSWAghhBBCCCGKTQILIYQQQgghRLFJYCGEeGQEBgYSGBhY5PevXLkSPz8/LC0tcXZ2LrmOmcDb25tBgwbd13OaatCgQXh7e5d1N+670rjuZcuWodPpiIyMLNF2S9qcOXOoXr065ubmNGzYsKy7I4R4SEhgIYS4J/3DkI2NDVeuXMm1PzAwkPr165dBz0rO6dOnGTRoEL6+vnzzzTcEBweX+Dn++usvpkyZQkJCQom3LR4sH3/8Mb/88ktZd6NINm/ezIQJE2jVqhVLly7l448/zvfYM2fO8NZbb9GyZUtsbGzuGTT93//9H40bN8bGxoaqVasyefJkMjIych2XkJDAiBEjcHV1xd7envbt23Po0KH71qYQoog0IYS4h6VLl2qABmijR4/Otb9du3ZavXr1yqBnxlJTU7XU1NQivXfRokUaoJ07d66Ee3XXnDlzNECLiIjItS8lJUVLS0srtXMXR1pampaSklLW3bjvXnnlFa1atWpFeq+9vb32yiuv5NqekZGh3blzR8vKyipe50rRu+++q5mZmZn0b2np0qWamZmZVr9+fa1hw4b5/v3WNE3buHGjptPptPbt22vBwcHaG2+8oZmZmWmvvfaa0XGZmZlay5YtNXt7e23KlCnaggULtLp162qOjo7a2bNnS71NIUTRWZRpVCOEeKg0bNiQb775hokTJ+Lh4VHW3cnFysqqyO+Ni4sDuO9ToPSsra3L5LymsLS0LOsuPDLMzc0xNzcv624UKC4uDltbW5P+PfXq1YuEhAQcHR359NNPCQsLy/fYcePG0aBBAzZv3oyFhXr8cHJy4uOPP+bNN9/Ez88PgDVr1vDXX3/x008/8dxzzwHQr18/atWqxeTJk/n+++9LtU0hRNHJVCghhMkmTZpEZmYms2bNMun47777jiZNmmBra4uLiwv9+/fn0qVLhv3z58/H3NzcaGrQ3Llz0el0vP3224ZtmZmZODo68u677xZ4vpw5FqGhoeh0On788UeCgoKoUqUKNjY2dOzYkfPnzxuO8/b2ZvLkyQC4urqi0+mYMmWKYf+mTZto06YN9vb2ODo68tRTT3HixIlc5z99+jT9+vXD1dUVW1tbateuzfvvvw/AlClTGD9+PAA+Pj7odDqjaSN55Vj8888/PP/887i4uGBnZ0eLFi3YsGGD0TGmXiPAuXPn6Nu3L25ubtjY2FClShX69+/PzZs3C7yvOXMNIiMj0el0fPrppwQHB+Pr64u1tTXNmjVj//79BbYFd6fW7dixg1dffZUKFSrg5OTEyy+/zI0bN3Idv3DhQurVq4e1tTUeHh6MGjUq13Qy/XS8gwcP0rJlS2xtbfHx8WHx4sV5njvndB39fQwNDS2w759++iktW7akQoUK2Nra0qRJE9asWWN0jE6nIykpieXLlxt+zvqfbX7nL8w1njx5kvbt22NnZ4enpyeffPJJgX3Wy8jIYPr06Yafl7e3N5MmTSI1NdWo70uXLiUpKcnQ92XLluXbpouLC46Ojvc898mTJzl58iQjRowwBAAAr7/+OpqmGd3DNWvWULlyZfr06WPY5urqSr9+/Vi/fr2hv6XRphCieCSwEEKYzMfHh5dffplvvvmGqKioAo8NCgri5ZdfpmbNmsybN4+xY8eydetW2rZta3hgatOmDVlZWezatcvwvp07d2JmZsbOnTsN2w4fPszt27dp27Ztkfo9a9Ys1q1bx7hx45g4cSJ///03AwYMMOz//PPPefbZZwFYtGgRK1euNDyArFy5kqeeegoHBwdmz57Nhx9+yMmTJ2ndurXRw+HRo0dp3rw5f/75J8OHD+eLL76gd+/e/PrrrwD06dOHF154AYDPPvuMlStXsnLlSlxdXfPsc2xsLC1btiQkJITXX3+doKAgUlJS6NWrF+vWrSv0NaalpdG1a1f+/vtv3njjDb766itGjBjBP//8U+Scj++//545c+bw6quvMmPGDCIjI+nTpw/p6ekmvX/06NGcOnWKKVOm8PLLL7Nq1Sp69+6NpmmGY6ZMmcKoUaPw8PBg7ty59O3bl6+//pouXbrkOs+NGzfo0aMHTZo04ZNPPqFKlSqMHDmSb7/9tkjXl5cvvviCRo0aMW3aND7++GMsLCx4/vnnjQK+lStXYm1tTZs2bQw/51dffTXfNgt7jd26dSMgIIC5c+fi5+fHu+++y6ZNm+7Z92HDhvHRRx/RuHFjPvvsM9q1a8fMmTPp37+/Ud/btGmDtbW1oe9F/XeX3eHDhwFo2rSp0XYPDw+qVKli2K8/tnHjxpiZGT+iPPHEEyQnJ3P27NlSa1MIUUxlPBVLCPEQ0OdY7N+/XwsPD9csLCy0MWPGGPbnzLGIjIzUzM3NtaCgIKN2jh07pllYWBi2Z2Zmak5OTtqECRM0TdO0rKwsrUKFCtrzzz+vmZuba7du3dI0TdPmzZunmZmZaTdu3Ciwn+3atdPatWtneL1t2zYN0OrUqWM0X/yLL77QAO3YsWOGbZMnT9YA7erVq4Ztt27d0pydnbXhw4cbnScmJkYrV66c0fa2bdtqjo6O2oULF4yOzT6XvqAci2rVqhnNyR87dqwGaDt37jTqj4+Pj+bt7a1lZmYW6hoPHz6sAdpPP/2U570rSM5cg4iICA3QKlSooF2/ft2wff369Rqg/frrrwW2p//71KRJE6O8kk8++UQDtPXr12uapmlxcXGalZWV1qVLF8P1apqmLViwQAO0b7/91rCtXbt2GqDNnTvXsC01NVVr2LChVqlSJcN59OfO+TPQ38dt27ble92apmnJyclGr9PS0rT69etrHTp0MNqeX45FzvMX5RpXrFhhdI1ubm5a3759c50ru7CwMA3Qhg0bZrR93LhxGqD9+eefRtdtb29fYHt5Kejvt37fxYsXc+1r1qyZ1qJFC8Nre3t7bciQIbmO27BhgwZov//+e6m1KYQoHhmxEEIUSvXq1Rk4cCDBwcFER0fneczPP/9MVlYW/fr1Iz4+3vDl5uZGzZo12bZtGwBmZma0bNmSHTt2AHDq1CmuXbvGe++9h6Zp7NmzB1CjGPXr1y9y/sPgwYON5ou3adMGUFONCvLHH3+QkJDACy+8YHQd5ubmNG/e3HAdV69eZceOHQwZMoSqVasataHT6YrU540bN/LEE0/QunVrwzYHBwdGjBhBZGQkJ0+eLNQ1litXDoCQkBCSk5OL1Kec/vOf/1C+fPl8z3kvI0aMMMrfGDlyJBYWFmzcuBGALVu2kJaWxtixY40+aR4+fDhOTk65poVZWFgYjQxYWVnx6quvEhcXx8GDBwt/gXmwtbU1fH/jxg1u3rxJmzZtiry6UGGv0cHBgZdeesnw2srKiieeeOKe91x/T7NPMQR45513AHKdp6TduXMHyDuXyMbGxrBff2x+x2VvqzTaFEIUjwQWQohC++CDD8jIyMg31+LcuXNomkbNmjVxdXU1+jp16pQhURrUw+jBgwe5c+cOO3fuxN3dncaNGxMQEGCYDrVr1y7DQ2tR5HzY1z8M5zWfP+d1AHTo0CHXdWzevNlwHfqHupJccvfChQvUrl071/Y6deoY9md3r2v08fHh7bff5r///S8VK1aka9eufPXVV/fMryhIUe+rXs2aNY1eOzg44O7ubphipr/GnPfBysqK6tWr57oHHh4e2NvbG22rVasWQInVjfjtt99o0aIFNjY2uLi44OrqyqJFi4p8Hwt7jVWqVMkVrJYvX/6e9/zChQuYmZlRo0YNo+1ubm44OzvnOk9J0wdkeeUypKSkGAVstra2+R6Xva3SaFMIUTyyKpQQotCqV6/OSy+9RHBwMO+9916u/VlZWeh0OjZt2pTnCjgODg6G71u3bk16ejp79uxh586dhgCiTZs27Ny5k9OnT3P16tViBRb5rcKjZZvLn5esrCxAzTt3c3PLtT97wmhZM+Ua586dy6BBg1i/fj2bN29mzJgxzJw5k7///psqVaqUyjkfNPmNIGVmZt7zvTt37qRXr160bduWhQsX4u7ujqWlJUuXLr1vqwoV954XdQStuNzd3QGIjo7Gy8vLaF90dDRPPPGE0bF5jYbqt+lXpCuNNoUQxSMjFkKIItGPWsyePTvXPl9fXzRNw8fHh06dOuX6atGiheHYJ554AisrK3bu3GkUWLRt25a9e/eydetWw+v7zdfXF4BKlSrleR36FaiqV68OwPHjxwtsrzAPddWqVePMmTO5tp8+fdqwvyj8/f354IMP2LFjBzt37uTKlSu5Vk66X/QjQnq3b98mOjrasAKV/hpz3oe0tDQiIiJy3YOoqCiSkpKMtumTcvVt6kdVciasm/KJ/dq1a7GxsSEkJIQhQ4bQvXt3OnXqlOexpv6sC3uNRVWtWjWysrJy3fPY2FgSEhJK7Dz50VfvPnDggNH2qKgoLl++bFTdu2HDhhw6dMgQ2Ovt3bsXOzs7wyhUabQphCgeCSyEEEXi6+vLSy+9xNdff01MTIzRvj59+mBubs7UqVNzfZKqaRrXrl0zvLaxsaFZs2b88MMPXLx40WjE4s6dO8yfPx9fX1/Dp5P3U9euXQ1r4ue10tHVq1cBtWxl27Zt+fbbb7l48aLRMdmvXz9Nx5RVmHr06MG+ffsMeSYASUlJBAcH4+3tTd26dQt1LYmJibmqEfv7+2NmZlZmS20GBwcb3ddFixaRkZFB9+7dAejUqRNWVlbMnz/f6D4uWbKEmzdv8tRTTxm1l5GRwddff214nZaWxtdff42rqytNmjQB7gaL+rweUKMVplRaNzc3R6fTGY1uREZG5llh297e3qSfc2Gvsah69OgBqBXQsps3bx5AiZ0nP/Xq1cPPz4/g4GCj+7do0SJ0Op2htgTAc889R2xsLD///LNhW3x8PD/99BM9e/Y05EqURptCiOJ5cMbxhRAPnffff5+VK1dy5swZ6tWrZ9ju6+vLjBkzmDhxIpGRkfTu3RtHR0ciIiJYt24dI0aMYNy4cYbj27Rpw6xZsyhXrhz+/v6AGiWoXbs2Z86cyVXf4X5xcnJi0aJFDBw4kMaNG9O/f39cXV25ePEiGzZsoFWrVixYsABQNTlat25N48aNGTFiBD4+PkRGRrJhwwZD0TD9w+37779P//79sbS0pGfPnrnyAgDee+89fvjhB7p3786YMWNwcXFh+fLlREREsHbt2lzLZt7Ln3/+yejRo3n++eepVasWGRkZrFy5EnNzc/r27VvMO1U0aWlpdOzYkX79+nHmzBkWLlxI69at6dWrF6ACtokTJzJ16lS6detGr169DMc1a9bMKIkZ1HSW2bNnExkZSa1atfjf//5HWFgYwcHBhiTxevXq0aJFCyZOnMj169dxcXFh9erVuYKuvDz11FPMmzePbt268eKLLxIXF8dXX31FjRo1OHr0qNGxTZo0YcuWLcybNw8PDw98fHxo3rx5rjYLe41FFRAQwCuvvEJwcDAJCQm0a9eOffv2sXz5cnr37k379u2L1O7Nmzf58ssvAdi9ezcACxYswNnZGWdnZ0aPHm04ds6cOfTq1YsuXbrQv39/jh8/zoIFCxg2bJghdwhUENCiRQsGDx7MyZMnqVixIgsXLiQzM5OpU6canb802hRCFEPZLEYlhHiYZF9uNqdXXnlFA4yWm9Vbu3at1rp1a83e3l6zt7fX/Pz8tFGjRmlnzpwxOk6/5GP37t2Ntg8bNkwDtCVLlpjUz/yWm825xKp+udSlS5catuW13Gz2drp27aqVK1dOs7Gx0Xx9fbVBgwZpBw4cMDru+PHj2rPPPqs5OztrNjY2Wu3atbUPP/zQ6Jjp06drnp6empmZmdHSnDmXm9U0TQsPD9eee+45Q3tPPPGE9ttvv+XqmynX+M8//2hDhgzRfH19NRsbG83FxUVr3769tmXLlvxup0F+y83OmTMn17GANnny5ALb0/992r59uzZixAitfPnymoODgzZgwADt2rVruY5fsGCB5ufnp1laWmqVK1fWRo4cmWvpYf2SxwcOHNCefPJJzcbGRqtWrZq2YMGCXO2Fh4drnTp10qytrbXKlStrkyZN0v744w+TlptdsmSJVrNmTc3a2lrz8/PTli5davi7k93p06e1tm3bara2thpg+Nnmt9xtYa4xp7z6mZf09HRt6tSpmo+Pj2Zpaal5eXlpEydO1FJSUnK1Z+pys/q/C3l95dWndevWaQ0bNtSsra21KlWqaB988IHRksN6169f14YOHapVqFBBs7Oz09q1a5fn/z+l1aYQomh0mvYAZ9kJIYR45CxbtozBgwezf//+XMXNiiowMJD4+Ph75rkIIYQoPZJjIYQQQgghhCg2CSyEEEIIIYQQxSaBhRBCCCGEEKLYJMdCCCGEEEIIUWwyYiGEEEIIIYQoNgkshBBCCCGEEMUmBfKArKwsoqKicHR0RKfTlXV3hBBCCCGEeCBomsatW7fw8PC4Z3FWCSyAqKgovLy8yrobQgghhBBCPJAuXbpElSpVCjxGAgvA0dERUDfMycmpjHsjhBBCiEdFTEwMS5cuZfDgwbi5uZV1d4QotMTERLy8vAzPywWRwAIM05+cnJwksBBCCCFEiUlKSsLGxgZHR0d5xhAPNVPSBSR5WwghhBBCCFFsElgIIYQQQgghik0CCyGEEEIIIUSxSY6FibKyskhLSyvrbogisLS0xNzcvKy7IYQQQgjxSJPAwgRpaWlERESQlZVV1l0RReTs7Iybm5vUKRFCCCGEKCUSWNyDpmlER0djbm6Ol5fXPQuDiAeLpmkkJycTFxcHgLu7exn3SAghhBDi0SSBxT1kZGSQnJyMh4cHdnZ2Zd0dUQS2trYAxMXFUalSJZkWJYQQQghRCuTj93vIzMwEwMrKqox7IopDHxSmp6eXcU+EEEIIIR5NEliYSObmP9zk5yeEEEIIUboksBBCCCGEEEIUmwQWolgCAwMZO3as4bW3tzeff/55GfZICCHE4yYlJaVEVm7MyMiQKbNCFIMEFqJE7d+/nxEjRpR1N4QQQjwm4uLi8PX1pXPnzsVq59atW1SrVo0OHTqgaVoJ9U6Ix4usCiVKlKura1l3QQghxGNkzpw5REVFERUVRXp6OpaWlkVq58CBA4Z2zp07R61atUq4p0I8+mTE4hEVGBjIG2+8wdixYylfvjyVK1fmm2++ISkpicGDB+Po6EiNGjXYtGmT4T3Hjx+ne/fuODg4ULlyZQYOHEh8fLxhf1JSEi+//DIODg64u7szd+7cXOfNORVq3rx5+Pv7Y29vj5eXF6+//jq3b9827F+2bBnOzs6EhIRQp04dHBwc6NatG9HR0aV0Z4QQQjwqNE1j/fr1ALRt27ZY06EiIiIA8Pf3l6BCiCKSwKKwNA2Sksrmq5BDs8uXL6dixYrs27ePN954g5EjR/L888/TsmVLDh06RJcuXRg4cCDJyckkJCTQoUMHGjVqxIEDB/j999+JjY2lX79+hvbGjx/P9u3bWb9+PZs3byY0NJRDhw4V2AczMzPmz5/PiRMnWL58OX/++ScTJkwwOiY5OZlPP/2UlStXsmPHDi5evMi4ceMKda1CCCEeP7t37+bcuXPY29vz22+/YW1tXeS2zp8/D0CbNm1KqntCPHZkKlRhJSeDg0PZnPv2bbC3N/nwgIAAPvjgAwAmTpzIrFmzqFixIsOHDwfgo48+YtGiRRw9epQtW7bQqFEjPv74Y8P7v/32W7y8vDh79iweHh4sWbKE7777jo4dOwIqcKlSpUqBfciZ2D1jxgxee+01Fi5caNienp7O4sWL8fX1BWD06NFMmzbN5OsUQgjxeFqyZAkA/fr1w9HRsVhtnTt3DoCaNWsWu19CPK4ksHiENWjQwPC9ubk5FSpUwN/f37CtcuXKgEp8O3LkCNu2bcMhj6ApPDycO3fukJaWRvPmzQ3bXVxcqF27doF92LJlCzNnzuT06dMkJiaSkZFBSkoKycnJhqJ1dnZ2hqACwN3dnbi4uKJdtBBCiMdCYmIiP/74IwCdO3dm06ZNeHp6Gv3uKwx9YHH27Fk6duxIly5dePfdd0usv0I8DiSwKCw7OzVyUFbnLoScCWw6nc5om75oXFZWFrdv36Znz57Mnj07Vzvu7u6GIeLCiIyM5Omnn2bkyJEEBQXh4uLCrl27GDp0KGlpaYbAIq9+yoocQgghCpKUlET//v05evQohw8fZs6cObz55ptFWvJc0zTD7zkHBwf+/PNPbty4IYGFEIUkgUVh6XSFmo70sGjcuDFr167F29sbC4vcfy18fX2xtLRk7969VK1aFYAbN25w9uxZ2rVrl2ebBw8eJCsri7lz52JmptJ59J8uCSGEEMXh7u7OkiVL0DSNBQsWAHD58uUit7dv3z7Onz9P06ZNmTt3LocPHyYiIgIfH5+S6rIQjzxJ3hYAjBo1iuvXr/PCCy+wf/9+wsPDCQkJYfDgwWRmZuLg4MDQoUMZP348f/75J8ePH2fQoEGGgCEvNWrUID09nS+//JJ//vmHlStXsnjx4vt4VUIIIR51Op0OLy8vAC5dulTkNurWrUuvXr3w8PAwfGD2888/l1g/hXgcSGAhAPDw8GD37t1kZmbSpUsX/P39GTt2LM7OzobgYc6cObRp04aePXvSqVMnWrduTZMmTfJtMyAggHnz5jF79mzq16/PqlWrmDlz5v26JCGEEI+otWvXcuDAAcO0Wf1CIsUZsciub9++hvMIIUyn02QyO4mJiZQrV46bN2/i5ORktC8lJcUwFGpjY1NGPRTFJT9HIYR4NKSlpeHp6Ul8fDxbt26lQ4cOxMTE4O7ujk6nIzU1tdBF8tauXcvZs2fp1q0bjRo1IioqCk9PTwCuXLmCh4dHkfsbHR1NcHAwI0aMwN3dvcjtCFFWCnpOzklGLIQQQgjx0Pi///s/4uPj8fDwoG3btgBUqlQJS0tLNE0jJiam0G3+8MMPTJo0ie3btwNqFP/JJ58EYN26dSXXeSEecRJYCCGEEOKh8d///heAQYMGGRYbMTMzM4wwFCXPIq8aFv3796dr166SvC1EIciqUEIIIYR4KFy8eJHNmzcDMGTIEKN906ZNQ6fTGdVFMkX2pWazBxZjxoxhzJgxxeyxEI8XCSyEEEII8VBYtmwZmqYRGBiYK4AYOHBgkdqMjo4mOTkZc3NzvL29S6CXQjy+ynQq1I4dO+jZsyceHh7odDp++eUXo/06nS7Przlz5hiO8fb2zrV/1qxZ9/tShBBCCFGKsrKyWLp0KQBDhw4tsXb106CqVauGlZVVrv2XL1/O9XwihMhbmY5YJCUlERAQwJAhQ+jTp0+u/dHR0UavN23axNChQw3LwOlNmzaN4cOHG147OjqWToeFEEIIUSZiYmKwtrbG2dk513MAwNWrVzlw4ACWlpZ06tTJ5Hbzyq/Qi4yMxMfHB0tLS+Li4nB2di76BQjxGCjTwKJ79+5079493/1ubm5Gr9evX0/79u2pXr260XZHR8dcxwohhBDi0eHh4UFYWBiHDx/G1tY21/5du3bRp08fmjdvXmKBhbe3N3Xq1OHUqVNs2LCBAQMGFP0C9JKSwN7etGM1DZKTTT9eiDL20KwKFRsby4YNG/Ic/pw1axYVKlSgUaNGzJkzh4yMjALbSk1NJTEx0ehLCCGEEA82GxsbwzKwORW1SN7kyZM5evQoY8eOzXN/iRbLi42FdesgNBTi4/M/TtPgyhUICVHHX7tW/HMLcR88NMnby5cvx9HRMdeUqTFjxtC4cWNcXFz466+/mDhxItHR0cybNy/ftmbOnMnUqVNLu8tCCCGEKKYff/yRixcvMnbsWMPysnnRBxbR0dFkZGQUeGx2dnZ2+Pv757v/2WefZcaMGWzevJmsrCzMzIrxmezVqxAVBTExcOYM1KunvsqXv3tMTAwcOwZnz0JqKmRkwNGjEBgIOl3Rzy3EffDQBBbffvstAwYMyFU1+e233zZ836BBA6ysrHj11VeZOXMm1tbWebY1ceJEo/clJibi5eVVOh1/QIWGhtK+fXtu3LjxSMwZfdSuRwghBMTHxzNq1Cji4+NxdHTk1VdfzffYypUrY2FhQUZGBtHR0SX2e93f3x9LS0uSkpK4fPkyVatWLXpjV66AjQ34+MCNG/D333DqFNSvD15ecO6cen3nDri7g6MjJCbCyZNQqxb8W6tDiAfVQzEVaufOnZw5c4Zhw4bd89jmzZuTkZFBZGRkvsdYW1vj5ORk9CWEEEKIB8vYsWOJj4/H39+fwYMHF3hs9iJ5pk6Hio2NZfjw4Xz22Wf5HmNpaWlY2vbMmTMm9jwPaWkqsNA/c5Qvr4IFS0vYuVNNeTp4EMqVU9v1C9E4Oan3Hj4MWVlFP78Q98FDEVgsWbKEJk2aEBAQcM9jw8LCMDMzo1KlSvehZ6Iw0tLSyroLQgghHhIbNmxg1apVmJmZsWTJkjyXgs2psHkWp06d4r///S8LFy4s8Dg/Pz8ATp8+bVK7eUpIUKMP2T/M1OmgQgXw8wM3N6hd23i/nqcnnD8PERFFP78Q90GZBha3b98mLCyMsLAwACIiIggLC+PixYuGYxITE/npp5/yHK3Ys2cPn3/+OUeOHOGff/5h1apVvPXWW7z00kuUzz5f8QFx+fJltm3bVujEsqJKTU1lzJgxVKpUCRsbG1q3bs3+/fuNjtm9ezcNGjTAxsaGFi1acMM5n64AACAASURBVPz4ccO+Cxcu0LNnT8qXL4+9vT316tVj48aNhv3Hjx+ne/fuODg4ULlyZQYOHEh8tmS0wMBARo8ezdixY6lYsSJdu3blxRdf5D//+Y9RH9LT06lYsSIrVqwA1FrlM2fOxMfHB1tbWwICAlizZo3RezZu3EitWrWwtbWlffv2BY5QCSGEeLjcunWLkSNHAmrUolmzZia9Tz/96dKlSyYdr18RqkaNGgUe99Zbb7F+/Xqee+45k9rNU0KCypewtMy9T6dTU6TyY2sL5uZw6JAavRDiAVWmgcWBAwdo1KgRjRo1AlS+RKNGjfjoo48Mx6xevRpN03jhhRdyvd/a2prVq1fTrl076tWrR1BQEG+99RbBwcH37RpMtWTJEqpVq0aHDh2oVq0aS5YsKfVzTpgwgbVr17J8+XIOHTpEjRo16Nq1K9evXzccM378eObOncv+/ftxdXWlZ8+epKenAzBq1ChSU1PZsWMHx44dY/bs2Tg4OACQkJBAhw4daNSoEQcOHOD3338nNjaWfv36GfVh+fLlWFlZsXv3bhYvXsyAAQP49ddfuX37tuGYkJAQkpOTefbZZwGVXL9ixQoWL17MiRMnDMHi9u3bAfULo0+fPvTs2ZOwsDCGDRvGe++9V6r3UgghxP0zceJELl26hI+PD9OmTTP5fcOHD2flypX07NnTpOMLWmo2u7Zt29KrVy/c3d1N7ksusbEFBw/34ukJFy6oPAwhHlA6TdO0su5EWUtMTKRcuXLcvHkzV75FSkoKERER+Pj45EocN9Xly5epVq0aWdnmRpqbmxMZGWkYti1pSUlJlC9fnmXLlvHiiy8CamTA29vb8OlP+/btWb16tWEE4fr161SpUoVly5bRr18/GjRoQN++fZk8eXKu9mfMmMHOnTsJCQkxuk4vLy/OnDlDrVq1CAwMJDExkUOHDhmOycjIwN3dnXnz5jFw4EAAXnzxRbKysli9ejWpqam4uLiwZcsWoyUFhw0bRnJyMt9//z2TJk1i/fr1nDhxwrD/vffeY/bs2fkmb5fEz1EIIUTp0xelA9iyZQsdO3YstXP16dOHdevWMX/+fN54441SOUd0dDTBwcGMqFQJd3t74xWgCisqCuzsoE8fqW0h7puCnpNzeihyLB52586dMwoqADIzMzl//nypnTM8PJz09HRatWpl2GZpackTTzzBqVOnDNuyP7y7uLhQu3Ztw/4xY8YwY8YMWrVqZVjnW+/IkSNs27YNBwcHw5d+Dmp4eLjhuCZNmhj1y8LCgn79+rFq1SpABUDr1683FB06f/48ycnJdO7c2ajtFStWGNo9deoUzZs3N2o3v3XNhRBCPFyqVKnCvn37WLlyZakGFWD6iIWmaaxdu5agoCCjEfdCSU6+m5BdVJUrQ3Q0ZPtgTYgHyUOz3OzDrGbNmpiZmeUasbjXnM6yNmzYMLp27cqGDRvYvHkzM2fOZO7cubzxxhvcvn2bnj17Mnv27Fzvyz5UbJ/HJyoDBgygXbt2xMXF8ccff2Bra0u3bt0ADP9hb9iwwbC6h15+ywcLIYR4dFhYWNCsWTOT8yqyS05OJjQ0lGvXrhlGxfOTlZVl+IDvXr+PdTodr7/+OnFxcXTr1i3Xh2YmycwEE2tr5MvcHFxdVV2LmjWLN/ohRCmQEYv7oEqVKgQHB2Nubg6ooOLrr78utWlQAL6+vobcBr309HT2799P3bp1Ddv+/vtvw/c3btzg7Nmz1KlTx7DNy8uL1157jZ9//pl33nmHb775BoDGjRtz4sQJvL29qVGjhtFXXsFEdi1btsTLy4v//e9/rFq1iueffx7Lf5PZ6tati7W1NRcvXszVrj4pr06dOuzbt8+ozezXIYQQ4vGUkJDAU089xeDBg8nIyCjw2OjoaFJSUrCwsMDb2/uebdeuXRu495KzCQkJbNy4kbi4OOMdJfXhWIUKqgbGkSOqQrcQDxAJLO6ToUOHEhkZybZt24iMjGTo0KGlej57e3tGjhzJ+PHj+f333zl58iTDhw8nOTnZ6NzTpk1j69atHD9+nEGDBlGxYkV69+4NqJU4QkJCiIiI4NChQ2zbts0QdIwaNYrr16/zwgsvsH//fsLDwwkJCWHw4MFkZmbes38vvvgiixcv5o8//jBMgwJwdHRk3LhxvPXWWyxfvpzw8HAOHTrEl19+yfLlywF47bXXOHfuHOPHj+fMmTN8//33LFu2rATvnhBCiLIQFRXFsGHDDNNlC0tfJC8zM5OYmJgCj/X09OTWrVscOXLEpCrdpi45u2vXLp566ikCAwPVhuRk9ee/i58Um04HHh6qrsXu3ao6txAPCAks7qMqVaoQGBhYqiMV2c2aNYu+ffsycOBAGjduzPnz5wkJCTFainfWrFm8+eabNGnShJiYGH799VfDWuGZmZmMGjWKOnXq0K1bN2rVqmVY69vDw4Pdu3eTmZlJly5d8Pf3Z+zYsTg7O2Nmdu+/VgMGDODkyZN4enoa5YEATJ8+nQ8//JCZM2cazr1hwwZDMl/VqlVZu3Ytv/zyCwEBASxevJiPP/64pG6bEEKIMrJ9+3aWLFlSYMG6gpibm+Ph4QGYVsvCwcHBaBS/IKYGFgcOHAC4O5UrIUH9aWtr0nlM4uQElSrBnj2webMawchPZiaEh8O2bWoKVVSUBCOi1EiOxSPMxsaG+fPnM3/+/Fz7AgMD0S8I9vTTT+f5/i+//LLA9mvWrMnPP/+c7/7Q0NB899WpU4f8FiTT6XS8+eabvPnmm/m+/+mnn87V73tVZRVCCPFg27FjB6CWdy2qKlWqcPHixRKvGWXqVCh9YNG0aVO1Qf/Q/+906BLj5KSWrz1zRgUvrVvDvx/AAapK94ULaspURISaNpWVpepoODlBlSpq5KNSJZW3IUQJkMBCCCGEEA+EkggsTC2SN2XKFGJiYnj11VcN9bQKoh+xOHv2LJmZmYa8yew0TTMesdA0NUJQWqysVBL35cuwcSO0aAENGqiVo44eVdW6zczAy+tujkdaGty6BadOqaDDxQWefloFGEIUkwQWQgghhChzV69e5eTJkwC0bt26yO3opxvfa8RizZo1nDhxwpBXeC/e3t5YWVmRkpLCxYsXDdNzs7t8+TKxsbGYm5sTEBCgHuATEwt/EYVhZgZVq8K1a2q60/nzqhhfVpYakcg5BcvKSiWAV6igXp89C2Fh0Lmzyt8QohgksBBCCCFEmdu1axcA9erVo2LFikVux5TAIisry1Ab6V41LPTMzc359ddf8fT0NIyK5KQfrahfvz62trZq5ECfvF3aKlRQxfPi4lS9Czs7097n4QGnT4OfnwpQhCgGCSyEEEIIUeZKYhoUQLdu3XB1dS0wKfvKlSuGpWarVatmcttdunQpcH+tWrWYNGkSrvqchdjY+zsKYGsLhbgeQK1WFRurRi08PUs+F0Q8ViSwEEIIIUSZ0+dEFDew8PPzM+RD5EdfcdvHx8ekpWZNVa9ePYKCgtSLrCy4eFE97N+vUYuiqlJFTaEKD4datcq6N+IhJoGFEEIIIcrcmjVriIuLw87UKTzFoA8sTJ0GpXfp0iV++OEHNE3j3XffLfjgmzfh+nWwt1f5Dw8ya2v1dfCgmg5lY1PWPRIPKaljIYQQQogHQqVKlXAogUJyv//+O8HBwSToa0jkcP78eaDwgUVMTAzvvvsun3/+ea598fHxbNq0iatXr+o3QFJSydavKE0eHmp1qXsspytEQSSwEEIIIUSZyq+uUVENHz6cV199Nd+aE7GxsQDUqFGjUO3qa1nExMTkClq2bdtGjx496N69u/4kasWmh2WlJQsLcHZWFb1v3Srr3oiHlAQWQgghhChTXbt2pXPnzhw5cqRE2tOv2pTfylArVqzg5s2bvPLKK4Vq18nJCXd3dyB3obz9+/cD/9av0OdXODoWtutlq1IluHoVjh8v656Ih5QEFqJIQkND0el0+Q4zCyGEEKZITk4mNDSULVu24FhCD+L6JWcLKpLn5OSU+3wJCbB5M6Sm5vs+fWJ4zsDCqDBedLSquO3kVJTulx0zMxVcHD2qpnIJUUgSWDyiAgMDGTt27APXlhBCCJHd3r17SU9Px9PTM8+ic0VRUC2LpKSk/KdeXbkC586pXIN86AOL06dPG7ZlZWVx8OBBAJrWqwe7d0NGhum1JB4kLi5qKlRYmKocLkQhSGDxmNI0jYyMjLLuhhBCiMdc9voVuhLKRyhoKtTo0aNp2rQpf/31V+43RkSo3Ih/V43Kiz7PIvuIxblz50hMTMTGxoa616/DhQuFryfxIPH0hBMnVFVuIQpBAotH0KBBg9i+fTtffPEFOp0OnU7HsmXL0Ol0bNq0iSZNmmBtbc2uXbsYNGgQvXv3Nnr/2LFjCQwMzLetyMhIw7EHDx6kadOm2NnZ0bJly3wT5YQQQoi8lFRhvOzymwoVGRnJypUrOXToEOY5C8Hdvq2mMFWurAKD69fzbDuvEQv9NKhGtWphcfKkWrL1YS405+CglpzdtUsFWkKYSAKLIkpKSsr3KyUlxeRj79y5Y9KxhfHFF1/w5JNPMnz4cKKjo4mOjjZ8evPee+8xa9YsTp06RYMGDYrVFsD777/P3LlzOXDgABYWFgwZMqRQfRVCCPH4SktLY8+ePUDpBBY5Ryxmz55NZmYmnTt3pnnz5sZvio1VU4A8PSExUSVf56Fly5YcPnzYkKwNdwOLpuXLQ4UKD+cUqJzc3VUtjp07H/wCf+KBIQXyiqigdbZ79OjBhg0bDK8rVapEcj7/KNu1a0doaKjhtbe3N/F5JEwVZim+cuXKYWVlhZ2dHW5ubsDdT1amTZtG586di9VWdkFBQbRr1w5QQctTTz1FSkoKNlJcRwghxD0cPHiQO3fuULFiRerUqVNi7dapU4fvvvuOqlWrGrZduXKFb7/9FoAPPvgg95uuXFF/mpurpOtTp6B+fbUMazaOjo40bNjQaNvrAwbQICmJuk5OKrB4FOh04OOjpoXt2QPt26vkbiEKIIHFY6Zp06Yl2l72UQ/9EnxxcXFG/5kLIYQQ+enWrRuVKlUyLb8iKQlOnoRGjXI98Gfn7OzMgAEDjLZ9+umnpKWl0aZNm9yjI+npEBkJ5cqp1xUrwqVLEBWlpjUVJC2NmtHR1KxVCwpZcO+BZ2GhckXCwlTAlCOgEiInCSyK6Pbt2/nuyzlvMy4uLt9jzXJE/9nzF0qDvb19rvPnHA1JT083uT1LS0vD9/pfCllZWcXooRBCiMfFk08+yaZNm0x/Q3S0esh1cQFfX5PfFhcXx9dffw3kM1oRF6eWh/13ChVWVqoWxT//5BlYbNy4kfXr19OhfXv+4+UFp09D9eqP5if6dnZQvrwatXBxuXegJR5rElgUUc4H9LI4tiBWVlZkZmbe8zhXV1eO5yiEExYWZhQwmNqWEEIIUaquXlVLwR47Bt7eBSZI//XXXxw7dozWrVuzadMm7ty5Q7NmzfKeDhwTo5aHtbK6u61CBTh/Hpo0yVXobt++fQQHB5N57Rqenp6EJSQQaGtL/Yd5JaiCuLqqFbN27oSnn747siNEDo9gaC1A5Wrs3buXyMhI4uPj8x1F6NChAwcOHGDFihWcO3eOyZMn5wo0TG1LCCGEMFV4eDjnz583/Q2aphKqK1RQ05bySa7W++yzz3jttdfYunUrb7/9Nr/88gtz5szJPeVK09TIRM6Ea2dnNYqRx3kMS84eOcLqI0d4Y8UKlm7ZYvq1PIyqVVNTw3btKrCAoHi8SWDxiBo3bhzm5ubUrVsXV1dXLubzH3DXrl358MMPmTBhAs2aNePWrVu8/PLLRWpLCCGEMNX06dOpWbMms2fPNu0NiYmqMnbFiiqx+PhxNV0pH9lXhjIzM+OZZ54xLDZi5Pp1VWXa2dl4u5mZCjbOnMl1Hr9/RyZOR0dzIDoagKY1ahTc/4sXYeRIeOst+Pvvh6/4nJmZSuY+eVKNXKSllXWPxANIpkI9omrVqmVYwk9v0KBBeR47depUpk6dWqi2vL29c+VmNGzYsFCrVwkhhHg8JSYm8tNPPwHQunVr09507ZqqNeHmpqZA/fOPmhaVz5x//dLo96yvFBurksI9PXPvq1hRrRYVEwMeHmpbZia1/q1xEZ+URPy/ReSa5pe4rf+9OH/+3ZWnPv4YatSAl15SieglVBiw1Flbq/t9+LAKNNq0gWxTp4WQEQshhBBC3FerV68mOTkZPz8/WrZsadqb9Eux60cSNE2NWuTzgZZ+xOL//u//WLp0af7tXryocivyeri3s1PTfiIi7m47dgz7M2fwyrasbDl7e2r8uzKikdRU+P579X1mJrRoAc89p4rPnT8PU6bAxInqOh4WdnYquDh4EHbvVrkpQvyrTAOLHTt20LNnTzw8PNDpdPzyyy9G+wcNGmSo9qz/6tatm9Ex169fZ8CAATg5OeHs7MzQoUMLXLFJCCGEEGVryZIlAAwdOtS0ZWZBLf+aPQ/C3R3Cw9W8/zzoAwvIvVqjQVKSGvXIOQ0qOxcXOHsW7txRQciePVC+PH7ZisU2rVEj93VERcG4cfBv8Tx69lRBxMsvQ3AwPPOM+rT/5EmYNAlmzFDTvR4Gdnbg5aWubfduFTQJQRkHFklJSQQEBPDVV1/le0y3bt0MFZ+jo6P54YcfjPYPGDCAEydO8Mcff/Dbb7+xY8cORowYUdpdF0IIIUQRHDt2jH379mFhYZErpy9ft26pqVBOTne32dmp+hP5jFrUq1cPFxcXAgICeOGFF/JuNzZWPcxnbzcnFxd1bn1uQXo6uLrily1wyZVf8fff8PbbcOEC6Avqduhwd1TE2RmGDlUBRvfuql7Evn3qPefOmXJHyp69vZo+tn+/CrYkuBCUcY5F9+7d6d69e4HHWFtb51nxGeDUqVP8/vvv7N+/31D47csvv6RHjx58+umneOjnQwohhBDigaAfrejZsyeVKlUy7U3Xr6v8ipzHu7urKUUNGqjvsylXrhwXLlzAzMzMaAl1I/rRjgKWrcXcXE2VOnVK1dGoVQuAj/r3J+TwYc5euUKz7PkVYWEwa5ZK+K5TB958U+Vo5KVCBZXQ3a2bek90NLz7LowYAV27Pvi5Fw4O6r7v3avuU+PGKg9DPLYe+ByL0NBQKlWqRO3atRk5ciTXrl0z7NuzZw/Ozs5G1aQ7deqEmZkZe/fuzbfN1NRUEhMTjb6EEEIIUXSpqakMHjyY6dOn57uQh6Zp/Pbbb4CaBmWya9fUg3rOAMDBAVJS4MSJPN/m4OCAXc5lZPUyMlTuREGjFXqVK6vgwMfHUASvopMTB+bOZfvHH9Pe318dd/UqfPqp6mu7dhAUVPA0Kz0fH5g7F5o3V/1auFAle2df1jUzU41mrF0L06bBV1/lH7DcT05O6v7s2gU//6yKBRai0K54tDzQq0J169aNPn364OPjQ3h4OJMmTaJ79+7s2bMHc3NzYmJicn3aYWFhgYuLCzEF/GObOXNmgasgCSGEEKJwli1bxrJlywBwc3Nj+PDhuY7R6XQcPXqU9evX07VrV9Mbv3gxd50JPTc39cDt768ecE119aqqU5F9dsO1a+oTd/30JT07O/i3dkV2jnZ2tK1fX71IT4fZs9XUKl9feOMNNcXJ1GVZHRxUDsa6dbByJWzdqla+at9eBU7Hj6uckOy2bIFOnaBfP1XErqyUK6emRsXEwIYNKv+iYUNVjdzigX7UFCXsgf5p9+/f3/C9v78/DRo0wNfXl9DQUDp27FjkdidOnMjbb79teJ2YmGhYlk4IIYQQhZOens6sWbMMr0ePHk2jRo2MZhTo2dnZ5Z/zkJfkZLUiVM6HfT0nJzWF6ORJ9XBtZuJkjJgY9dCvn7oTHw+vv64ekOfNg/LlTe8jwJIlKsnbwUFNZ8pexdtUZmbQty/UrAlz5qgRlewrUtnbQ/36ULcuHDkChw5BSIgKQrp3VytOFbbfJcXCAqpUUQFWdDT89puqjh4QcM8q6eLR8UAHFjlVr16dihUrcv78eTp27IibmxtxcXFGx2RkZHD9+vV88zJA5W1YyxxAIYQQokR8//33REZGUqlSJZo1a8aGDRsYOnQohw8fxuzfB/3U1FSsrKxMXwVK79o1lbzt7Z3/Me7uqrbC7dsq38LLq+AAQ9PUA7u9/d1tW7aoaVUpKfDJJ2qVJlMfhkNDYeNG9f3bb6tRlOJo0AA+/xyWLVPX7u+vtlWvfrdPzz6rgqnvvlOjGb/+Cps3w9NPq32mTPEqDZaWajnatDS14taFC2qqV0CAqt5tauAnHkqF/ulOnjyZCxculEZf7uny5ctcu3YN938TtJ588kkSEhI4ePCg4Zg///yTrKwsmjdvXiZ9FEIIIR43zs7O1KxZk3feeYfvvvuOZ555hp9++skQVADMmjWLunXrsmbNmsI1rs+vsLBQOQd5zd93clLBRHg4/PIL/P67eqjNmeuRnKymVR08qEYsypVT27OyVGChd+IErFhhWv8iI2HBAvX9f/4DeYzSFEmFCvDOO6rWhX4UI2egU7euyuOYPl0llaemqhyM4cNV/YycU6fuJysrFVB4eal7vn69Cr4uXiywYrp4uBV6xGL9+vUEBQXRrl07hg4dSt++fYv86f/t27c5f/684XVERARhYWG4uLjg4uLC1KlT6du3L25uboSHhzNhwgRq1KhhmJdZp04dunXrxvDhw1m8eDHp6emMHj2a/v37l/6KUHfu3N9y9lZWYGt7/84nhBBCmOiZZ57h6aefJiMjA2tr61x1qbKysli6dCkXLlwgNXtCsimuXFHTlW7dUisoVayoRhRyTjWysVG5DcnJcOaMCjJq1VIP5LduqToYMTEqB0LT1JQl/fSqY8cgLk6NYIwYAZ99pnIdateGggr4JSWp1ZzS0lQF7WxTuO8bnU6NBjRooOpKfPedGo1ZvVrlO/TpA089pe5PWbC2VgHGnTtqBa9//lFVxwMC8q52Lh5qhQ4swsLCOHz4MEuXLuXNN99k1KhR9O/fnyFDhtCsWbNCtXXgwAHat29veK3Pe3jllVdYtGgRR48eZfny5SQkJODh4UGXLl2YPn26USCzatUqRo8eTceOHTEzM6Nv377Mnz+/sJdVOHfuqMj7xo3SPU925curYjomBhe3bt3iww8/ZN26dcTFxdGoUSO++OILo5+RpmlMnjyZb775hoSEBFq1asWiRYuo+e+yeampqQwbNoz169fj5ubGwoUL6dSpk+H9c+bM4eLFi3z55Zf37E9iYiKzZ89m7dq1REZG4uzsTP369Xn99dd59tln0el0BAYG0rBhQz7//PNC3hwhhBBlzdzcPM9CdH/++SebN2/mwoULlCtXjj59+pjeaEqKqjXh6KimOiUmqq///Q8GDsz7PXZ26sH19m018nDihPqE3MZGjWz4+OROKN68Wf3Ztq1Klo6IUCMfX3yhpu/k9QB8+7YKQKKiVLDzzjtlm0eg00GzZtCkiaor8f33Kphavlxdy4ABZbuEra2tceB35Qr06CHBxSOmSDkWjRo1olGjRsydO5dff/2VpUuX0qpVK/z8/Bg6dCiDBg2inH54sQCBgYH5LkkHEBIScs82XFxc+P777wvV/2JLS1NBha3t/fkEICVFnS8tzeTAYtiwYRw/fpyVK1fi4eHBd999R6dOnTh58iSe//4j/uSTT5g/fz7Lly/Hx8eHDz/8kK5du3Ly5ElsbGwIDg7m4MGD7Nmzh02bNvHiiy8SGxuLTqcjIiKCb775hgP6iqIFSEhIoHXr1ty8eZMZM2bQrFkzLCws2L59OxMmTKBDhw44m7IcnxBCiAfKli1bOH/+PIMHD1Yf+iUlqU+l69YFS0tCQkLo0aMHWf9OfRkwYAC2hRl9v3ZNBRJVq8LRo3e3//wztGqlcg7y4+CgRiuysgqe13/rlipoB9C5s/rz5ZdVIvbJk2pEYs6cu7/vk5NVPsMvv6jrtbCA994ru5yGnMzM1L1p0QJ27IAfflAjNQsXqpGZ0aPLdgaEPvALD1f1L55+umiJ7uKBVKwMGk3TSE9PJy0tDU3TKF++PAsWLMDLy4v//e9/JdXHB5eNjRo2Le2vQgYvd+7cYe3atXzyySe0bduWGjVqMGXKFGrUqMGiRYsA9bP7/PPP+eCDD3jmmWdo0KABK1asICoqyjCEferUKXr16kW9evUYNWoUV69eJT4+HoCRI0cye/ZsnEz4j3TSpElERkayd+9eXnnlFerWrUutWrUYPnw4YWFhOOS30ocQQogHlqZpTJw4kZEjR/LJJ5+ojadOwbZtarUioF27djRu3NjwnmHDhhXuJNevq/oNlpZ3A4tKldS2L780rdrzvZKFt29XeRs+PuoTdVDBwoQJarbAhQvqoTw1VU2PGj4cVq1SQUW1avDRR4aieQ8Uc3M1+rJwIQwerF7v3AnjxqmRjLJWtaqaFnX8eFn3RJSgIgUWBw8eZPTo0bi7u/PWW2/RqFEjTp06xfbt2zl37hxBQUGMGTOmpPsqTJSRkUFmZiY2OQISW1tbdu3aBah8lpiYGKOpTeXKlaN58+bs2bMHgICAAHbt2sWdO3cICQnB3d2dihUrsmrVKmxsbHj22Wfv2ZesrCxWr17NgAED8sx7cXBwwELWuBZCiIfO5s2bOXDgALa2trz22mtqatDRoyp/Yd8+OH8eGxsb1qxZg4+PD8888wyNGjUq3EmiolRQERenPnU3M4OpU9WHbuHhalpycWga/PGH+r5zZ+NpQi4uMH68OmdoKAwaBEuXqhEOT0+174svVL2GB5mFhVolKihIXdOlSyq4+Pd5oMxYWqr+HDigfr7ikVDowMLf358WLVoQERHBkiVLuHTpErNmzaJGjRqGY1544QWuXr1aoh0VpnN0dOTJJ59k+vTpREVFkZmZyXfffceePXuIjo4GMBQQrJyjmFDlypUN+4YMrBNsTwAAIABJREFUGUJAQAB169YlKCiIH3/8kRs3bvDRRx/x5Zdf8sEHHxiS6a9cuZJnX+Lj47lx4wZ+fn6leMVCCCHut6CgIABee+01XF1d1dSh+Pi7RdF274b4eKpVq0Z4eHiuhG5ATXM6ccK4wrReWpqqh+DoqKbwgJra5OkJ+qrd33+vgo+iCg9X+RSWlqpSdk7168Mrr6jvk5LUaMmbb6pVoNq0ebiWTq1bV+WE+PurXNFPPoH//ldV+i4rrq4qIN23r2z7IUpMof9F9OvXj8jISDZs2EDv3r3zTNSqWLGiYT6lKBsrV65E0zQ8PT2xtrZm/vz5vPDCC0ZL/92LpaUlX331FREREezfv5/WrVvzzjvvMGbMGA4fPswvv/zCkSNHaNGiRb4jVAXl0AghhHg47dixg507d2JlZcW4ceNU3sGxY2rqkJmZeviPj1efiqek5F274tIlVURtwwaVPJ1zQZTr11Xg4eh4dxpUQID6s2NH9X1aGnz1Ve5lZU2lH6148kl1nrz07g2vvaYqaS9apM5tSpL2g/gcVL48TJumlq8F+L//UyMvhw8X/R4WV9WqKpn79OmyOb8oUYUOLD788END8q94cPn6+rJ9+3Zu377NpUuX2LdvH+np6VT/N9FNX0AwNjbW6H2xsbH5Fhfctm0bJ06cYPTo0YSGhtKjRw/s7e3p168foaGheb7H1dUVZ2dnTst/GEII8ciYMWMGoEa2PTw8VML21avqE2hQU4p8fNQoxt69xg/ZWVmqavSGDSo528dHPVhu2GBcZfraNRU4WFndDSz8/e+2P2qUWsr02LG7qzoVRmqqSm6Gu0nbedHp1OpFnTurkQ1T3LihrrGwS+veD+bmahRm0qS7U8omT1avT5y4//2xtlaJ7/v23d/VNkWpKNIY3uXLl1m4cCHvvfceb7/9ttGXeLDY29vj7u7OjRs3CAkJ4ZlnngHAx8cHNzc3tm7dajg2MTGRvXv38uSTT+ZqJyUlhVGjRvH1119jbm5OZmYm6f8WKUpPTycznwQ6MzMz+vfvz6pVq4jKY7j69u3bZMjwpxBCPDT++ecf/vjjD8zNzXn33XfVyoVHj6pP/LN/km9pqYqjHT6skrpBjWyEhsLWrSpg8PZWC5TUrAk3b6oCagcPqqTsmBg1pSoqSgUZlpaQfVqtmxu89JL6ftkydUxh/PWXmt5UufLdgKWkxMWpaytsn+6nFi1UYnevXurenjgBEyeqIOPcufvbl8qV1QjVvn0P5kiPMFmhs2a3bt1Kr169qF69OqdPn6Z+/fpERkaiaZrRyg+ibIWEhKBpGrVr1+b8+fOMHz8ePz8/Bg8eDIBOp2Ps2LHMmDGDmjVrGpab9fDwoHfv3rnamz59Oj169DAk3rVq1Yrx48czePBgFixYQKtWrfLtS1BQEKGhoTRv3pygoCCaNm2KpaUlO3fuZObMmezfv1+WmxVCiDKSnp7Of/7zHxo3bswHH3xwz+Pv3LlD586dsba2xtvbW63qExOjlhDNycFBzaHfvVt98n/ypFoJqGpV9Wm5npmZ2nbtmlpV6to11aaTk0ruBRVU5CzI+/TTaqWjs2fVNKX33ze9ToN+GlTHjiWbK5GQoK7b01MFVQ/y77fy5WHYMDXd68cf1T05fFh9de6slqa9H3UvdDoVhJ46pQKy2rVL/5yiVBQ6sJg4cSLjxo1j6tSpODo6snbtWipVqsSAAQPo1q1bafTxwZWS8sCe5+bNm0ycOJHLly/j4uJC3759CQoKwjLbMO6ECRNISkpixIgRhloTv//+e67VpI4fP86PP/5IWFiYYdtzzz1HaGgobdq0oXbt2gXWEnFxceHvv/9m1qxZzJgxgwsXLlC+fHn8/f2ZM2eOSTVPhBBClI7ffvuNdevWsW7dOt5///288yGyqVevHpv1U49SU9VohYND/nkHbm5qus22ber4mjVzF6jTq1BB1Tk4ckS99va+Ow2qQYPcx5ubq4fft99Wn3avX68eku8lKkoFRDqdCixKUlycKlLn56celO/Xs0JxVKwIr7+uqnSvXq1Glf74QyV8l/T9yY+dnQoc//5bfe/ldX/OK0qUTitkdq2joyNhYWH4+vpSvnx5du3aRb169Thy5AjPPPMMkZGRpdTV0pOYmEi5cuW4efNmrroMKSkpRERE4OPjc/eB+yGovC2M5flzFEIIwVdffcXo0aMBtWJgztUCC3TqlMqN8PXNP1gANbXp1i0oV860T8AzM9U0JQcHVawuMRFmz4Y6dfI+fuNGWLxYjTzMmKFWcyrIihWwZo0KACZPvnd/TKWvDN6nj8o3WbuW6Ohogm/cYISfH+52diV3rtK0dq2q2O3goJLjy5e/P+fNyoKLF9XUrJYt1c/xYVp56xFV0HNyToUesbC3tyctLQ0Ad3d3wsPDqVevHoCheNojz9ZWPeT/ex/uCysrCSqEEEKUuPPnzxu+1+fOFSQ9PV2Nfqenq9GE/2fvzuOiqvf/gb9mhn3fQUAWBRV3yH1fc8m0stv1qj8t1+yWpVbmTTO1XLppVlpdzW/dbt42r6mZSy645q64LyCbqCiyiKwCw++Pl8OwDMjAGQR8Px8PHsCZw5mDwnDe5/NebG0rDioAriwYkxKk0TANKiaGF+q6OozyDBrEIGfvXk7J/uQTzkgw5N49YPduflxR0XZVJCayW5UuOGvalKlfdc0zzzDFLDoaWLUKmDmzZp5XreYqVVISsHMnU+I6duQKhqgTjA4sOnXqhAMHDiAkJASDBw/GjBkzcPbsWaxfvx6dOnUyxTnWTtbWcqEvhBCizrv4oLB69erV8PX1fej+3bt3R3R0NH785z/RJzmZF4KmokuDatGi4uBF1yUqNpaTsj/6iCsXpb8mKgpYvJiFwi4uQPv2yp1rRgYDoAc3WwGwzsLaum6kQxWn0bC97owZrI85fJjF3jXF3Z3BxPHj/L/q3p0zREStZ/T60rJly9CxY0cAwLx589C3b1/89NNPCAgIwJo1axQ/QSGEEEKYzoULFwAAIeWlGZUSGxuLpKQkOCUmMie+si1Yq6Ki+orSrKyAd97hhfyFC0x30iksBLZvB95+mzUQXl5MgVLy3G/eZAF78ZbtLi5194K4cWNO7AaYZpaRUbPPb2vLf8+4OM47uXKlZp9fVInRKxa6OQgA06K++uorRU9ICCGEEDXnv//9Ly5cuICQkBCkpKTApbwUIrAjlG7+UUBBQcmLaKUVFLDAGqhcYAFwheD117kqsWEDU5HateOFsa69eocOwBtvsH5AKZmZDFJatixZQ6LrdnT5ct1sozpiBHDoEIvdv/2WhfI1ycyMwcWNG0xfs7MDvL1r9hyEUaQippJkgnTdJv9/QghhWLdu3eDi4gIPDw+88MILFe4bFxcHALC3tYWzhQXr/0wlKorNUuzsjEu36tJFf6f9s8+AN99kUKFW6wfDVTaoyMtjnn9MDO+Yp6QYnlCta7dr6KJXt2KRmVn+82Rl8futbXOdLC31wcQff3AYYU1TqRgwZmUBERF1M0B7jFRqxcLZ2fmh7ed0UlJSqnVCtY3mQfu8+/fvw1pqKuqsrKwsACjRblcIIQR5enqioKAAMcUnXxug6/wY0KBBpa8LqkyXBtWyZfmtbMszZgwDgfPnmUrj6MgAo02bir+usJBzKO7dY12Erog8OJipOZcu8bgeHvpOSVlZDFpatDDc8UoXxNy9qy/qLi4/n52Q3NyYplXb7si3bAkMHAhs2wasWMFgrfQ8kZrg48N/+2bNgGLZM6J2qVRgsXz58qKPk5OT8cEHH2DAgAFFE5oPHTqE7du3Y86cOaY5y0fIzMwMNjY2SEpKgrm5OdTS9qxOKSwsRFZWFm7fvg0nJ6eiQFEIIQSwZcsW3LhxA8EPOi7Fx8cjPz8fZuUUShcFFm5u1atPyM/ndOdLl7ga8WD4agnG1FeUptGwnmL+fAYGU6dyRkZF8vJY/G1nx2F9DRvyYt/NjfUbAIOHc+fYger2baBBA3YwCgoCHlb4np/P9K7if4cKC7ka0qgR4O/P+RGl96kNxo4Fjh1jHck333Co3sM6gSnNxoYB3MmT/L+RG4W1UqV+KsaOHVv08fDhwzF//vyintcAMHXqVKxYsQI7d+7EtGnTlD/LR0ilUqFBgwaIiYkpWgIWdY+TkxO8TJkLLIQQddCqVauwceNGfPrpp7CwsMD9+/eRkJDAidoGFAUWTk76i+3KKCjgRfuZM3w7f75kp6SnngLGjdNfLObl8eIdKLnKcP8+cP06uwY9LJ3J2ZltZysjIwO4do0BQvfuPL4hrq5Az54cHHfuHAMjtbpy8xYcHLgaUjzAuXmTqynduwP29iw61wUstYmtLTBlCjttbdnCydx/+xvPuyaDIF9ftsCNjOT/gah1jA43t2/fjiVLlpTZPnDgQLzzzjuKnFRtY2FhgeDg4KL5HaJuMTc3l5UKIYQwQNdqtkWLFvD390dkZCRiYmLKDSwaNWqEfn36IKxBg4oDi8JCICFBH0icPVu2q5C9Pe/Unz7NIXtRUZyX4ObGYuf79xkcFF8JuH2b21JSeFHu7c2L3uq4dYupTx06sKVqZQImd3egd28O7Ltzp3JTon19+T3qAou7dxlcDRyor8No04a1DB4etW/VokMHTudeu5b/9suWAb/8AowaxX+3msjoMDdnQHnyJFe6ZL5FrWN0YOHq6oqNGzdixowZJbZv3LgRrg9bZqzD1Gq1TGwWQghRb+Tm5hYNxwsJCUFgYGBRYNG7d2+DXzNp0iRMGj4c+PFHwxfgf/7JLkJnzgCpqSUfs7bmnf3Wrfnm78+L0WPHeJF6+TK7Nb31Flc0AO5XvG4hI4PF2W5uDFauXGExb4MGxl9k6lZRrK05KK95c+Mvjr28Kt8Zq0ED3mnXpUTdvMnvpUkT/T5BQSxQvnPHcD3GozZwIFdsNm8G1q/nKs/ixQwQX3wRaNvW9Ofg5cV/x4sXOTld1CpGBxbz5s3DhAkTsGfPnqJ5FkeOHMG2bduwevVqxU9QCCGEEMqLjIyEVquFo6MjGjRogMDAQAB4aAE3MjOB3NyyBbw//cS72ToWFryjrwskgoIM34Vv354pS4sXM81l7lx9mlOrVvr97t3jdl9f3vX39mbNw+nTXAlQq/XBysPcv8/n8vUFevRgYbCpubvzvO/c4WpFSAi/9+KBk60tg6/du7l/bazrtLYG/vIXTjvfuBHYtIn/lu+9x+0jR5p2tUWj4b/jqVMMaHRF9KJWMDqwePHFFxESEoLPPvsM69evB8A7HQcOHCgKNIQQQghRu+nSoEJCQqBSqdC9e3ekp6ejVfGL+WIKCgqQm5sLm4wMpjoVv+j99Vd9UDF4MNCtG2dIVLbA1ssLWLIEWLUK2LEDSE/n9uKF28nJDBx0czZ0MyJ8fNhVae9etn2tTFeluDh2eurblylZNcHCgsHVzp0cPtetm+F2vU2aMFhKTi6/1qM2sLNjGtTTTwPff8+uUb/8wlWkN99k7YipuLnxec6c4QqKqDWqVNLfsWNHrC1+V0IIIYQQdUrpidujRo3CqFGjyt0/MjISISEhCPL1RWTxRi2bN7NTEACMHg08ZBZGuSwtgddeY0Dyr3+xM5MuzUir5SpJUFDZlq5qNfPts7OBrVv5vqL28MnJTJvq1Knmggodf38GNF27Ak5Ohvext+eqxd69vIA2dVvf6nJwYO1FixZsR3v6NDBtGrtyNWtmmudUqfizceECf16kOUutUQvX2IQQQghhasVXLCpD1xHK2sxMnwa1fTtXGQAGFFUNKop78klOeS7eKCY9nRewFbV0bdKEdRLx8YaH2AHsNpWUBISFPZqLUV9fdsDy9694v6ZN9UXqhuTksJC9Ng1/7dkTWLqUK0h37nAQ4ebNpjtHR0cGkTI0r1aRwEIIIYR4DH366afYtWsXnn/++aJtBQUFiIuLQ76BCdBFrWadnRlY7N4NfPEFH3z2WabFKMXevmSaUEoKL8YdHMr/Go2GNQvOzrzoNuTaNSAw8OGD8kypMilCjo5ctUhKKnlhnpvLuRcJCSwAv3bNdOdZFX5+DC66dGGR+qpVXMUwVXDh48NViwerb+LRk8BCCCGEeAx5enqiT58+RUXbAODh4YGAgICiblHFlQgsIiI4gbmwkHfgX3zRdCk7BQVcaajMtGVXVwYXd+/yIry4u3c51K1Dh0czOdpYTZsywEhN5fcfF8fVGF9fYOhQtrsFyl/VeFRsbNg2ePx4pqnt2AHs22e653JwAA4e5HwT8chJYCGEEEIIAIDPg+5IhjpD6YbEBlhasgZCq2Xa0sSJpq0DSEtjwXZlOzeFhDAtKj5ev62gALhxgysVfn6mOU+lOTsztevGDa5SeHqyUHroUBZ/BwcziEpKKjlssDZQqYBhw4ARI/j56tX6gnyleXry+9+3j53DxCNlVGCRl5cHMzMznDt3zlTnI4QQQggTO3bsGGbNmoUtW7aU2F5Ry9miFYv8fKa5NGjAol1Tt0RNSeFqRWXnVJibc1XC1paF2gDvZjdsyNqKuqR5c3bGeuop4JlnGEyYFeu7ExbGlryxsQyeapvhw5nClp7O4MJU/P2ZHnbwIH82xSNj1KuBubk5/Pz8UKDQD+++ffvw9NNPw9vbGyqVChs2bCh6LC8vDzNnzkSrVq1ga2sLb29vjBkzBjdu3ChxjICAAKhUqhJvixcvVuT8hBBCiPooPDwcixcvxvfff19ie6UDC4ApOaYOKvLy+P5hxc6leXpyeNqdO0wlys8HOnase5OaXVwYVDRrZrh1r5kZ6xl8fRlc1Dbm5uz0pVazy9WJE6Z5Ho2GncHOneNUbvHIGP2K8O677+If//gHUhTI6cvMzESbNm2wcuXKMo9lZWXh5MmTmDNnDk6ePIn169fj8uXLGDp0aJl958+fj5s3bxa9vfbaa9U+NyGEEKK+Kt1qFoWFQHo6Gj2oYygdWGi1Wjz99NPo164dAu7f58aa6KqUmqofhmesVq1YqB0by7v+xWpJ6hV7ew75s7YGbt161GdTVpMmTOECgJUrgaws0zyPlRXg4QEcPcqBfeKRMHqOxYoVKxAVFQVvb2/4+/vD1ta2xOMnjYgUBw0ahEGDBhl8zNHRETt27Cjz3B06dEB8fDz8iuVI2tvbw0t6GAshhBCVUqbVbHQ0cPIkAh9cwJcOLNRqNVatWsVJy/PmcWNN/N1NS+O8iaoUW1ta8mvNzbl6UdvnQVSHjw9nY+zYwRQw3eTy2mLUKODIEQ4w/O474OWXTfM8zs6ss9i/n4Xvrq6meR5RLqMDi2eeecYU51Epd+/ehUqlglOpoTKLFy/GggUL4Ofnh5EjR2LatGkwM6vS7D8hhBCiXissLCwKLJo3b87c/LNngevXEdiyJQDDqVAoKOCFvq5uwZjAIiODd9OtrNjFx9b24WlUublMcalOsbWPD1c76nNQodO8OVO/jh4tW4vxqFlZAX//OzBnDocY9ujB8zWFhg05lXv/fk6BNzTdXJiM0T91c+fONcV5PFROTg5mzpyJv/3tb3Ao1sd66tSpCAsLg4uLC/7880/MmjULN2/exLJly8o9Vm5uLnKLtaFLN1WnAiGEEKKWuX79Ou7duweNRoOgoCB2T4qNBXJyEGhjgxEjRiAwMBD5+flFN+nu3r0Li5wcWGdnswsRwDqGysjMZGejli1ZxJuSws/Vat5Zd3Bg7UPpi/+UFMDdnUXi1fE4BBUA/z07duQMD928jtqkTRugXz9g507Otli+3DQX/SoVi/0jIxkwP/GE8s8hylWlcDYtLQ3r1q3D1atX8dZbb8HFxQUnT56Ep6dnUas6JeXl5eGFF15AYWEhvvzyyxKPTZ8+vejj1q1bw8LCApMnT8aiRYtgWc7S6aJFizBPt5QrhBBCPEZ0qxVBQUGwMDNjwatKBdjbwzYpCT/88EOZr1m8eDEWL16Md3r0wKLsbG6sTGCRlcVuPU88AXTvzhWIu3d5Z/32bc5mSE1l16YH51AUaKSns+tRbbrzXttZWzO4+O03/jtXZhhfTRo3jgXcCQnAzz8Do0eb5nnMzQE3N+D4cRa2VzYIFtVmdPH2mTNn0KRJEyxZsgQff/wx0tLSAADr16/HrFmzFD9BXVARFxeHHTt2lFitMKRjx47Iz88v6l5hyKxZs3D37t2it2u1bXKlEEIIYSIl6isSElhf4eXFC/qkJK4wlKL7m+qu1XKDi8vD6x6ys7ka0rYt0K0bAwSVCnByAoKC2M3or38FXniBxb3t2vHC+M4d4PJlps/4+ir5rT8e/Py4OnDjRu1rQWtnB0yezI9/+YXT203FzY0/y0ePSgvaGmT0bYDp06fjxRdfxEcffQR7e/ui7YMHD8bIkSMVPTldUBEZGYnw8HC4VqIIJyIiAmq1Gh4eHuXuY2lpWe5qhhBCCFGfvfrqqxg2bBju5+ZytUKr5QqBpSWQlISCW7dw3cwMGo2mKAuhqNWsLq3oYfUVOTlcjWjThvn0hlqlAkzfcXXlW5MmvBBOSWFwkZPDLj/COCoVV3quXeNKUG0bCNilCzBkCLB5M6e3m5tzNcsU/PwYpAYEsEuYMDmjVyyOHTuGybposxgfHx8kJiYadayMjAxEREQgIiICAIvFIiIiEB8fj7y8PDz//PM4fvw41q5di4KCAiQmJiIxMRH3H7S6O3ToEJYvX47Tp08jOjoaa9euxbRp0zB69Gg4Ozsb+60JIYQQ9Z5arYa/vz+CbWyAqCh9K1eNBigsxJwFC+Dv749FixYVfU1RYFFYyA0VpZbk5upbvPbsaVwevUbDuoqQECA01PRzMuorW1sOCczPZ+F8bTNhAqe2a7XA0qXAoUOmeR5LS6aDHT3KgFWYnNErFpaWlgaLna9cuQJ3d3ejjnX8+HH07t276HNdvcTYsWPx/vvvY9OmTQCAtm3blvi68PBw9OrVC5aWlvjxxx/x/vvvIzc3F4GBgZg2bVqJugshhBBClFJYCJw/zwvP4kPjbGwQ8GBFX9cZKjs7u+jGYdEMCw8P3hHXBRrFZWayULtnz6q1iRXKaNyY/w+nTrFLlKEgLSuLNS42Nqxvqal6FrWaU9vz8oDwcOCf/wT+8Q+mwynN05Ndoo4dA/r3l2DVxIz+CRo6dCjmz5+Pn3/+GQCgUqkQHx+PmTNnYvjw4UYdq1evXig09KL0QEWPAUBYWBgOHz5s1HMKIYQQj6vk5GS8/PLLaOHnh7mBgVCV7rjk4IDAUoFFfHw8AMDeygrOuhuLdnZciTA0uM7ODmjfnjUS4tFRqfj/kJAA3LzJ1rs6OTn6gnkvL85+iI3lCoK1Nett7Oy4gmQqajUwdSqDiwMHgEWL2I621M3kalOp2IL2wgVOcG/WTNnjixKMDiyWLl2K559/Hh4eHsjOzkbPnj2RmJiIzp0748MPPzTFOQohhBBCARcuXMC6detwzNMT77/9dtlBanZ2CHywghEbG4vCwkJ9GpSLC1R37nA/JycWxw4eXINnL4zm4MAuUVu2sJherWZRt1bLFY3WrXnRnZPDupY7d1gbk5TE/Ro1Mu2qk0YDTJ/O4OLIEeCDD4D33+dKi5JsbBgwHTnCYPghjYBE1RkdWOgmYh84cABnzpxBRkYGwsLC0K9fP1OcnxBCCCEUcuHCBQBAiKur4ToJtRp+Tk5QqVTIzs7GrVu34O7ujvEjRqDB7dvAnj3cz8GBnaFE7dekCVcjTp3iKlJAAIvqAwL0aUE2Nix09vNj4fe9e5zinZBg+nkYZmbA228DCxeyFe2CBZzurvTKQoMGTIk6fhzo1UtSokykysl03bp1Q7du3ZQ8FyGEEEKYUNHEbTc3/V3bggLg3//mRWW/frBwcoKviwuuJScjJiYGnTt3xtdz53Kf3buZAmVjU/tmJAjD1GoWcqtUXKUICHh4LYW9PVcz4uKA+/dNP73a3Bx45x2uWJw+zcBiwQK2JVaKWs32xRER/Nl/4onHZ3hiDapSuLZr1y4MGTIEjRs3RuPGjTFkyBDs3LlT6XMTQgghhIIuPujCGBIcrN948iSwYQPw1VcMMhwcEPhgNUJXZ4GMDCA5mR97efGCrHjRt6jdXFyAAQN4oV7ZAu2AAKZJ3bxp0lMrYmkJvPsu0Lw5GwDMnQvofv6UYmfHFL4DB/hz/5BaXmE8owOLL774AgMHDoS9vT1ef/11vP7663BwcMDgwYOxcuVKU5yjEEIIIaorMREXzp4FUCqw+PNPvr9/nwPtbG0xolUrvDNpEkJCQhAXF4fsmzc5yRlgRyiNhi1NRf1lZsZVi9xc1kDUBCsr4L33gKZNmY713nvsPqYkZ2fWCB04wNURCS4UZXQq1MKFC/HJJ5/g1VdfLdo2depUdO3aFQsXLsTf//53RU9QCCGEENV0+zbubtqEhAe9/EMaNuT2/Hz2+NeJigICAzGla1cW/YaGwsvLC7du3UJE585oA/COr5WVBBaPg8BArlokJvJ9TbCx4WrFnDnA1at8v3Ch4Q5kVeXqyoBi3z4GyTI8TzFGr1ikpaVh4MCBZbY/+eSTuKu7myGEEEKI2uHOHWDHDqRdu4ZGnp5o6OYGF3t7Pnb+PO8M60RF8b29PRAXh+zMTNy6dQsA0DA7m4+5ujJtRVKh6j9zc150Z2fX3KoFwJSlefPYHjYlBZg9G3jwc6gYNzc+z549/D0QijA6sBg6dCh+/fXXMts3btyIIUOGKHJSQgghhFBASgq7+9y6Bf927XDik0+wdsYM/eO6NChnZ75/EFho7exwLSYGP3/7LYAHMyxSU7mPkxMDD3PzGvomxCPVqBFXC5S+sH8YBwdg/nwWXN+5A8yaBURGKvsc7u5sQ7tnD3DpkrLHfkwZnQrVvHlzfPjhh9izZw86d+4MADh8+DAOHjyIGTNm4LPPPivad+rUqcqdqRBCCCEqLzUV2LkTBQkWslsDAAAgAElEQVQJ0DyYvOxkZ4fuLVrwca0W0A2Z/etfWbwdEwPk5eFGdjb8Zs8uOlSAszNUSUn8xN5eWs0+TiwtWWuxbRtT52pqOjfAgHfBAq5YXL8OzJwJTJrEQnSlOjp5erJAPTycnaOaNFHmuI8po3861qxZA2dnZ1y4cKGoHzYAODk5Yc2aNUWfq1QqCSyEEEKIRyEzE9i1C5lRUej5f/+Hcf37Y8qgQVAVvxi7dInBh40N0L8/sHYt06Li4uDdqBEsNBrcLygAAAQ4OOi7Azk4cNVCPD4aN+YF+O3bhmsdUlP51rCh8itZrq7Axx8Dn37KQPiLL4CLF4FXXlFueF+DBhwIuHs3gwsl29w+ZowOLGKUbv0lhBBCCGXduIHCmBi8/McfOHH1KhKSkzGie3d9bQUAHDrE9+3b82KwcWP2+I+KgjooCAFubrjyIP0lwMqK+7q4cKaBFG4/XqysuGqxfTsDDI2G2+/eZWG3vT3bEF+9yp8jpYMLW1umQv36K/Ddd1xdiI7mNqWKur29ORBw925+f6YeDFhPydhBIYQQor65cwdfHTmC7/fuhUatxs9vv10yqCgs1AcWXboAaWm8OASK6iwCvbyKdg/QDUjz8uIdXSncfvwEBbHVcFISkJ4OXL7MFa527YDhw4GnnuI+0dFMmVKaSgU89xxToxwdObxv+vSSXc2qy9eXReq7dvH4wmgSWAghhBD1SWEhju7ejdc3bAAALB47Fj1atiy5z9WrTGuxtATCwjgAT5f+8aBANrBBg6Ld21tb8wM3N36NrFg8fmxsuGqRksJANDSUAUWvXvoOS3378k7/1asctmgKrVoBy5cDISFAVhawZIl+eKMS/PyAnBxg507lZ2g8BiSwEEIIIeqR65cu4fmlS5FXUIDnOnfGjGeeKbuTbrXiiScYKGRn86IR4JC83FwEenoCAEa2a4fuutoMV1eZYfE4a9qUgcRzzzGI8PAo+biDA9CvH9vERkWZLrhwdQU+/BBo1owrDA+CaMX4+RXVKeH6dWWPXc9JYCGEEELUE1lZWeg7ZAiupaaiibc3vnn99ZIF2wDToHRtZh90d0RhIS+m7O15MRgTUxRYxCQm6u8IOznxzrVSRbOibrGx4eDEYqtZZTg6Mrjw8zPtyoWZGTBiBD/etk0/GV4JKhUQEMBj7trFwnRRKRJYCCGEEPWEjY0NJg8dCn9nZ/wxfz4cDNVCXLvGu7BmZizcLijghZSNDXPMASAqCqGNGmHmc8/hlaFDmTYF8KJRWs2Kh3F2ZnDh7c3gQqs1zfOEhjKFLzcX2LRJ2WOrVEzrun0bOHbMdN9DPWN0YLFt2zYcOHCg6POVK1eibdu2GDlyJFIlohNCCCEencJCTOvWDWfnzIF/6TQVHd1qRdu2DCays/nex4ftQgEgKgpB3t5Y/OKLGN23LweUARJYiMpzcWEbY123KFNcmKtUwAsv8OPNm1krpPTxGzbkZG7dVHpRIaMDi7feegvp6ekAgLNnz2LGjBkYPHgwYmJiMH36dMVPUAghhBDlS09Px5QpU5CWlsYLq+Rk2JcXVAAlu0EBDCxsbblaoSvgLn4RlZTEi0JLSxboSkcoUVlubgwuPDw4fLGwUPnn6NCBNR3Z2QwulGZjw7qio0fZBUtUyOjAIiYmBs2bNwcA/O9//8OQIUOwcOFCrFy5Elu3blX8BIUQQghhWEZGBgYPHoyvvvoKI0aMYC1ERkb5xdU3b/ICT63mBRnACzIXF65GPPj7joQEbgc4pwDg/AK1Wgq3hXE8PJgW5ezMVrRKBxdqtX7V4rff2ClKad7e/N05ccI0wVE9YnRgYWFhgawH/2k7d+7Ek08+CQBwcXEpWskQQgghhGlptVoMGzYMBw8ehJOTExYuXMhWoAUFwA8/AKtWcXWi+F1W3WpFq1bs4AOwtaaHBy/QmjZlgbZWy4tAQB9YeHhIq1lRNV5eXLlwduZ8CKUvzrt0YSrfvXss5FaaWs3jnzkDxMYqf/x6xOjJ2926dcP06dPRtWtXHD16FD/99BMA4MqVK/DVFX0JIYQQwqQOHTqE3bt3w8bGBtu2bUNYWBjbbt68Cfz8M3favJl54o0bM5g4cYLbdd2gdBwd+d7dnSlRaWlMh2rRAngwfRuurgwsJBVKVIW3N1vU/vEHg4uAAOWOrdEAzz8PfPopp3M/9ZTyncvs7bkiePgwV+/k98Ago1csVqxYATMzM6xbtw5ffvklfHx8AABbt27FwIEDFT9BIYQQQpR18uRJAECfPn3QsWNH9t2/c4dpTADvEjdsyLvDUVG84IqPZ6DRqRP3ycsDzM31U7cdHUsUcANgoAIwXcramm9CVIWvL9OirK35s6iknj25qnb3LrBjh7LH1mnYkL9fp06Z5vj1gNErFn5+fthsoDjmk08+UeSEhBBCCPFwusAiLCyMG1JSmAoSE8PPn3oKGDaM28+eZRrHxYts0anr7JSVxYs8XVqUoyNXNwB9YKFbsXBy4lvpuRhCGMPPj8HFrl3sFhUQwBWH6jIz46rFF18A69cDAwYwaFaSRsOAPSKCBeOSqVOG0YEFwLzOqKgo3L59G9pS7cN69OihyIkJIYQQonynHtw1DQ0N5YbkZCA/H7h8mZ83a8b3Li68m9uzZ9mDZGeX7PTk4ACEhPDj69e5CqKrsbC3l1azQhkBAQx89+0DrlzhRboSqUV9+wI//siVu/Bw4EEdsKKcnPQpUS1bsvOViwvrMITxgcXhw4cxcuRIxMXFobBU8Y1KpUKBqSYsCiGEEKLISy+9hKNHj6Jdu3bckJAApKdz1cLCAmjU6OEHycrifrpVCI2GKxaurrx4OnVK3x3K2VmfMiVEdXl5Mbj480+uprm68q06zM2B554Dvv4a+O9/geBgDrlTmp8fU7liY7ni5+TE4MjTk+lYupqlx5DRgcXLL7+Mdu3a4ffff0eDBg2gkiVRIYQQosa9/vrr+k9ycjghWFcPERRUuTSQ/PyyF3MeHuyAk5wM6AbiurryeFKwKpRkawv06cOfr8OH2Sq5YcPq3f0fMAD4/Xf+Lrz1FvD3vwO9eyt3zgB/Fxo3Zv1SdjYD+qNH2U3NzQ14+unqB0l1lNH/c5GRkVi4cCFCQkLg5OQER0fHEm9CCCGEqGHJyby40bXC1KUzVaSwkG+6+godR0d97vjx43zv4cGLKWk1K5Sm0QBhYcDgwbzzHxnJlslVZWkJ/POfrCW6fx/45BPgyy/ZqEBpKhWDbS8vro40acKBkpcuKf9cdYTRgUXHjh0RJWPNhRBCiEfm+PHjuHTpkj79WFdfceUKP69MYJGbyzSO0ulNxQu479/nezc3Th+WwEKYiq7uwsODF+fV4eAAvPceMGIEP9+6FXjnneof92FUKp7/hQtAaqppn6uWMjqweO211zBjxgx8++23OHHiBM6cOVPizRj79u3D008/DW9vb6hUKmzYsKHE44WFhXjvvffQoEEDWFtbo1+/foiMjCyxT0pKCkaNGgUHBwc4OTlh/PjxyMjIMPbbEkIIIeqMV199FSEhIVi3bh033LzJO7LXrvFzXeF2RUp3hNJxcCj79a6uDCwkFUqYkrMz0Lo1W8aWag5kNI0GGDmSAYadHVdC3njD9K1inZ05B0bXROExY3RgMXz4cFy8eBHjxo1D+/bt0bZtW4SGhha9N0ZmZibatGmDlStXGnz8o48+wmeffYavvvoKR44cga2tLQYMGICcnJyifUaNGoXz589jx44d2Lx5M/bt24dJkyYZ+20JIYQQdUJ+fn7RjbxQXbrHjRussQBYH1E6WDAkK4urE6UHiZmbszjVw0O/zcmJ+0rnG2FqwcFcIVNqdaFdO6ZDBQWxscH8+ayHMBWViud/7hwDpMeM0cXbMbr+2AoYNGgQBg0aZPCxwsJCLF++HLNnz8awYcMAAN999x08PT2xYcMGjBgxAhcvXsS2bdtw7Nixoq4Yn3/+OQYPHoyPP/4Y3t7eip2rEEIIURtcvnwZ2dnZsLOzQ1BQEAOKe/c4zRio3GoFwIJvT0/Dj3l5MUDRBSsODo9tMaqoYba2nBK/ezcnwSsRzHp6AosXczL3/v38ePZs1naYgqsr6ywuXwY6dDDNc9RSRv9v+fv7V/imlJiYGCQmJqJfv35F2xwdHdGxY0ccOnQIAHDo0CE4OTnpW+0B6NevH9RqNY4cOaLYuQghhBC1hW5+Rdu2baFWq1lfkZenH2gXEsJA48yZigtWtVqmbRhSvIAbkFazomY1aaJveawUCwtg+nSgSxfWIy1cyMGRplB81eLePdM8Ry1VpTDw6tWreO2119CvXz/069cPU6dOxdWrVxU9scQHA3k8S91N8fT0LHosMTERHsWXagGYmZnBxcWlaB9DcnNzkZ6eXuJNCCGEqAvKTNzWtZgtPhjv7t2K00kKCnjxU16w4Oion4OhK9qWwm1RU+ztgRYtGFiUmplWLRoNMGMG0L49UwgXLGChtSm4unJQX6na4PrO6MBi+/btaN68OY4ePYrWrVujdevWOHLkCFq0aIEdO3aY4hwVt2jRohItchs2bPioT0kIIYSoFF1gERoayhWJGzcYSNy/zyJVX1/WT3h5cbuhC7PsbMOF2zqOjpwq3KMHC2AtLCSwEDWrWTOulKWkKHtcc3Ng5kygbVumA86bp++mpiS1mud/9iwn2D8mjA4s3nnnHUybNg1HjhzBsmXLsGzZMhw5cgRvvPEGZs6cqdiJeXl5AQBu3bpVYvutW7eKHvPy8sJtXf7nA/n5+UhJSSnax5BZs2bh7t27RW/XdF00hBBCiFpMq9UWpUKFhYXxois9vWQ3KJWKwUTDhoCLi+G2l9nZ7PBU3oqFpSVXPP76V6BfP2k1K2qeoyNXLZKSyl+1SE4GLl7UT4evLAsL4N13GTxnZwNz5wLR0dU/59Lc3Vmn9BiNaTA6sLh48SLGjx9fZvu4ceNwQcHlpMDAQHh5eWHXrl1F29LT03HkyBF07twZANC5c2ekpaXhxIkTRfvs3r0bWq0WHTt2LPfYlpaWcHBwKPEmhBBC1HaFhYX47rvv8N577yEkJIQBRU6O/sKlWTP9aoSfn35gV2nZ2QwcNJryn6xBA+6Xm8tAQ1rNiprWtCkDjLS0ktu1Wg6DzMriPnFxxg/Vs7QE5szh70xmJvD228A33zBQV4pGw/M/c8b44KeOMjqwcHd3R0RERJntERERZeodHiYjIwMRERFFx4uJiUFERATi4+OhUqnwxhtv4IMPPsCmTZtw9uxZjBkzBt7e3njmmWcAACEhIRg4cCAmTpyIo0eP4uDBg3j11VcxYsQI6QglhBCi3tFoNBg2bBjmzZsH88JCdp5xdNRP+g0J4UWSnR1bxAYHc7WhdCpGTg7vplbEyYl3inNymDJlZnQjSSGqx8WFP9PFs1Nyc1m34OzMgXp9+wL+/vqp88awtuZqRevWTCX89Vdg4kRg7VpAqZloHh5AYiKgcC1ybWX0q8TEiRMxadIkREdHo0uXLgCAgwcPYsmSJZg+fbpRxzp+/Dh69+5d9Lnu68eOHYtvv/0Wb7/9NjIzMzFp0iSkpaWhW7du2LZtG6ysrIq+Zu3atXj11VfRt29fqNVqDB8+HJ999pmx35YQQghRtyQkcDXC1pZFomo1A4mbN3kX18yMdRYBAUzzCAws+fWOjhUf39GRKSP37lW+ha0QSgsJAc6f56qFSsWf75AQoFs3Br8A0L07sHkzL+ArSIU3yNaWRdwnTwLff88A4KefeLxnnwWefpoBSFVpNAzMT55kPVRpVlZse2tnV/XnqEWMDizmzJkDe3t7LF26FLNmzQIAeHt74/3338fUqVONOlavXr1QWEG1v0qlwvz58zF//vxy93FxccF///tfo55XCCGEqIvWrl0LR0dH9OjRAw669Cdd15lGjXiRkpenv7hSqRgUXLnC7ebmfG9m9vAheo6OTH9KTX14ECKEqbi58Wf4zz8ZBHTpwq5OFhb6fRo04PYdOxgIG9saWaUCnniCF/iHD3PFIj6egUZ4OLB0afVSAT09gevXDQ/My85m0NS3b72oYzI6sFCpVJg2bRqmTZuGew9689pLb2shhBDCpAoLCzFt2jQkJSXhyM6d6BATw4uu3bu5Q0iIvo1s8fkU/v688EpKAry9H164rWNtzYDizp16ccEj6rAWLZgO1aYN64ZUqrL7hITwZ/XoUaBxYwbRxlKpgM6dOdTuwAHWXFy/zgBj0qSqn79Gw5onQ/Ly2CparQb69KnztUzVGmdob28vQYUQQghRA27cuIGkpCRoNBq0srfn3U9HR3bFAfRFqLa2+hQRgBdYzZuzKFWrZcFrZedSeHtLRyjx6Lm7A8OHM8XPUFAB8MK8Y0emA0ZHV2/+hUYD9OwJvP46P//9d/3vmdLMzbnaePEiV0dyckzzPDWkUisWYWFh2LVrF5ydnREaGgpVef+p0PfXFkIIIYRydH9fQ0JCYB0byxWHnBwgJoY76AILR8eyqxGBgVzFSE3likVgYPkXaMU5OzNIqeN3UUU9UFEHMx0rK9Zb3L3LGqTqzikLDWWK0q5dwIoVwPLlVVsJeRgLC/5Onj/P77N3b3atqoMqFVgMGzYMlg++wWHDhlUYWAghhBBCebr5FaFNmrCA1ceHKRRaLVOi3N3ZdlY3y6I4BwduP3yYj7m5Ve5JXVz4Vk8KS8VjwNWVhd1btuhX9apj3DjgxAm2dv75Z2DUKGXOszRLSwYXZ85w9aVXr5J1JHVEpQKLuXPnFn38/vvvm+pchBBCCFEO3YpFWIMGQH4+L0R06RkhIXyv1ZbfRjYoCDh9misWlU1j9vIC+vc3zV1aIUylcWNO1j58mEFxZVY7ymNvz/qKjz4C1q0DunZlpzVTsLLisU+fZnDRs2ed+90zusaiUaNGSE5OLrM9LS0NjRo1UuSkhBBCCFFS0YqFjQ3vygL6+RXNmrEPv7l5yfqK4jw9edFibf3wjlDFVafVphCPSlgYV/WuX6/+sbp2BTp1YnOEzz83fhifMXTDLSMi2KK2jjE6sIiNjUWBgX/Q3NxcJCQkKHJSQgghhNC7c+cO4uPjAQBt7e1Z+6DVlhyMl5HBu7PFO0IVp1JxP09P4wILIeoiW1sWc+fnV3/YnUoFTJ7MY0ZGAr/9psw5lsfGhoP1jh6tc4P1Kt1udtOmTUUfb9++HY7FctYKCgqwa9cuBJYeviOEEEKIanN2dsa5M2dw6T//gaOzM9Mkzp1jsbalJVcibt7Ud3EqT2AgaybqaGGoEEZp1Aho1Yo1Ek2a8PemqlxdgRdfBFauZPvZTp2MH8ZnDGdnBkQHDnAVUrdKWctVOrB45plnAHCOxdixY0s8Zm5ujoCAACxdulTZsxNCCCEENBoNWri7o4W/PwuvCwuBf/+bD/bsyYF32dlM/aiISlV+qpQQ9Y1KBbRrx8JrXcOD6njySWDfPuDsWXaIeu8903ZM8/XlCsn+/cDAgRXfNKglKh26abVaaLVa+Pn54fbt20Wfa7Va5Obm4vLlyxgyZIgpz1UIIYSo97Kysgw/EBurH2538CA7QllZASNHMtAoLORqhBBCz8GBKVHZ2ZzhUh0qFfD3v3PF78IFYMYMTug2FZWKq4yRkcCRI0x/rOWMXhOKiYmBW2Xb1AkhhBDCKH/5y1/QuHFjTJw4ET/88AMSExPx1rRpWP7ZZ0g1N+ek3u++487PPstgQhdwyGqEEGU1acIhkdeuVW9wHsB0ww8+4Mrh9esMLvbuVeY8DTE350rLyZOmG9KnoEqnQulMnToVQUFBmDp1aontK1asQFRUFJYvX67YyQkhhBCPk4KCAhw6dAipqamIjo7G119/XeLxUU8+yf78iYkMKJ59lg9kZrItpgQWQpSlVgPt23No3q1b1a+NaNoU+OQT4OOP2Rp26VI2Uhg3zjTtYe3t+Tt+8CBrL7y9lX8OhRi9YvG///0PXbt2LbO9S5cuWLdunSInJYQQQjwutm3bhrVr1wJgLUVcXBy2bNmCN998E2FhYUVDaZt5ecHd0hL46Sd+4ciR+pzrjAxeLFWnX78Q9ZmLC9ChA4fm5eZW/3iOjsD77wMvvMDPf/8d+Mc/gKSk6h/bEC8vpnLt3QvExNTatCijVyySk5NLdITScXBwwJ07dxQ5KSGEEOJxsGvXLjz77LPIzc2Fp6cn+vXrB3t7ewwaNAiDBg0CAKTs2oUja9eieadOnPybkQH4+wN9++oPlJfHNrJCiPKFhLBW6coVpkdVl0YDjB7NFYxly1j3NHMm8NlnpplWHxDAmo6NGznwsnVrFnhXp9uVwow+k6CgIGzbtq3M9q1bt8qAPCGEEKKS9u/fj6FDhyInJwdDhw5Fz549y+6UmgqXmBgM6tYN/ioVsHkzt7/4on51oqBAuj3VMgkJCQgPD5f5XrWNmRlXLezsACVvhrdvz9QoLy8e95tvlDt2cRoNi7m9vYGoKGDDBmD7dtZ6VLd2RCFGr1hMnz4dr776KpKSktCnTx8AvOOydOlSqa8QQgghKuHw4cMYPHgwsrKyMHDgQPz0008wN5SbfeYMkJrKO6Iff8xhX23bcqqwTmZmxYPxRI1as2YNJk2aBK1WC7VajRUrVjzqUxLFeXqyBe3u3UxnUqomwssLeOMN4J13gB07gB49gDZtlDl2adbWQOPGTI26dIlD9Jo2ZetpCwvTPGclGb1iMW7cOCxduhRr1qxB79690bt3b3z//ff48ssvMXHiRFOcoxBCCFEvnDx5En369EHPnj2RkZGB3r17Y/369bA0NLDu5k3g/HmgQQN9L3uVCnjpJb7XycjgaoUpUi+EURISEoqCCoCt+j/88MNHfFaijFateGGudKvY5s2BwYP58cqVytRyVMTGhilRrq6crREXZ9rnqwSjVywAYMqUKZgyZQqSkpJgbW0NO3kxE0II8ZgqLCzEv//9b0RHR5d5rKCgAF26dMFTTz0FALC3t0d4eDgAoG/fvti4cSOsra3LHlSrBSIigJwc5lAvWsTtffowFaK4zEygRYuSwUY9lZCQgMjISAQHB8PX1/dRn04ZkZGRRUGFTunPRS1gYcHZFomJXBFUcrVvzBjg6FEee+1adooyNTs7pnlduMCA6RHWXFQpsMjPz8eePXtw9epVjBw5EgBw48YNODg4SJAhhBDisbJlyxa89NJL5T4+fPjwosAiKCgIa9asQdeuXdGkSZOijk9lxMWxwNTHh6sWFy7wYmjUqLL7arXsqV/PlU4xWrVqFcaPH/+oT6uE4OBgqNXqEsGEuhYV1opifHyA0FCuBNrb88JcCTY2wJQpwIIFwKZNQPfuQHCwMseuiKcnXzeUmDBeDUb/K8bFxWHgwIGIj49Hbm4u+vfvD3t7eyxZsgS5ubn46quvTHGeQgghRK00ePBgrF69Gl9//TXatWtX5vE2xfKsVSoVxj3sDmZeHnDqFFcgbGyAP/7g9l69ygYQubkMOOp54bahFKPJkydjwIABtWrlwtfXF6tWrcLkyZNRUFAAjUaDd999F7dv337UpyYMaduW6VDXrpVdCayO9u1Z77B3L/D55+wYpVTgUh4bG752REbWrcDi9ddfR7t27XD69Gm4uroWbX/22WelxkIIIcRjR6VSYcKECZgwYYIyB4yKYp/6gADWT/z5J7f3719239RUwMGh3hduG0oxKigoQFRUVK0KLABg/PjxGDBgAKKiohAUFASNRoNVq1Y96tMShlhbMyVq0yYgPZ2/S0qZMIHTsmNjgfXr9fMuTMndnSudbds+spsNRq/P7d+/H7Nnz4ZFqarzgIAAXL9+XbETE0IIIWqzkydPIj09XdmDpqQAJ07w7qOFBbBvH3D/PuDnV7bvfm4uh32FhQGGir/rEV2KUXEajQZBQUGP6Iwq5uvri169etW6oEcY4O/P7k3Xr7PrmlIcHQHdDfcff+SqiKk5OfE1wUC9V00xOrDQarUoKCgosz0hIQH29vaKnJQQQghRm6WlpWHIkCFo3rw5zp8/X/0DpqcDR47wzubNm+wEBQA7d/J9//5li7Pj4xlshIRU//lrOV2KkebB7A6NRoN//etfcuEuqk+lAp54gt2VIiPZDEEpPXuytW1+PudcKH0jojTdPJsLF0zfkaocRgcWTz75ZIl5FSqVChkZGZg7dy4G61psCSGEEPXY22+/jZs3b8LW1pbDYW/fBrZt491CY2RkAMePA+vWcXVCo2E/eo2G6VBRUczN7t275NclJwO2thz2pVQf/lpu/PjxiI2NRXh4OGJjY01SuC2D7R5TdnbAoEGsjbhxA0hKUua4KhULuW1s+Ls8bRqDF1NycwNu3WIK1iNgdGCxdOlSHDx4EM2bN0dOTg5GjhxZlAa1ZMkSU5yjEEIIUWvs2bMHq1evBgCsXr2a7WIvXmSLyZ07K3dXMjeXBdrr1gEP2s+iSRNeFOhWJnbs4PuOHUvmfufl8cLniSc4lOsxYsoUozVr1sDf3x99+vSBv78/1qxZo/hziFrM2porDP368XcsJoZT7avL3R1YvJirkElJwMyZvAlhqknZZmZMo7x4kR3japjRgYWvry9Onz6Nd999F9OmTUNoaCgWL16MU6dOwcPDwxTnKIQQQtQKWVlZRY1KXn75ZfTo0YOrFFeucN5EdDSDi4yM8g+SkgJs38798vIYULi7l+w9f/8+sGcPPy5dtB0fDzRqxCFfQhHldZ2SlYvHjFoNtG4NPPUUfycjIzlLproCAtgZqlMnpkV98QWwfLnp0pU8PPg6ceOGaY5fgUoFFmFhYUhNTQUAzJ8/H/fv38eoUaPw0Ucf4YsvvsCECRMMD/gRQggh6pH3338fUVFR8PHx0a/SR0cDaWlcbWjcmJ/v3m04Vzs6GvjtNwYijRpxxcHQnINDhxicuLmxsFQnLY2pTx061MqC7bqaSlRR1ymg7n5fomReclYAACAASURBVIp8fRlctGzJlCIlggtbW2DWLODFF/k7Hx4OvPWWaS7+bWwYwJg67cqASgUWFy9eROaDF8h58+Yho6I7MUIIIUQ9FBMTg6VLlwIAvvzySzg4OPCO4/nzTFVSqXjR36gRcPkyg4usLH5xfj5rKbZuZcAQHMx0hfLoirb79WO9BcC0jMREtpJs2NCE32nV1OVUooq6TtXl70tUg4MDf/9atODdfyVSl1Qq4LnnODzP0ZFBy1tvMe1KaW5uvIHxYGGgplRqjkXbtm3x0ksvoVu3bigsLMTHH39c7oTt9957T9ETDAgIQFxcXJntr7zyClauXIlevXph7969JR6bPHmyDOoTQgihqNzcXIwYMQI5OTl4+umnuTE+noWSJ0+ybWXnzvrg4tIl3pns0IGPnz3L9AoXl4qfKDEROH2aFyF9++q3JyQwoAgNNd03WUUPG2CXkJCAyMhIBAcHm6Q+orrHNzTY7l//+hcA1InBfMJEdKuDuoJupVL+W7ViKtTChVxVeO89fqzkDQMnJ742RUezHquGVCqw+PbbbzF37lxs3rwZKpUKW7duhZmBCYIqlUrxwOLYsWMl2tueO3cO/fv3x1/+8peibRMnTsT8+fOLPrexsVH0HIQQQohmzZph7dq1+g1aLQsko6OB//6XRZNffMH0JgsLTvI9f54XJElJDDwqkza8axfft24NeHry48xMPl+HDkxzqGUqSiXavn170cW5Wq3GqlWrFO3otGbNGkWOX3qwna+vL8LDwysczGfqgEnUAm5ubBm7cycv1itaaTSGqyswbx4wZw5w9SrfL1wIeHsrc3xd69lz55g26ebG5zRxFzlVYaFxaztqtRqJiYmPrFD7jTfewObNmxEZGQmVSoVevXqhbdu2JVrgGis9PR2Ojo64e/cul7aFEEKIh7l+nXMntm8Hfv+d27p3Z2qDTm4uW8N6eupTmipSUMChWnfuAG++CfTowe1xcRyS9/TTZedZ1AIJCQnw9/cvcRGu0Whw6NAhdOrUqcz22NhYRS7Ey3vemjh+ZQOmmzdvYtWqVZg0aRIa6OaTiLolLw/YsoU3EZQeypieDrz7Ln/H3dwYXCjV7a2ggIP5cnMZEDk4sH7E25urp+7ulTzFyl8nV2lA3qMKKu7fv4/vv/8e48aNg6rYC+vatWvh5uaGli1bYtasWcjS5bSWIzc3F+np6SXehBBCiPLExcXh7NmzKHEvLjKS3ZvOndNv27+f9RU6lpb8I16ZoAIAIiIYVNjbs4MMwNzu7GymV9WSoKJ0MXN5A+wyMjIqLIqurocVXVdXed8XYDhFSoq76yldSpS1NW8UKMnBAZg/nxf8d+5w5UKpORoaDTtSNW3K1yGtlq9XW7YA//sfcOyYstPGUclUqNIiIyMRHh6O27dvl/mFVjoVqrgNGzYgLS0NL774YtG2kSNHwt/fH97e3jhz5gxmzpyJy5cvY/369eUeZ9GiRZg3b57JzlMIIUT98uWXX2LJkiVF9X1IS2NhpJkZ7zSqVByudfQo8M03wKJFFQcBhYWGi0F1syt69dKnXGRmcoBXLbnbXV7qkaFUooSEBKjV6jJ3/IMUuuurK7o21fGBqqVIiXqoQQMgLAzYu5fBgJIpRc7OLOj+xz+Amzf1aVEPq8cyhqVlyVWKlBR+LykpQJcuvJmhAKNToVavXo0pU6bAzc0NXl5eJVYOVCoVTp48qciJGTJgwABYWFjgt99+K3ef3bt3o2/fvoiKikLjxo0N7pObm4vcYr2D09PT0bBhQ0mFEkIIYVBoaCgiIiLwn//8B6NHj2Yx9q5dLKj+7DN2eZo1C3j5Za5izJrFQm5D4uN50VBRm8lPP2WNhm5/b2/gmWce+YpFVVKP1qxZU6YoWpcypESNQkXHr+j7qM7zGvPvIKlQ9UhuLttFX7/OFUSlJSXxteP2bRZyf/gh6yRMJSeHnal8fZl26eNjcDeTpkJ98MEH+PDDD5GYmIiIiAicOnWq6M2UQUVcXBx27tyJCRMmVLhfx44dAaDCZVBLS0s4ODiUeBNCCCEMuXXrFiIiIgAATz75JC8uLlxgu8hTp7hTaCjzo595hp//+9+GUwyuXwdmz644qHjiCX1QAbBlbePGJg0qKjunoSqpR+PHj0dsbCzCw8MRGxtbdNGvVBvX8o5fHiWet7wUKVmtqOcsLZkSZWbGwZhKc3cHPviAryXXrrFb1L17yj+PjpUVb4rcugVs3sw0qWpO6zY6FSo1NbVER6aa8s0338DDwwNPPfVUhfvpXvzlroAQQggl7HiQnhQaGsoawytX+IfY3581EXyQ7597jsXcN24A27YBQ4boD5SYyKAiLY15z7Nn8w97acVTErKy2AXKhH/TjOmqVNXUI19f3xIX3Q9rT2us0scvj5LPayhFSjwG/Pw4tPLPP1nU7eCgXKcogIXburSo2Fhg7lx+bmur3HMUp9HwxsWtW0zFvHMH6Nq1yqleRq9Y/OUvf8Eff/xRpSerKq1Wi2+++QZjx44t0eb26tWrWLBgAU6cOIHY2Fhs2rQJY8aMQY8ePdC6desaPUchhBD10/bt2wEwHRdaLVcrzM2ZonTvHgs6mzblzjY2wMiR/PiHH/TTt5OSGEgkJzPFYcEC9sR3cCj7VnxlIiWF+7m6muR7K+9Cu7yVC6Xu1Ju66LqmntfX1xe9evWSoOJxExbGFrQqFVchL1/mkLvkZAYb1eXjw9cIBwcgKoptaR/SmKjaPD0Z1Jw4wTbaVWT0ikVQUBDmzJmDw4cPo1WrVjAvFdFMnTq1yidTnp07dyI+Ph7jxo0rsd3CwgI7d+7E8uXLkZmZiYYNG2L48OGYPXu24ucghBDi8aPVaotupg0YMICFlfHx/COsq/dr04YBREwMp/Q++SQfS0gA1q3jqsW77zJv2ttbP3W3MjIz2R1KbfR9wEqp6EK7vItlJe7UP2zlQ6n5EKWPUxPF3uIxYGMD9OnDeqrkZN7lv36dK5UxMUxlqm7htZ8fXyvefZfDNhcsAN5/n+lYpmJnx9emY8cY3FThhobRxduBxfM+Sx9MpUJ0dLTRJ/GoyRwLIYQQhpw6dQphYWGwtbVFSkoKLA4eZF1FcDDwzjtcvZgyhQGFhQVXNLy92R3qgw+4suHhwYsOT092i3Jzq9yTZ2czGHn+eeX62pdi6jkQFSmv6FqpgXflHacqxd7VIcXbj5GcHA7F3LePF+XOztU/ZmQku0RlZfEmxpw5yqZelVZYyHTPli2B/v0Bjcao62SjA4v6SAILIYQQhuTk5GD//v2Ii4vDhOHDgZ9/ZuqTpSUwahQHUK1axSFXwcFMiWjShCkSs2cDZ8/yQG5uDCp0k7Qr48YNpkK88ELl52BUQU1faBeXkJBQpj2tEoHOw45T+nlNSQKLx4xWCxw/Dhw4wJsKlV2drMjFi6y1yMlhCtasWaadoJ2dzRXXQYOAkBDTdoUSQgghHhdWVlbo378/OxJGRzOAcHICzpxhUOHtzYsHgKkLzs5AaioDi5de4p1FFxeuXhgTVACs3wgONmlQARjfVUlJpWsUlKqBeNhxpDZCmIxazYv/Ll1YEK3EEOaQEHaIsrBg0PLPfyo+2K4Ea2sWix89anT3q0rVWEyfPh0LFiyAra0tpk+fXuG+y5YtM+oEhBBCiFpPl+Lg5MSgQddePTSUd/dsbBhYpKczP9nVFQgKAr78kqsOxuZF5+byIsLbW/nvxYDKdlUyNaVqIKSWQjxSajXb0hYUAIcO8TWjugPoWrZkvcUHHwCHDwOffAJMn266Gw9eXkyJOn6cgVIlVSqwOHXqFPIeVLmf0vXsNkD1iAf3CCGEEErZv38/NmzYgGeffRbd3N3Z2alRI+YgF59fkZGhL3ps0oTpT5mZvOOnm3JrrNRUBie61ZDHhK7rVOnULGODHqWOI0SVqdVsvKDVAkeOsBjazq56xwwNZW3XokXA/v1Mh5o61TTNHdRqDs47d86oQvRKBRbh4eEGPxZCCCHqq19++QWff/45MjMy0G3AAK46mJmx9uHWLX7cqhU/DwzknUNPT863iI4uOeTOWOnpPLaZ0c0b6zyl5kPInAnxyGk0QOfO+roLBwfeLKhOINC+PfDmm8BHHwG7dzO4eOUV0wzQtLNjy+vjxyv9JVJjIYQQQhhQNL+iXTsWMupWD3SrFSEhzEUuKNA/plJxe35+xf3stVqudBjqn3L/Pi9IfHwU/G7qFqVqIKSWQjxyZmast+jTh0HAlStc/azOhOsuXYBp0/h6s3078PXXhl9LlODjwzbblSSBhRBCCFFKbGwsrly5Ao1Ggz4eHgwUrK35YPE0qPx8BgHF20r6+XFS9u3bhg+u1QJXr/JO4JUrTHsqTpcGZaIWs0KIGmZuzteL558HevfmtitXOAOjqgFBz55MgwI4N+fbb00TXOhWYitJAgshhBCiFN1qRef27eGYmKivlcjL07eQDQ3V11I4Oem/2Nyccy0yMsreldRqmSbl4cFWju3bsz/95cv67jF377KWw5TtJIUQNc/OjoXQzz8P9OjBGxNXrlR9WnffvkyDAoBffwV++EG5cy3OiNciCSyEEEKIUnSBxZNt2vCCX9eL/tIldoFydGQNRUYGgwpb25IHCAzUt57VKSzUT+Xt35/BQ8+ewPDhwBNP8HkuX9YXTVZDQkICwsPDkZCQUK3jCCFMwMEB6NiRv/sBARygWVUDBwITJvDjH3/krJ1HSAILIYQQopi8vDzs2rULADDA01PfYhbQt5lt25YBQFYWc5BLF07a2wPNmgF37vBzXVDh7Az061cytcDdnekRzz7Lybq+vtVKg1qzZg38/f3Rp08f+Pv7Y82aNVU+lhDChJydufJZUMAbFlU1dCgwdiw//v57YONGZc6vCowOLPz8/DBmzBisWbMGV69eNcU5CSGEEI9Mamoq/P394erkhCdKt4zV1VeEhfF9YSHrIQwJCmJdRkYGEBvLu5T9+rH+whAvLz4+bJjxcy8eSEhIwKRJk4rmN2i1WkyePFlWLoSorQIDOQizOqsWAFc/Ro7kx2vWAFu2VP/cqsDowGLhwoWwsrLCkiVLEBwcjIYNG2L06NFYvXo1IiMjTXGOQgghhOnk5vJu4YM3D3t7HN+3D7+9/TY0trb6AVS3brE+AuCKRW4uA4DihdvFeXgwzSEykqlS/fo9vNOTSsXBeFWk1ORqIUQNUav5emJhUf0p3X/9KwMMAPjqK2DHjuqfn5GMbpA9evRojB49GgBw8+ZN7N27F5s3b8Yrr7wCrVaLgoICxU9SCCGEMImbN4GdOxkkFGMBoLO5ub6NbF4e8M9/8uMWLRhMJCeXLdwuTtd69u5doGtXoGFD030fD8jEaSHqIB8foHlz4MQJplFWdSaFSgWMGcPXq02bgBUruK1vX9PMuTCgSpN3srKycODAAezZswfh4eE4deoUWrZsiV69eil8ekIIIYQJJSbyrUEDbDxxAmeuXcPMIUNgYWbGWgcrK+73f//H7i22tsDrr3NbRgbTGCpaYQgMZD1F6eJuE5GJ00LUUW3aAFFRbENdXnplZahUwPjxDC62bgU++4w3T0aPBlq2VO58y2F0YNGlSxecOnUKISEh6NWrF9555x306NEDzuUtBQshhBC1UWEhax9sbZGq0eDlf/8biampsHNywrRhw/T77d0L/P47P54+XV9YnZNTfr2EjkpVY0GFjkycFqIOcnEBWrcG9u3jimh1pnOrVMDkyVz9+PVX4MIF4B//YPAyahQbS5iI0YHFpUuXYGtri2bNmqFZs2YICQmRoEIIIUTdk57OIXaOjnjrm2+QmJqKpj4+mDJokH6fuDimEwDACy9w7gTAeRRqdfn1FY+Yr6+vwYAiISEBkZGRCA4OloBDiNqmZUu2tL516+E3LR5GreYqxaBBbEG7Ywdw+jTf2rfnY4GBypx38ac19guSk5Oxe/dudOrUCdu3b0fXrl3h4+ODkSNHYvXq1YqfoBBCCGESSUnAvXvYHR2NNQ+KHL9+7TVY6VKbsrKAxYtZf9G2LfC3v+m/NiuLHZ/Kq6+ohaQNrRC1nK0tZ9rcu1d2aJ5Wy+1JSWUHb1bE1RWYMgX44gvWWqjVwLFjwIwZwObNik/rNjqwUKlUaN26NaZOnYp169Zh69at6N+/P3755Re8/PLLip6cEEIIYTKJicjKz8fEL74AALwyeDC6NW/OxwoLmZt8/ToH2r35pr47FMD6CgcH/eC8Wk7a0ApRRzRpwkYPN24AmZlsMHHlCusv7t1jUHHtmvHH9fJifdiKFUCHDpz6vWoVsHQp0zoVYnRgcfLkSSxbtgxDhw6Fq6srOnfujDNnzuC1117D+vXrFTsxIYQQwmTy84HYWMwND0d0YiJ83dywaMwY/eMbNwJ//gmYmQEzZzKIKC4zk51cqpMHXYOkDa0QdYSFBefkqNVAairrJDp35nybv/5Vv+qQnFy14/v6Au++ywJvtZo1HW++CSh0k8HoGosOHTogNDQUPXv2xMSJE9GjRw841pE7NkIIIQQAIDkZMVFRWPYgBeqrKVPgYGPDFYoffgD27+d+EyYATZuW/fqCAq5k1BHShlaIOqRRI+CppxhUODmVvIHh5AR07Ajs2cN0TBsb44+vUjFQCQpiG+34eKZGTZ3K1tjVYHRgkZKSAofSd26EEEKIuiQpCf42Noj49FNcuHYNT/n5AZ9+CoSH6/OXn36ahY+l5eUxLaqWFm4bIm1ohahD1GrA37/8x9u25YpFRARTp4qnaRqjRQvgk08YXJw7ByxZwoBj7Fiu1laB0V/l4OCAtLQ0rFu3DlevXsVbb70FFxcXnDx5Ep6envB52FRRIYQQwkRycnJgaWkJ1cOGQcXHQ21lhVb29mh17hz/uObn87EOHYCRI3nX0JDMTMDOrk4VbgPShlaIekOj4cpCWhpbZjduXPVjOTsDCxYA333H1rQbNwKRkcDbb7MFrpGMTg49c+YMgoODsWTJEnz88cdIS0sDAKxfvx6zZs0y+gSEEEIIJURERMDHxwevvfZaxTtmZbEgsrCQS/9btzKoaNuWd+5mzy4/qABYuO3sXOPzKZTg6+uLXr16SVAhRF1nawt07873iYnVO5ZGA7z0EjBrFtOrLlwApk3jKoaRjA4spk+fjpdeegmRkZGw0k0kBTB48GDs27fP6BMQQgghqis1NRXPPPMMAgICMHz48Ip3TkpCfGws/t/nn+Pbe/fYL37hQmD+fMP1FKVlZwPe3sqcuBBCVFWDBly5yMxkx6jq6twZWLaMaVipqbzJ8uuvRrWkNTqwOHbsGCZPnlxmu4+PDxKrGzEJIYQQRtJqtRgzZgzi4uKQkpKCtm3bVvwFt25hb1QUvr96Ff8CWEvRsmXJfbKygKtXeScwK6vkH9bCQvaGF0KIRy0khF2kEhL4WlVdPj5cue3Vi/Vm33wDfPllpb/c6BoLS0tLpKenl9l+5coVuLu7G3s4IYQQoloWL16MzZs3w9LSEv/73//g7OyMwsJCw3UWWi0QG4t9V68CAHoAQKdOZfeJjweCg/mHOjmZqxRqNdMErKzqVOG2EKIeU6nYJSonBzhzht3qqnvjw8qKqVDNmgFffw2cOFHpLzV6xWLo0KGYP38+8h5MBFSpVIiPj8fMmTMfvvwshBBCKGjnzp2YM2cOAGDlypXw8vLCs88+ixYtWqDQ0PJ9WhqQnIx9ly8DALp7e5dtG5uYCLi7/3/27jwsyqp94Ph3Zth3FBEQQUwUc88t1NyNzDY1MzWz3KoXy+Vt821RKzNbf5W2uKRZmlkupbmkgJq74kpuoCKg4IIKuADCnN8fB0bZFJQl7P5c11w6z5w5zxkeL3nuOefct/7G7skn4YkndOrHe+7R65mrVat0G7eFEHcwW1td36JTJ/1lyPHjJavOXRiDAR58ECZNKtEXKSUOLD755BMuXryIp6cnV65coUOHDtSpUwdnZ2cmTpxY0u6EEEKIWxIfH0+/fv0wm80MHjyYIUOGUKVKFVatWsWBAweIKmzj4enTJCUmcjg1FQPQtkOHvK+np+v1yq1aXcsf7+Ghv7nr0AH69YOHHgJr63L5jEIIUSwmEzRvroMBNzed2Skz8/b7rVcPxo0rdvMSBxaurq6sXr2apUuX8sUXXzBixAiWL1/OunXrcCzlDBnjx4/HYDDkeQQFBVleT09PJzQ0lKpVq+Lk5ETv3r05depUqY5BCCFExUhNTaVTp060adOGN998k/DwcNLT0y2vb9++nZSUFJo1a8aUKVMAsLOzo2PHjgCsWrWqYKcnT/JXdDQAjQH3nLaA3jsRF6fXLBe1idvK6tYKUgkhRHmoVUvPsNarB0ePQiHbF0rM2bnYTW+t+gXQrl072rVrd6tvL7YGDRqwZs0ay3Or6wp2jB49mj/++INffvkFV1dXRowYQa9evdi4cWOZj0sIIUTZWrRoEWvXrgVg8+bNTJw4EVtbWx544AGWLFlCr169+Ouvv6hWrRr29vZ66n/vXkK8vFiBDixefvnlax1mZkJcHOv37AGgvYuLzqqS68wZcHWFli1vveCUEEJUNHd3CAnRdSi2b9dFPcsp4cQtBRZhYWGEhYVx+vRpzPnWcH333XelMrBcVlZWeHl5FTiekpLCzJkzmTdvHp07dwZg1qxZ1K9fny1btnBv/s14QgghKpVly5YBem+fi4sLYWFhJCYmWvb4AbRu3Vr/5fJl2LwZdu8mJGfPxF/r13P54kUcnJx0mzNnICWFM0lJGID2jRtfO1lmpk6v2LVrwT0XQghR2djYQJs2OuHE+vV6WWc5JJ0ocWAxYcIE3nnnHVq0aIG3t/fNq5vepujoaHx8fLCzsyM4OJhJkybh5+dHZGQkV69epWvXrpa2QUFB+Pn5sXnz5hsGFhkZGWRkZFieF5blSgghRMVq06YN8fHxvPnmm7Rs2RKlFIcPH+bKlSt5G54+rX9xHj0K/v7UCwzEb+ZM4s6dY93UqXQfOVJnOTlzBtLSmJ+WxteAba9e1/o4fhzq1oUGDcr1M96uhIQEoqOjCQwMlKJ3Qoi8DAZo1kzP5v71lz5WxsFFiQOLb775htmzZzNw4MCyGE8erVu3Zvbs2dSrV4/ExEQmTJjAfffdR1RUFElJSdjY2OCWLzNH9erVb1pPY9KkSUyYMKEshy6EEOI2jRkzhjFjxlieGwwG6l2/90EpOHwYNmyAlBSdHtbKCgMQ0qIF0//8k1VLl9I9MBDat9fBw5EjkJWFe40acNddup/kZJ3tqVWrSrUpe+bMmQwfPhyz2YzRaGTatGkMGTKkVM8hgYsQlZzBoDPaZWfDxo165sLVtcxOV+LAIjMzkzZt2pTFWAro3r275e+NGzemdevW+Pv7s2DBAr2e9haNHTs2zy+r1NRUataseVtjFUIIUU7MZh1I7N+v86vb2kKdOvoXaI4HW7Rgf3w8dzdsCLGxuv3Vq6ioKAygK8waDHrt8ZkzOuNTIctu/6kSEhIsQQXoIoHPPfccISEhpRYAlEfgIoQoB0aj3jtmNuslowYDuLiUzalK+oahQ4cyb968shjLTbm5uVG3bl1iYmLw8vIiMzOTCxcu5Glz6tSpQvdkXM/W1hYXF5c8DyGEEP8MSil+//13UlJScg/owODIEf1L8ZdfYP58/e1blSrg45MnqAB47N572TB5MsN79NBBR3o6nD5Nh8hI2gN7/P11wzNnwNcXmjQp3w95m6KjowvscczOziYmJqZU+i8qcElISCiV/oUQ5cxo1IX0WreGxERISyuT05R4xiI9PZ1p06axZs0aGjdujHW+aeNPP/201AaX38WLFzly5AgDBw6kefPmWFtbExYWZinMd+jQIeLi4ggODi6zMQghhChbBw8e5NFHH8XJyYmzhw9ju3mzDgAuXtQBhKOjztNeo0aBgKJQBgP4+nLxyBE2KUU24F6/vn4tLQ1atNCzHrehvJcMBQYGYjQa8wQXJpOJOnXqlEr/NwpcZEmUEJWU0Qj33quXRW3bppdEeXrq46WkxIHF3r17adq0KUCB4kOlvZH75Zdf5uGHH8bf35+TJ08ybtw4TCYT/fr1w9XVlSFDhjBmzBiqVKmCi4sLL774IsHBwZIRSgghKrHcbFBt27bF9tAhiInRQYSXV4l/AZ5LS+NgQgJt6tdn8+rVZAP+9vb4eXrqTFL29nrG4jZUxJIhX19fpk2bxnPPPUd2djYmk4lvv/221G76yzpwEUJUEJNJZ4tydYXdu+HQIZ2K1sOjVAKMEgcWERERt33S4kpISKBfv34kJydTrVo12rVrx5YtW6hWrRoAn332GUajkd69e5ORkUFISAhfffVVuY1PCCFE6csNLB7u1En/0vP2vqX1wPtiY2k6ahQu9vacmTWL9QcOANA+t9BqcrIOVnJ+p9yK8tjrUJQhQ4YQEhJCTEwMderUKdXzlXXgIoSoQFZWevlnnTpw8CDs2aMTYXh46CDjNiYKbrlAXnmYP3/+DV+3s7Nj6tSpTJ06tZxGJIQQ4rakpsKxY9CoUaHfjp07d85S5LRH7do6k5OPT8F+lNJ/3uAXYP2aNXGxt+fCpUvsWLWK9Tn1L9q3aaPff+mSziR1G9/SVfSSIV9f3zI7T1kGLkKIfwBHR2jeXKfa3r8f9u3TX+b4+kJu/Z8S+kcHFkIIIe4gSuksTvv26T0NuTMH11m5ciXZ2dk0rF+fWufPF5xNuHgRfvsNli3TU/qNGkHjxvqRbxO3lclE1yZN+HXTJn4PD2drzvH2DRvqfpycbnsZ1J2+ZKgsAxchxD+Es7Pe1B0UBDt3wq5d1/ZflJAEFkIIIcqFOnGCubNnU8/NjZZbt+pAIN8SJ8syqJYtdSao3LoVV67A0qWweLGeaci1caN+gJ7Cb9w4TzASkpnJr8CkY8cAqO7kRKCPj05BW7v2bReLkiVDQog7W1IuUAAAIABJREFUhqurTr3t4QGbNunZZT+/EnUhgYUQQoiyZzYT9v33DJwzB0c7O7aFhnJ3zZrQqZNlliE7O5uVK1cC8JC3tw4QMjNhxQpYuFAHGqB/0fXvr4OCvXv14+BBvWci3z7AkOv+3t5kIig4GINSkJGhC+SVQtIRWTIkhLhjGI16JtjdXVfrjo4u0bIoCSyEEEKUvWPH+PG33wC4lJ5Or3nz2F6tGs7+/pYK2CaTiV27drHy669p7e6ulzq9+CIkJek+vL11QNGunX4NoH596NtXBwoHD0JUVJ4ZjZpA/bVrOXDxIi8+8QSP9+sH589fS1dbSsp6yZBUwBZClCtfX+jRQ89cbN168/Y5JLAQQghRtq5e5cqWLSzauxcAZ3t7/L28yASdS93LS28iBPyrVOG5unX1+5Ys0UFFlSowYAB07nwtoMjP1lZnOSmk0N39BgMHli7lz+RkHgc4dw7uvrvMKs+WNqmALYSoEC4u0KUL2NgU+y0SWAghhChbhw+zbOVK0jIy8Pf0JOK99/CrVg0T6Gn2Xbv0LEROW0sa2BUr9LHQUGjZ8pZP/3TnzjT09yekWTNdGCo7GwICbvtjlYeKTGcrhBBYW+siosUkgYUQQoiyc+kS7NrFkkOHAOh3330EeHlde93bm9MbNpCmFC9NnEhPHx+GtmkDy5frDdsBASX6pVaYe+66i3tylluRnKzXDpfiMqiyVNHpbIUQoiQksBBCCFF2DhyAkyeZ9cor9Nu3T2/YzpGemcmLP/zAH1u28HSnTixftYr0OnUY2rWrzgAFev9EKWywtjh3Du65BxwcSq/PMnSnp7MVQtxZbr92txBCCFGY8+dh927w8MDGzo6HWrak9nWzFVnZ2Ww6cIDE1FQm52zsfqhZM1i5Us901KwJ995bsN/sbJ0h6tQp3S63WN7N5BTIo1at2/xg5Sc3na0pZ2+JpLMVQvyTSWAhhBCibOzdC+fPo6pWLfRlJ3t7Fo4di5O9veXYQ8HBugAewOOP69SHZrOu2J2QoKvCHj2ql0nZ2uoZiMOHISZGBxpXrhQ9nvPndX52b+/S/JS3JCEhgYiICBISEm7adsiQIcTGxhIREUFsbKxs3BZC/GPJUighhBCl7+RJ+PtvTtrZcd/zz/Pkfffx7oABGM+f1znRbW0BCPL1ZdZLL9Fn8mSa1a5N4N9/6yDCywvat9d/P3lSZyfx8ND7LapV0393cNDBwtmzOntUXBycOQPp6Tp7lIuLfuSci/PnoW3ba8/LQWFpYm8ly5NUwBZCVAYSWAghhLgtSimysrKwtrbWB65e1WlkMzL4+e+/OZqUxLqoKIyLFsGPP+r0sa+9Zqmq/Xjbtuz94guqOzrCK6/oPh5/XAcHiYk6I1SzZjpIyL/fompV/ahXD7KyrgUaiYk60Dh5UhfZs7bWj+v2eJS1wgKIkJAQyfIkhLhjSWAhhBDill2+fJn//Oc/ZGRkMG/ePAwGgy5Sd+QI1KrFvK++AqD/5cswZ45+09mzMHYsDB0K3buDwUCjWrV0Jqhz5/RsRKdOeh+FkxM0aACurjcfjJWVns2oVk0Xzrt6VWeBOnv2WoBxfUaqMlRUmth58+ZJlichxB1LAgshhBC3JD09nb///pu5c+eSlZVFmzZtePHJJ2HHDnB35/CZM+yIicEE9Dl+XM8YDB4M+/bpaq7ffKOzRoWG6qBg4ULdca9eum1S0rWlT7fC2loHEl5e0LBhqX3u4igqTazBYLhhliepsC2EqMxk87YQQogSi46OpmrVqnz66ad89NFHAIwZM4ZN33+v90VUq8a82bMBuB+o5ukJkydDjx56GdTgwXpj9rp18PLLMH++3h/h5gbduuk+HB31zEMllJsm9nomk4ng4OAiszzNnDkTf39/OnfujL+/PzNnzqyIoQshxC2TwEIIIUSJLVy4kMuXL5OcnMzIkSPp27cvWVlZ9Hn3XU47O6OmTmXe1q0ADAgIgM8+g9zaCwYDPPYYTJyoi9XFxcGCBfq1nj315upTpyAwEKpXr6BPeHtulCa2sCxPRS2dKk7WKCGE+KeQwEIIIUSJLVq0CIDevXtjMBiY8dFH1Pf25mRqKk+OH8/WVauIBuxNJh59/31wdi7YSYMGOuC4+2793NkZHngALl7UwUWDBuX3gcrAjdLE+vr60rFjR8typxtV2BZCiMpC9lgIIYQokbi4OLZv347BYODRRx8FsxmnAwdY2L8/raZMISIxkY+B4fXrY6pVCydHx6I7q1IF3nsP/vpLF66zt4foaB1UlNNG65Io6R6I4qaJlQrbQog7gcxYCCGEKJHc2Yp27drh5eWli9Pt30/9Fi2Y6eWFB/BcrVp8O2kSX73wws07tLLSWaACAnQlbWtrHVjkTy1bBkpSqK4s90BIhW0hxJ1AAgshhBAlcv0yKC5e1DUrbG1hyxaeOH6cGGtrur3+ut6cXVJJSXDXXVCjRimPuqCSBArlsQdCKmwLISo7CSyEEEIUW1JSEhs2bACgZ8+eEBurgwEHB8i5MXft3x98fEre+eXLepaiYcMyn624WaCQfyajvPZA5N97IYQQlYnssRBCCFFs1tbWfPDBBxw6dAi/mjVhyRK9L2L6dD17cdddOuPTrUhKgtq1oRxuqm8UKKxatarQitmyB0IIIW5MZiyEEEIUW9WqVXn11Vf1sqHcitZHj8LGjXrp04svgsmkq1wnJkJaGuS7gS8gt0K2Unq24laWUJVQUXUmHB0dC53JAGQPhBBC3ITMWAghhLg1CQk6IMgphEfPnnrGAeD4cfDw0IFFYqI+5uQELi5gZ6ePp6VBeroORFxcoEkT8PMrl6HnbpZ+7rnnyM7OtgQKFy9eLHImY8iQIYSEhBATE0OdOnUkqBBCiHwksBBCCGFx+PBhqlevjqura4HXIiIiOHHiBA899BBuzs5w+DCsXg3nzuk9FU8+qRuePaurZoeE6CDi7Fk4fVrvx7hwATIydJDh5wc1a+oAxMNDty1HhQUKCQkJN1zyVNz0sUII8W8kgYUQQgiysrJ48803mTx5Mp6envz888907NgxT5vPPvuMpUuXMm7cOMYPHw6RkZCzkZsRI3RmqKtXdSDRufO1qtmurnrvRevWOrC4fFlX3L5RfYtykj9QKGomQ4IJIYS4OQkshBDiDnHlyhXs7e1L/L5Tp07Rr18/IiIiADh9+jRdu3Zl7dq1tGvXDoC0tDT+/PNPICfNbHz8taCibVu9NwIgLk4HEY0aFTyR0agL4lWpUvIPV45kyZMQQtyaf/Tm7UmTJtGyZUucnZ3x9PTkscce49ChQ3nadOzYEYPBkOfx/PPPV9CIhRCi/CmlmD59On5+fuzbt69E7920aRP33HMPERERODo6MufLLxn4wAN06dyZ4OBgS7vly5eTkZFBYGAgDevVg7//ht279YshIfrP8+fBxkbPTNjYlNbHqxCS9lUIIUruHx1YrFu3jtDQULZs2cLq1au5evUq999/P5cuXcrTbtiwYSQmJloeH374YQWNWAghKsby5cs5e/YsQ4cOJTs7u9jvmz9/PidPniQoKIjt69Yx0NOT7zt3Zsn771syIKWnp/Pee+8B0KtXLwxJSRAeDleugKcnNG4MWVk6XWyzZuVS3E4IIcQ/zz96KdTKlSvzPJ89ezaenp5ERkbSvn17y3EHBwe8vLzKe3hCCFHhlFIYDAamTp1KREQE27Zt44svvmD06NHFev/HH3+Mu7s7L48ejfO2bXD8OAZnZ+z37tUZnqpUYdiwYURFRQE5y6Di4mDLFt1B1656idPx41CrFjRtWqqfLyEhgejoaAIDA/PMHhR1XAghRMX5R89Y5JeSkgJAlXzrc+fOnYuHhwcNGzZk7NixXL58uSKGJ4QQ5e5///sfPXv2JCEhgY8++giAN954g6NHjxbaPjMzky+//NIyq2FjY8OECRNwPnIE9u7VmZp8fHSmp+3bybhyhcScdLEBAQG0aNgQNm2CI0d0dewuXSA1VXfeqpUulldKZs6cib+/P507d8bf31/XzrjBcSGEEBWr0gQWZrOZUaNG0bZtWxrmbhIE+vfvz48//khERARjx47lhx9+4KmnnrphXxkZGaSmpuZ5CCFEZZOZmcnMmTNZsmQJp06dYujQoXTq1IkrV67w3HPPoZTK0z4rK4t+/frx0ksvMXz48GsvxMbC1q065auDgw4YataEAwewjYtj1apVzJs3j99//x1DYiLkbOKmaVO9EfvECV2Dwt//pmNOSEggIiKChISEm7YrrFDd9u3bCz1+s/6EEEKUvUoTWISGhhIVFcX8+fPzHB8+fDghISE0atSIAQMGMGfOHBYvXsyRI0eK7GvSpEm4urpaHjVr1izr4QshRKlbunQpZ86cwdvbm+7du2MwGJg2bRp2dnasWbOG2bmF69BF3p5++mkWLVqEjY0NTzzxhH7h/Hmd3clshqpVr3Xu4KDTx27diunSJfr166e/1ImJge3bdZtu3eDMGfDygubNdUByAyWZaYiOji60UN2GDRuKLGAHxQ9chBBClL5KEViMGDGCZcuWERERcdO1tK1btwaw/JIpzNixY0lJSbE84uPjS3W8QghRHnJvzAcNGoRVRgYcPEid2rV55513MBgMxMbGAvpb/WHDhvHTTz9hZWXFr7/+SkhICGRm6qAiKUnPUOTn46Nfi4wEpeDiRVi6VC99cnHR2Z9SUuDuu3XBuxsoagaiqAAgMDAQozHvryiTyUS7du0KPV6nTh1ZIiWEEBXsHx1YKKUYMWIEixcvJjw8nICAgJu+Z3dO+kNvb+8i29ja2uLi4pLnIYQQlUlCQgKrVq0CYPDgwXDgAKxaBVu3MnrkSCIjI5kwYQJKKUJDQ5k1axZGo5GffvqJhx9+WAcK27fDwYMQEKA3YEdFQWgofP65zvJkNOqAY98+OHZML3nKqXVBp066graDQ7GWQBU1A1HUl0C5hepyM1PlFqpr2bJloccBWSIlhBAV7B+dFSo0NJR58+bx22+/4ezsTFJSEgCurq7Y29tz5MgR5s2bx4MPPkjVqlXZu3cvo0ePpn379jRu3LiCRy+EEGVn9uzZmM1m2rdvT2DNmnpDtVKweTNWRiPNWrYE4LXXXuObb77BYDAwZ84cHn/8cV0de+dOHVj4+IC1NSxaBHPm6CVR8fE6aPjvf/VMRHKy3oORlaUDGLi2DKp2bb03I5/8WZtyZyCuDy5yZxqKUlShusKOR0REFBm4SNYoIYQoH//oGYuvv/6alJQUOnbsiLe3t+Xx888/AzqbyZo1a7j//vsJCgriv//9L71792bp0qUVPHIhhCg7ZrOZ7777DtA32cTG6pv8WrX0Tf6mTbBrFyhFhw4dsLOzY8aMGQwYMADS0iAsDP76S2+8Nplg0iSYPVsHFS1agJWVXiL1xRf6mK+vDjYWL9bP69bVtSquXoXAwAJ7KwpbklTUDMTNbvqLKlSX/3hRS6duFLgIIYQoXf/oGYv8GU3yq1mzJuvWrSun0QghxD9DVlYWY8aMYcGCBTzesycsWQIzZkCjRjBwoJ65+OsvMBrp8eCDHDlyBB8fH72Uaf16SEjQQUhSEnzwAZw8qYOJYcPggQf07MQHH+hlT9bW8J//6EBi40Y9gG7ddDraKlV0etrrFLWXIiQkpMgZiNKQG7g899xzZGdnFwhcpO6FEEKUvX90YCGEEKIgGxsbRowYwYgRI3SxukWL4NAh/cjIgKFDdXCxfj2YTPg0bKj3SWzaBOnpepZh/XqYOlVv4PbwgNdf1zMRAPfeq5dBffKJTi1rbQ333QenT+tMUffdp4OTe+8tULfiRnspfH19LY+yUFTgMnPmTEuwYzQamTZtmp7pEUIIUaoksBBCiIqUnq6XMfn46GVJJXXw4LUq2KCzNllbw6BBetnS2rU6CDh4EFxd9WbsadNgxQrdvlkzHUTkT2Jx3316T8X//R/88YcOSgDatdNLn6yt9abvfG5lL0Vpyh+43GgGRWYuhBCidP2j91gIIcSdbENYGP27d2f266/rm/fYWB0M3MCiRYuYPn06aWlpelP1mjV6SZONjQ4mdCP46SeoVk1vvo6KAm9vHRCMHXstqOjbF95+u2BQkatTJ70MCnS9C7i2advHR9evyOdW91KUVf2JkmajEkIIcetkxkIIIcrZjh07eOuNN1iZU8H6V5OJ9rVrU/vYMbjrLmjcWM8sFFJw7t1332X37t2kp6fzYuvWkLvPrFUraNgQnn0WZs2C+fP1rEKfPjrA2LULPv5Yb952coIxY/RG7etlZup9GBkZejbC1hZCQvQm7WnT9L6MevV0kbygIJ2OthAl3UtRlkuVKnoGRQgh/k0ksBBCiHJ09uxZ2rVrR0ZGBlZGI97u7sQnJzN27Vp+fvFFfdN+9KjeB9GkiZ4ZyLF27Vp2796NjY0NA3r1gmXLYM8e/WLTpnoGwWiEp5/WqWN/+EEHFxkZMG+e3ndx113w2mt5ZxuuXoXERN2uVi0dUOzffy24eOghaNAA3N11cTw3twKbtvMr7l6Ksl6qdLNN3UIIIUqPBBZCCFGOPFxdeeGhhzgXG8vbAwdyyWCg6ahRLNiwgdGPPMK9QUFw+bKuFxEbqzdIN27MlcxMhg8fDuiCeFVSUyE8XAcDPj76Rr9OHb0v4t57dbDw00+Qk5YWgPvvh+HD9bIp0G0TE+HKFT1D0rSprktx9aoOQvbv14GIjc21/RTR0XpfRs7yqdvNtnSzzd6loSyzUQkhhLhGAgshhChDly9f5tlnn+Wtt96iYVAQbNrEp61bY+jTR1etzs5meLdu2NnaEpg7O+HgoGcszp7VKV+Tk3ln5Uqio6Px9vZm0nvv6eM7duj2992nb/Tr1dMZnpYv10uYMjNh4UIdGDz/PHTtem1gZ87olLE1akDnzjqAsLbWr1lZ6WNK6QCnTh39WmamnhGpXRsonSVM5bVUqSyzUQkhhNAksBBCiDKSnp5Oz549+fPPP9mxYwcHp0/HeudODL6+Ok3rqlUwcyZfK4WhYUNduK5xY70cyWTSQYKDA7uWLeOjL78EdOFQt0uXIDJSL5kyGvWSKU9PvUTJ3R1at9Z1LPr21bUtvLyuLakym3WKWmtr6NIF7r5bL3fKz8FBBxfZ2XqWok4dHehUrw4+PqW2hEmWKgkhxJ1DAgshhCgDmZmZ9OnThz///BNHR0fmvPwy1jt36ht8a2td1TosDAAD6EAhMhIFKEdHjI0a6b0NjRvz4qpVZJvN9GnenEfr1s2bYrZFC7C2JsHBgei1a/WSpHvu0bMRUVE66MhNY5uRoZdXeXtD+/Z6+dONODrq4MNs1ns/QG8St7Yu1SVMslRJCCHuDBJYCCFESWRl6Zv2s2f134OCru1ZsDTJol+/fixbtgw7OzuWffghbS9f1jf0Fy/qqtbHjunZhqeegnvugX372L1lC/89cIDHLl3ixS1b9FKnsWP5buRIXp09my9799azHEYj7NypT9axIzMjIxn+2mt5lyT16aNTxMbH6xmQCxfg1Ck9Q9GuXdEpZvNzdtZLqMxm3V/Opu3SXsIkS5WEEKLyk8BCCCFuRCkdSJw5o2/M4+IgJUUXtlNKP2/bVi9BQn9r//TTT7No0SJsbGz4bfJkOqan6yVEBw7ognOXLulidS+/rGcUAGrXZqutLeF//80eBwcG3n03bjt2wAcfUPfNN1nyxhu6XWqqrpqdmgru7iS4uTH8o48wKwXkW5J03326Psbhw3qWpF07aN782l6K4nJx0fUrTp2CqlUBWcIkhBCiIAkshBDiRqKiYONGHQyYTPob/OrVwc5Ob2Y+eFDPBrRrB7Vq8eGHH/LTTz9hZWXFwsmTuT87G6pUgd9/h19+0X0GBemUrzk36bmGdOvG50uXciA+nvd9fRmQkUGTfftg4kQYP17XqXBxuTZb0bkz0fHxlqAil2VJUseO0KYN7NunlzDVqVNobYzrFZnlydVVP64fryxhEkIIcR2pvC2EEEVJSIBNm3RAERioMyd5euqgAvQSqMBAHVgsXw47dzLihRfo3Lkz8ydN4iHQMxkLFlwLKh5+WAcKuUHF+fN61iM5GSuzmQ+feQaAj5YsoVlUFGO9vHQA8847esYjOflaYNGpE4EeHhjzFarLsySpUSPo1UuP8yZBxcyZM/H396dz5874+/szc+bMm/6IfH196dixowQVQgghJLAQQvwL5fuGv1CpqTqzUkaGzqpkMFxL//r55zBiBFenTGHmypV0/PZbMm1sICIC523bWPPRR/S2ttbf8C9eDCtW6PePHAnDhumlSCkpcOiQriHh62vJ1tTDxYWOgYE5w1RUf+ABvVwqPR0mTNBVtc1mvVfCyQlff3+mTZmCKWeDdoElSQaDzkB1E0VleUpISLi1n7EQQoh/HVkKJYT498jK0pWq4+IgONhSffrMmTNs3LiR1NTUa+3276dxVhZNnZ1hxQou7NnD74mJlq5SgC/i4sjJlcScPXsY2q4d7NmDwcpKL1n64w+9BApgxAidYSktTRelc3DQGZ0aNtRpZdPT4exZDGfP8qmVFW1fe42WNWrwYvfu0KOHXgr19996fwXoPQ8XLkD9+gy5/35CHn64REuS8i95Ko9CdUIIIe5sElgIIf45MjL0rEBaWuGvOzrePEVqUS5e1Hsl9u3Tz5OT2enuzrPvvMPevXsLfcvb9evT9MABAE4CgwppUw0Ya23NAFvba4Xt0tNh6VJdnA50cbr27XU9CFtbPQPRqJHeq5HLzk7PXPj60qxpU0707o3jli2YEhL03oi33tLBxcGDegaiTRsdIOX8PEqSVamwwnYhISHlUqhOCCHEnUsCCyFExbl6Ve8ZOHsWTpzQj9RUXZStME5O0L27JeVpsZ08CX/9RcLevTjcdRdVqlWDxES8o6IsQUXDhg31jfnly5aq1HVyggruvx/HunV54K+/LDUhDAYDHWrXJnTfPpwOHoRPPtEF655+GlauhHnz9HuHDIEHH9SZmerW1VmZvL0t+x2K2iztXrOmroD922/6Z1S1KowbB3Pm6KAkO1sHWtcHJ8VQ1JKn2NhYyfIkhBDitkhgIYQoP7k1IM6c0fsXjhzR2Zays/UNu6Oj/tY/t6Db9YxGfTO+fr0uHOfmdvPzmc16w/PGjSQmJtJ51iwcjEZWv/km1WrUwNvRkaXPPEOrTp3wfOghvfdi2bK8sw2DBkHv3vgDK7p00TMfqal6b4TZDI8/Dj/9BEuW6P0U27frTd+gg4xHH4UzZ0jIziY6K4tAsxnfnKCisJmDIUOGANcFHDVq4HvwoF5a5egIL7xATgOoVs2S5rYoJVnyJFmehBBC3A4JLIQQZSe3qNrZs5CUpJfuREbqG/ejR0ven7+/3vy8YQPcf3+BwnR5pKfD1q2wcyenzWa6fPUV0adPUwtIHzVK3/A/9hgPPfSQLla3bJmeEfntNx0gAPTvD488om/iL13SwY2zs55x8PfXxediYuDZZ6FePV1NOzeoePJJHXRkZjJz5UqGL1xYYOlRYTMHISEhrFq1Km/AMXw4Q2xtISDg2ue7dAnuvfeGmZ5uZcmTFKoTQghxqySwEEKUHqX0huKzZ68Vk7twQX+7n5AAq1frDcigl/nUqHHDG2OlFPuvXsXaYKDu+fNw/DisWMFFpXCqWrXoG+vkZD0jcvgw51JS6PbZZxy4cgVfIByomZGhU8D+8Qf07Kk3Rycn6+AiN6h4/HHo3p2EyEiirawIbNkS39yN1k5OeiYgPp5AoxHfhARdJM/fH2bP1hmbevYEIGHvXob/+muBAnbz5s0rdOZg8+bNBQOO6dMJGT8e35QUnWkqI0MHVTdYBiVLnoQQQpQ3CSyEEMWXnq6/oY+N1bMR+V25opc65X677+SkN2IvXqxnDwBMJlSXLsR06ECCUgT6+ODr4cHBhASyzWYSzp7lz127iElMZMuhQ5xOSWF4SAhvNWtG9OTJeG7cSLPNm2k6fTqdQ0Lo1Ls37dq1w9HRUfd/5Iie0YiKIiUigvsjI9kLeAFhbdsSMHiw3kQ9d66ecfjxRz2DEhwMq1bpPh59FLp2ZeayZQz/5ZcCS5UKzAQ8+SRDQkL05us337z280hOJjo1tdACdgaDodCZA6VU4UuVXF3xTUzUP9MLF/QSqGrVirxUpb3kqcjCeUIIIUQOg1LFSeh+Z0tNTcXV1ZWUlBRcXFwqejhC3J6rV/UG5HxVkm9VQkICx2NiaOvpqVO1njjBG6tX4+/hwbPBwVhfvx/CaNT7KGJidPalvXv1fgogwWAgIiiIcDc3fo+K4lxO5iejwcC0Z57h9V9/5Wwh2aDsbWxo7u/PppgYzEphAPL/p2Vtbc29rVvTuV49+lSrRoODB0n77TdClGIz4GFtzdqxY2nQosW1N2Vn6wBk3jyd/jVXSAjcfz8JWVn4jx1b4MZ/8+bN3HvvvXmPG43Ejh2Lb5Mm14rn5fwcEoKC8H/ssQL9xMbGsmrVqgIzByEhIfj7+xdsv38/vpGROnDLyICWLeG++2543QrtJza2xIHBjfaCCCFuLDExkWnTpjF8+HC8vb0rejhClFhJ7pNlxkKIO8jRPXuY9t57dPD1pfvw4RAUdNNqy7k+//xz9uzZY3l+9epVtm7ZQnRMDH4eHsSGhmI4cYLo6Ggmr19PtlJMXryY8QEB9PfywmQ265mAEycsfSwCVhsMhNvZcfjKFb2ROh+zUjz3/ffUdHfnbL7XjAYDq0ePpv2HH1q+9Vfoyp6fOjiwu25dwhISiD97lr82bOCvDRuo4edHg7g4zgLxNja4W1mxZtIkGly/PwH0BvEOHfQSpvBwvQzq7rv1zbqvL9EmU6Hf+G/YsKHgcbOZGBcXfGNjdbpZk0nPhgQE4NutW5FLj4qaOSi0fd26ut9ly/SSsxo1bng9fX19S2XJU1FLqkJCQmQety6jAAAgAElEQVTmQgghRB4SWAhxB0hISODd117ju59/Jis7m8nAa3v38t5bb2EVHKxrJ+Rz8uRJfDw9dV2EY8dYM3cuy7ZvL9DOCFRPSSF1wgRcgZrAJ8D7wNErV3h6/37e37+f8UANoJ3RCHfdBY0bM2nzZnacPAlXrmA0GmlesyZ3ubszf/fuPOfINpt56fXXGfPKK3mOm5Vim4dHgaVEZqDJ5cuMNJtR//0vR41GwlevJnzDBrrExYGVFQHDh7O+aVPOXbxIk/xBxfWU0ilgAwP1LEPjxhAcTOD584UuVWrXrl3hm58feQR27dL7QKpW1UFAq1ZgZ3fDpUeFbZYusn3t2np8cXHFSjNbGlmepHCeEEKI4pLAQojyZjZDZua1JTO34fTp07w/cSLffPMNGZmZADQwGPhbKX6JjOR/69fjkpoK7dpBlSqW982cPp0RL77I9w0a8MTevZCVxSCg3XV9G4B6QAfA7epVfdDfn7OBgTQ2GvnLwYHFcXFMjoriYGYmTwImg4G/x4+nXk6GoYGOjjQ+fpwAV1d6BgbSoGlTEqpXZ8EDDxS8Ye/QofAb+cKOGwzUsbGBvXsx1K7NXW5u3LV+PcMyM8HTE157DerUIeDSJQKcnHSK2BwJyclEnzhBoKMjvnZ2ehO5i4uuMeHvr/80mfB1dCz0G/+WLVsWPhMQFKQzRi1bpmcr7rsvT72NkmZbKrS9waCDlZo1derZW+2nBAIDA6VwnhBCiOJRQqWkpChApaSkVPRQxB3i7Nmz6ssvv1Rz585VJ0+evPbCxYtKrV6t1Ny5Sh07dtvn2bZ+vUKvDlLtXVzUq6CMOc8NoGZUr67U8OFKff+9UkePqssXL6rBjz5qec+ToOJBhef8qa57xIMKr1JFxXfsqNSrryo1Z46aMWKEMhoMClBGg0HNGDJEXfj4Y/XI3Xdb+jSAmjFwoFLffKNmDBx4rb3RqGZ89ZVSSqkZM2Yok8mkAGUymdSMGTOKf9xoVDOeeEKpl17KM14FSjVrptSPPyq1ZImKf+stFT5ypIqfNEmpb74pOB6DQc0YN06phASlrlxR8fHxKjw8XMXHx+f5GcfHx6uIiIjiH1+9WoWPH6/iDx267etbmKLGWZaKui5CiJs7efKkGj9+fN7fBUJUIiW5T5bAQklgIXJkZyt15oxSBw7oR1ZW8d6XlKRUVJRSly6pCxcuqLfffls5OztbbrQBVb9+ffX9F18otWCBUh9+qNTnnys1ZYpSO3YU/zzXM5uVOn5cqV9+Uf8LDlZ/Vq+u4q4LKnIfppwA4ctq1dQ7rVurZu7u+qYa1ERQ06pVy3ujHRqq1MKFasbgwXmPP/GEin/rLcsxS/9Go9o2ebIyGo15j5tMaltERKHHc2+IS3zDnns8NlapP/5Q6uOPlerRQwcUBoNS/foptXixUkuWqBlPPJE3oJkyRcUfOFDkeGbMmGF5zWg03vKNc2n1U1H930hR10UIcWMSWIjKTgKLEpLAonJLTU1VkydPVmlpaSV6X3xcnOr/+OPqub591fIxY1Tak08q1aiRindyUuG1aqn4iRP1DENR0tKUGj1axbu4qHA/P7V3xAhVxdXVctPauHFj1bx5c2XIucH9bsAApXr0UPFubmp2QIAa3ry5+mPwYJW6eLHuqxiyrl5VY4YPV8dmz9bBybBhSjk6KgUq3NExz01z7mO2waCsrnvuAWp19eoq/sUXCwYKBoPaNnJkoQHEz++8U2j/n376aYmOR0RElOg6FerMGaWmT9czEcOHK/X++0r9/rtSv/2m4seNKzh+k0n9/PPPhY5nwYIFNwyAiis+Pr5U+qmo/oUQZUMCC1HZleQ+2Vis9VKVwNSpU6lVqxZ2dna0bt2abdu2VfSQbiohIYGIiAgSciv13uLxsh5Pabt8+TKffPIJffv2tZx3+fLlPProo8TExBS7nytXrvDJJ59Qu3ZtXnvtNd578kkiPvuMhJgY9u/fz8GDBwu0X7RoEZjNzBw7Fj8/P+b9+ivf/vwzD376Ke7z51Nn3z5qXrxI59hY/N94g5mNG8P69XlPnJ4OH38Mfn5M/+wz/FNT6RwXR9MpU6iVmcndnp78+tln7NqyhR0bN5K8aBGLW7XiwSVLmPnHH/hfuMAzx44xLTKSHt99h3uvXrRp3pw3R41i//79RX5e88mTDO7enU+nTaPrqFEc/e03IqZPJ+HSJahbl8ChQzHmywBlMhjo/L//8WlQELbA3VZW/PHUU3T99luiq1cvWF9BKTY4OhY8bjZjCArCaMz7X8b1m5mLe7xU1uZ7eECLFpCWRkLz5kSYzSScOQOxsUSnp9+wbkT+8aii6kaU4N8i3HiTc2ko6/6FEEKI21b2cU7Zmz9/vrKxsVHfffed+vvvv9WwYcOUm5ubOnXqVLHeXxEzFkUtaZgxZUre4xMmKLVvn5oxYULe49Omle548p+3lPtXSqn09HQ1ZcoU5e3tbfnG9dVRo/J8C2tvZ6eWLFlSdCdpaSr94EE19cUXlfd1y408c9b2k7PMp76DgzIaDGrQgw+qA/v2qa8mTlQ+bm4KUL+5uBRYMlTUI3cpkXrkEaX271fq66+V8vZWGaAmFdLeCCoWlLrnHr1U54UXlHJ1texZuNl5DQaDGtCjhzoZHq7Uvn2Wh/mnn9SwwEDLmF64ri8jqBm9euklTJ99lnct/JtvKjVliprRv/+1n0/O3oj4N94o9Jv9bdu23XDJ0C3vjSjttfmZmQX3fDzzjIov4fhLayZAZiyEEIWRGQtR2f3rlkK1atVKhYaGWp5nZ2crHx8fNWnSpGK9P/cH9r/XX1dRe/cqc1aWUtnZKv74cRW+Zo2KP35cqRMn9KbQIUNUfFCQCm/TRsWPH6/Unj16jXz+9tnZlsf1xxN37VJfPv10wRtKUN75bhpyH94Gg+Wm8Pob2OM//6zU4cMqft8+Fb569Q3PW+D48uUqfssWpf78U8UPG3bj/vfuLXn/OccvHDqk5r/4ohoeFKS8rKws/dcymdQnjo4Fzpv7eG3YMHVs585r41y9WqmXX1bj/PyU13U3w36gProuqLjZww/UB0W89kaXLoUv3eHaZubVoD4FFVDEtcrf3rIp2sNDhXfvXmj7n0B9B6pbznMXUOdcXZVyd1fK3V3FOTqqntf9O/mykADFZDSq+OhopVTBtfDxGzcWurQpfsGCgoFIMQKCW94bUco3wDe60S7p+EsrACrrTc6yiVqIykcCC1HZ/asCi4yMDGUymdTixYvzHH/66afVI488Uqw+cn9glm+/7e1VK09Py82qAdSMnJvFd667iTWgvzleaG+vXriuvdFgUDM6d1aHBgxQLzRsWOyb3pI+3EC1sLfPM54RjRsrNWKEmtG5s+Vm0gDqhYYN1cLu3QuMZ0bOzW9h/bvm6z/3c139z3/y9FNU/0ZQbxTSrwHUNzc4b2Htc3/+7fId/7ppUxX+xBOFvu8rLy/V4LobagOorxs0UPHPPVf4JuT58wveqBqNKv7ZZ9WonE3Pef6dODlZ9k/kbz/Dx+fajILBoGb06qXi33uv8P4HD87T/vrPOz3fOZ+tWVOF9+lTeEBTxN6F8PDwG7Yv74CgtNzq5ypKaX3esv65/dOvixAiLwksRGVXksDCoFS+xciVzMmTJ6lRowabNm0iODjYcvzVV19l3bp1bN26tcB7MjIyyMjIsDxPSUnB77p884UxAmuqVaPzmTPFGpcRGApMK+S1OiYTMdnZeY4ZgCnPPUfot98WPN6lC6FhYcU6L0A40AV9l3UzBiCsShW6nDtXrPZG4A+gezHHYgAamExE5fu8RmDNgAF0mTs3z3mNwBvAu4Wcd03VqnROTi54/JVX6PLxx1z/T9loMLDmpZfo8vnnBfqPevddwk6eZOQ332BWCqPRyOeff87TTz/NnDlzGDlyJGazWR//6CM6BwTQoFevAp9t56pVbDx8uED7LrVr06Bnz7znNRiIWruWsL1787b/+GO6BAQUbA+s6duXzj//XPDzvvyy/rzXHzcaiYqKokYh1ZhPnDhBgwYN8v58btC+srhTP5cQ4s6SlJTErFmzePbZZ/Hy8qro4QhRYqmpqdSsWZMLFy7g6up6w7b/ysBi/PjxTJgwoTyHKYQQQgghRKUVHx9/04Krlb7ytoeHByaTiVOnTuU5furUqSK/GRg7dixjxoyxPL9w4QL+/v7ExcXdNBITlU9upB0fH4+Li0tFD0eUMrm+dza5vnc2ub53Nrm+dwalFGlpafj4+Ny0baUPLGxsbGjevDlhYWE89thjAJjNZsLCwhgxYkSh77G1tcXW1rbAcVdXV/mHfwdzcXGR63sHk+t7Z5Pre2eT63tnk+tb+RX3i/dKH1gAjBkzhkGDBtGiRQtatWrF//3f/3Hp0iWeffbZih6aEEIIIYQQ/wp3RGDRt29fzpw5w9tvv01SUhJNmzZl5cqVVK9evaKHJoQQQgghxL+Cafz48eMrehCloVWrVowePZq33nqLYcOG3XRzSX4mk4mOHTtiZXVHxFoiH7m+dza5vnc2ub53Nrm+dza5vv8ulT4rlBBCCCGEEKLiGSt6AEIIIYQQQojKTwILIYQQQgghxG2TwEIIIYQQQghx2/71gcXUqVOpVasWdnZ2tG7dmm3btlX0kMQtmDRpEi1btsTZ2RlPT08ee+wxDh06lKdNeno6oaGhVK1aFScnJ3r37l2gsKKoHD744AMMBgOjRo2yHJPrW7mdOHGCp556iqpVq2Jvb0+jRo3YsWOH5XWlFG+//Tbe3t7Y29vTtWtXoqOjK3DEoriys7N56623CAgIwN7enrvuuot3332X67d4yvWtPNavX8/DDz+Mj48PBoOBJUuW5Hm9ONfy3LlzDBgwABcXF9zc3BgyZAgXL14sz48hysi/OrD4+eefGTNmDOPGjWPnzp00adKEkJAQTp8+XdFDEyW0bt06QkND2bJlC6tXr+bq1avcf//9XLp0ydJm9OjRLF26lF9++YV169Zx8uRJevXqVYGjFrdi+/btfPvttzRu3DjPcbm+ldf58+dp27Yt1tbWrFixgv379/PJJ5/g7u5uafPhhx/yxRdf8M0337B161YcHR0JCQkhPT29AkcuimPy5Ml8/fXXTJkyhQMHDjB58mQ+/PBDvvzyS0sbub6Vx6VLl2jSpAlTp04t9PXiXMsBAwbw999/s3r1apYtW8b69esZPnx4eX0EUZbUv1irVq1UaGio5Xl2drby8fFRkyZNqsBRidJw+vRpBah169YppZS6cOGCsra2Vr/88oulzYEDBxSgNm/eXFHDFCWUlpamAgMD1erVq1WHDh3UyJEjlVJyfSu71157TbVr167I181ms/Ly8lIfffSR5diFCxeUra2t+umnn8pjiOI29OjRQw0ePDjPsV69eqkBAwYopeT6VmaAWrx4seV5ca7l/v37FaC2b99uabNixQplMBjUiRMnym/wokz8a2csMjMziYyMpGvXrpZjRqORrl27snnz5gocmSgNKSkpAFSpUgWAyMhIrl69mud6BwUF4efnJ9e7EgkNDaVHjx55riPI9a3sfv/9d1q0aEGfPn3w9PSkWbNmTJ8+3fL6sWPHSEpKynN9XV1dad26tVzfSqBNmzaEhYVx+PBhAPbs2cOGDRvo3r07INf3TlKca7l582bc3Nxo0aKFpU3Xrl0xGo1s3bq13McsSte/tlrJ2bNnyc7OLlCdu3r16hw8eLCCRiVKg9lsZtSoUbRt25aGDRsCkJSUhI2NDW5ubnnaVq9enaSkpIoYpiih+fPns3PnTrZv317gNbm+ldvRo0f5+uuvGTNmDP/73//Yvn07L730EjY2NgwaNMhyDQv7/1qu7z/f66+/TmpqKkFBQZhMJrKzs5k4cSIDBgwAkOt7BynOtUxKSsLT0zPP61ZWVlSpUkWu9x3gXxtYiDtXaGgoUVFRbNiwoaKHIkpJfHw8I0eOZPXq1djZ2VX0cEQpM5vNtGjRgvfffx+AZs2aERUVxTfffMOgQYMqeHTidi1YsIC5c+cyb948GjRowO7duxk1ahQ+Pj5yfYW4w/xrl0J5eHhgMpkKZI05deoUXl5eFTQqcbtGjBjBsmXLiIiIwNfX13Lcy8uLzMxMLly4kKe9XO/KITIyktOnT3PPPfdgZWWFlZUV69at44svvsDKyorq1avL9a3EvL29ufvuu/Mcq1+/PnFxcQCWayj/X1dOr7zyCq+//jpPPvkkjRo1YuDAgYwePZpJkyYBcn3vJMW5ll5eXgWS5GRlZXHu3Dm53neAf21gYWNjQ/PmzQkLC7McM5vNhIWFERwcXIEjE7dCKcWIESNYvHgx4eHhBAQE5Hm9efPmWFtb57nehw4dIi4uTq53JdClSxf27dvH7t27LY8WLVowYMAAy9/l+lZebdu2LZAe+vDhw/j7+wMQEBCAl5dXnuubmprK1q1b5fpWApcvX8ZozHu7YTKZMJvNgFzfO0lxrmVwcDAXLlwgMjLS0iY8PByz2Uzr1q3LfcyilFX07vGKNH/+fGVra6tmz56t9u/fr4YPH67c3NxUUlJSRQ9NlNALL7ygXF1d1dq1a1ViYqLlcfnyZUub559/Xvn5+anw8HC1Y8cOFRwcrIKDgytw1OJ2XJ8VSim5vpXZtm3blJWVlZo4caKKjo5Wc+fOVQ4ODurHH3+0tPnggw+Um5ub+u2339TevXvVo48+qgICAtSVK1cqcOSiOAYNGqRq1Kihli1bpo4dO6YWLVqkPDw81KuvvmppI9e38khLS1O7du1Su3btUoD69NNP1a5du9Tx48eVUsW7lg888IBq1qyZ2rp1q9qwYYMKDAxU/fr1q6iPJErRvzqwUEqpL7/8Uvn5+SkbGxvVqlUrtWXLlooekrgFQKGPWbNmWdpcuXJF/ec//1Hu7u7KwcFB9ezZUyUmJlbcoMVtyR9YyPWt3JYuXaoaNmyobG1tVVBQkJo2bVqe181ms3rrrbdU9erVla2trerSpYs6dOhQBY1WlERqaqoaOXKk8vPzU3Z2dqp27drqjTfeUBkZGZY2cn0rj4iIiEJ/3w4aNEgpVbxrmZycrPr166ecnJyUi4uLevbZZ1VaWloFfBpR2gxKXVf6UgghhBBCCCFuwb92j4UQQgghhBCi9EhgIYQQQgghhLhtElgIIYQQQgghbpsEFkIIIYQQQojbJoGFEEIIIYQQ4rZJYCGEEEIIIYS4bRJYCCGEEEIIIW6bBBZCCCGEEEKI2yaBhRBClKOOHTsyatSoih4GSUlJdOvWDUdHR9zc3MrlnAcPHuTee+/Fzs6Opk2bEhsbi8FgYPfu3WV+7vHjx9O0adMSvefy5cv07t0bFxcXDAYDFy5cKJOxrV27tkz7F0KI8iKBhRBC/At99tlnJCYmsnv3bg4fPlwuN7fjxo3D0dGRQ4cOERYWRs2aNUlMTKRhw4a33Ofs2bOLFRi9/PLLhIWFlajv77//nr/++otNmzaRmJiIq6vrrQ7TorDAsk2bNqXWvxBCVCSrih6AEEKI8nfkyBGaN29OYGAgAPv37y+VfpVSZGdnY2VV8NfLkSNH6NGjB/7+/pZjXl5et9RXSTk5OeHk5FSi9xw5coT69evfVuBTHDY2Njf8OQghRGUhMxZCCFFGLl26xNNPP42TkxPe3t588sknBdr88MMPtGjRAmdnZ7y8vOjfvz+nT58G9I11nTp1+Pjjj/O8Z/fu3RgMBmJiYgo97/bt2+nWrRseHh64urrSoUMHdu7caXm9Vq1aLFy4kDlz5mAwGHjmmWfo1KkTAO7u7pZjAGazmUmTJhEQEIC9vT1NmjTh119/tfSVO9OxYsUKmjdvjq2tLRs2bCgwJoPBQGRkJO+88w4Gg4Hx48cXWApVVF979uyhU6dOODs74+LiQvPmzdmxYwdr167l2WefJSUlBYPBYOm3MPmXQj3zzDM89thjfPzxx3h7e1O1alVCQ0O5evUqoGcWPvnkE9avX4/BYKBjx44AZGRk8PLLL1OjRg0cHR1p3bo1a9euzXOujRs30rFjRxwcHHB3dyckJITz58/zzDPPsG7dOj7//HPLeGNjYwudLVq4cCENGjTA1taWWrVqFfi3U6tWLd5//30GDx6Ms7Mzfn5+TJs2rdDPLoQQ5UYJIYQoEy+88ILy8/NTa9asUXv37lUPPfSQcnZ2ViNHjrS0mTlzplq+fLk6cuSI2rx5swoODlbdu3e3vD5x4kR199135+n3pZdeUu3bty/yvGFhYeqHH35QBw4cUPv371dDhgxR1atXV6mpqUoppU6fPq0eeOAB9cQTT6jExER14cIFtXDhQgWoQ4cOWY4ppdR7772ngoKC1MqVK9WRI0fUrFmzlK2trVq7dq1SSqmIiAgFqMaNG6s///xTxcTEqOTk5AJjSkxMVA0aNFD//e9/VWJiokpLS1PHjh1TgNq1a9cN+2rQoIF66qmn1IEDB9Thw4fVggUL1O7du1VGRob6v//7P+Xi4qISExMt/RZm3LhxqkmTJpbngwYNUi4uLur5559XBw4cUEuXLlUODg5q2rRpSimlkpOT1bBhw1RwcLBKTEy0fKahQ4eqNm3aqPXr16uYmBj10UcfKVtbW3X48GGllFK7du1Stra26oUXXlC7d+9WUVFR6ssvv1RnzpxRFy5cUMHBwWrYsGGW8WZlZVk+9/nz55VSSu3YsUMZjUb1zjvvqEOHDqlZs2Ype3t7NWvWLMv4/f39VZUqVdTUqVNVdHS0mjRpkjIajergwYNF/rsQQoiyJoGFEEKUgbS0NGVjY6MWLFhgOZacnKzs7e3zBBb5bd++XQGWG+QTJ04ok8mktm7dqpRSKjMzU3l4eKjZs2cXeyzZ2dnK2dlZLV261HLs0UcfVYMGDbI8z39zq5RS6enpysHBQW3atClPf0OGDFH9+vXL874lS5bcdBxNmjRR48aNszwvKrDI35ezs3ORn3fWrFnK1dX1pucuLLDw9/dXWVlZlmN9+vRRffv2tTwfOXKk6tChg+X58ePHlclkUidOnMjTd5cuXdTYsWOVUkr169dPtW3btshxdOjQocD1z/+z79+/v+rWrVueNq+88kqeANPf31899dRTludms1l5enqqr7/+ushzCyFEWZOlUEIIUQaOHDlCZmYmrVu3thyrUqUK9erVy9MuMjKShx9+GD8/P5ydnenQoQMAcXFxAPj4+NCjRw++++47AJYuXUpGRgZ9+vQp8tynTp1i2LBhBAYG4urqiouLCxcvXrT0WVwxMTFcvnyZbt26WfYoODk5MWfOHI4cOZKnbYsWLUrU943k72vMmDEMHTqUrl278sEHHxQ4961q0KABJpPJ8tzb29uyDK0w+/btIzs7m7p16+b5eaxbt84ypt27d9OlS5fbGteBAwdo27ZtnmNt27YlOjqa7Oxsy7HGjRtb/m4wGPDy8rrh+IUQoqzJ5m0hhKggly5dIiQkhJCQEObOnUu1atWIi4sjJCSEzMxMS7uhQ4cycOBAPvvsM2bNmkXfvn1xcHAost9BgwaRnJzM559/jr+/P7a2tgQHB+fpszguXrwIwB9//EGNGjXyvGZra5vnuaOjY4n6vpH8fY0fP57+/fvzxx9/sGLFCsaNG8f8+fPp2bPnbZ3H2to6z3ODwYDZbC6y/cWLFzGZTERGRuYJSADLxnB7e/vbGlNJlHT8QghR1v6/vfsJSWUNwwD+dCRbVEsziLIIJ5QIMgiMXNWihKA2RUhmuImQQDTIRYG5UGpp5CpMIoqCoEiIwIhiFmYERZAI/XWXii2kKIjO4nK9dIvOPccrtXh+y/n3vTOLgWfej28YLIiI8qC2thaFhYUIh8OoqqoCAKTTacRisWxXIhqNIpVKwePxoLKyEgBwdHT07lp6vR7FxcXw+XzY3t7G/v7+p2OLooi5uTno9XoAQDweRzKZ/PQcqVQKAG++iKvVahQVFeH29jZb81cRBAGCIMBqtaK/vx9+vx89PT2QSqVvas6nxsZGvLy84O7uDjqd7sNjGhoaEAqF4HQ6P9z/X+pVqVQQRfHNNlEUIQjCu0BDRPSdcCoUEVEelJSUwGw2Y2xsDLu7uzg7O4PJZMKPH/+8dquqqiCVSuH1enF5eYnNzU24XK5315JIJDCZTHA4HFAqldBqtZ+OrVQqsbi4iPPzc4TDYRgMhl9+SVcoFCgoKMDW1hYSiQQymQxKS0tht9thtVoRCARwcXGB4+NjeL1eBAKBP3swv+nx8REWiwV7e3u4ubmBKIqIRCJQqVQA/lodKZPJIBQKIZlM4uHhIW+1CIIAg8EAo9GI9fV1XF1d4fDwEG63G8FgEADgcDgQiUQwMjKC09NTRKNR+Hy+bLCrrq5GOBzG9fU1ksnkhx0Gm82GUCgEl8uFWCyGQCCA2dlZ2O32vN0bEdH/gcGCiChPZmZmoNPp0NXVhfb2drS2tqKpqSm7XyaTYWFhAWtra1Cr1fB4PO+Wlv2b2WzG8/MzhoaGfjnu/Pw80uk0NBoNBgYGMDo6irKysk/PqaiogNPpxPj4OORyOSwWCwDA5XJhYmICbrcbKpUKHR0dCAaDqKmp+Y0n8eckEglSqRSMRiMEQUBvby86OzuzHYGWlhYMDw+jr68PMpkM09PTea3H7/fDaDTCZrOhrq4O3d3diEQi2a6UIAjY2dnByckJmpubodVqsbGxkf0Xh91uh0QigVqtzk59+zeNRoPV1VWsrKygvr4ek5OTmJqayi4BTET0XRW8vr6+fnURRET0uYODA7S1tSEej0Mul391OURERO8wWBARfWNPT09IJBIYHBxEeXk5lpaWvrokIiKiD3EqFBHRN7a8vAyFQoH7+/u8T/MhIiLKBTsWRERERESUM3YsiIiIiIgoZwwWRERERESUMwYLIiIiIiLKGYMFERERERHljMGCiIiIiIhyxrDytcsAAAApSURBVGBBREREREQ5Y7AgIiIiIqKcMVgQEREREVHOGCyIiIiIiChnPwE9Oe2XC0FV/AAAAABJRU5ErkJggg==\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "def plot_forecast(samples):\n", " duration = len(empty_data)\n", " forecast = samples[\"S\"].size(-1) - duration\n", " num_samples = len(samples[\"R0\"])\n", "\n", " time = torch.arange(duration + forecast)\n", " S2I = samples[\"S2I\"]\n", " median = S2I.median(dim=0).values\n", " p05 = S2I.kthvalue(int(round(0.5 + 0.05 * num_samples)), dim=0).values\n", " p95 = S2I.kthvalue(int(round(0.5 + 0.95 * num_samples)), dim=0).values\n", "\n", " plt.figure(figsize=(8, 4))\n", " plt.fill_between(time, p05, p95, color=\"red\", alpha=0.3, label=\"90% CI\")\n", " plt.plot(time, median, \"r-\", label=\"median\")\n", " plt.plot(time[:duration], obs, \"k.\", label=\"observed\")\n", " plt.plot(time[:duration], synth_data[\"S2I\"], \"k--\", label=\"truth\")\n", " plt.axvline(duration - 0.5, color=\"gray\", lw=1)\n", " plt.xlim(0, len(time) - 1)\n", " plt.ylim(0, None)\n", " plt.xlabel(\"day after first infection\")\n", " plt.ylabel(\"new infections per day\")\n", " plt.title(\"New infections in population of {}\".format(population))\n", " plt.legend(loc=\"upper left\")\n", " plt.tight_layout()\n", "\n", "plot_forecast(samples)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "It looks like the mean field guide underestimates uncertainty. To improve uncertainty estimates we can instead try MCMC inference. In this simple model MCMC is only a small factor slower than SVI; in more complex models MCMC can be multiple orders of magnitude slower than SVI." ] }, { "cell_type": "code", "execution_count": 15, "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "INFO \t Running inference...\n", "Warmup: 0%| | 0/800 [00:00, ?it/s]INFO \t Heuristic init: R0=2.05, rho=0.437\n", "Sample: 100%|██████████| 800/800 [01:44, 7.63it/s, step size=1.07e-01, acc. prob=0.890]" ] }, { "name": "stdout", "output_type": "stream", "text": [ "CPU times: user 1min 43s, sys: 1.2 s, total: 1min 44s\n", "Wall time: 1min 44s\n" ] }, { "name": "stderr", "output_type": "stream", "text": [ "\n" ] } ], "source": [ "%%time\n", "model = SimpleSIRModel(population, recovery_time, obs)\n", "mcmc = model.fit_mcmc(num_samples=4 if smoke_test else 400,\n", " jit_compile=True)" ] }, { "cell_type": "code", "execution_count": 16, "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "INFO \t Predicting latent variables for 90 time steps...\n", "INFO \t Forecasting 30 steps ahead...\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "samples = model.predict(forecast=30)\n", "plot_forecast(samples)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Advanced modeling \n", "\n", "So far we've seen how to create a simple univariate model, fit the model to data, and predict and forecast future data. Next let's consider more advanced modeling techniques:\n", "\n", "- [regional models](#Regional-models) that couple compartments among multiple aggregated regions;\n", "- [phylogenetic likelihoods](#Phylogenetic-likelihoods) to incorporate genetic sequencing data;\n", "- [heterogeneous models](#Heterogeneous-models) with time-varying latent variables; and\n", "- [Complex compartment flow](#Complex-compartment-flow) for models with non-linear transitions." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Regional models \n", "\n", "Epidemiology models vary in their level of detail. At the coarse-grained extreme are univariate aggregate models as we saw above. At the fine-grained extreme are network models where each individual's state is tracked and infections occur along edges of a sparse graph (`pyro.contrib.epidemiology` does not implement network models). We now consider an mid-level model where each of many regions (e.g. countries or zip codes) is tracked in aggregate, and infections occur both within regions and between pairs of regions. In Pyro we model multiple regions with a [plate](http://docs.pyro.ai/en/stable/primitives.html#pyro.primitives.plate). Pyro's [CompartmentalModel](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel) class does not support general `pyro.plate` syntax, but it does support a single special `self.region_plate` for regional models. This plate is available iff a `CompartmentalModel` is initialized with a vector `population`, and the size of the `region_plate` will be the length of the `population` vector.\n", "\n", "Let's take a look at the example [RegionalSIRModel](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.models.RegionalSIRModel):" ] }, { "cell_type": "code", "execution_count": 17, "metadata": {}, "outputs": [], "source": [ "class RegionalSIRModel(CompartmentalModel):\n", " def __init__(self, population, coupling, recovery_time, data):\n", " duration = len(data)\n", " num_regions, = population.shape\n", " assert coupling.shape == (num_regions, num_regions)\n", " assert (0 <= coupling).all()\n", " assert (coupling <= 1).all()\n", " assert isinstance(recovery_time, float)\n", " assert recovery_time > 1\n", " if isinstance(data, torch.Tensor):\n", " # Data tensors should be oriented as (time, region).\n", " assert data.shape == (duration, num_regions)\n", " compartments = (\"S\", \"I\") # R is implicit.\n", "\n", " # We create a regional model by passing a vector of populations.\n", " super().__init__(compartments, duration, population, approximate=(\"I\",))\n", "\n", " self.coupling = coupling\n", " self.recovery_time = recovery_time\n", " self.data = data\n", "\n", " def global_model(self):\n", " # Assume recovery time is a known constant.\n", " tau = self.recovery_time\n", "\n", " # Assume reproductive number is unknown but homogeneous.\n", " R0 = pyro.sample(\"R0\", dist.LogNormal(0., 1.))\n", "\n", " # Assume response rate is heterogeneous and model it with a\n", " # hierarchical Gamma-Beta prior.\n", " rho_c1 = pyro.sample(\"rho_c1\", dist.Gamma(10, 1))\n", " rho_c0 = pyro.sample(\"rho_c0\", dist.Gamma(10, 1))\n", " with self.region_plate:\n", " rho = pyro.sample(\"rho\", dist.Beta(rho_c1, rho_c0))\n", "\n", " return R0, tau, rho\n", "\n", " def initialize(self, params):\n", " # Start with a single infection in region 0.\n", " I = torch.zeros_like(self.population)\n", " I[0] += 1\n", " S = self.population - I\n", " return {\"S\": S, \"I\": I}\n", "\n", " def transition(self, params, state, t):\n", " R0, tau, rho = params\n", "\n", " # Account for infections from all regions. This uses approximate (point\n", " # estimate) counts I_approx for infection from other regions, but uses\n", " # the exact (enumerated) count I for infections from one's own region.\n", " I_coupled = state[\"I_approx\"] @ self.coupling\n", " I_coupled = I_coupled + (state[\"I\"] - state[\"I_approx\"]) * self.coupling.diag()\n", " I_coupled = I_coupled.clamp(min=0) # In case I_approx is negative.\n", " pop_coupled = self.population @ self.coupling\n", "\n", " with self.region_plate:\n", " # Sample flows between compartments.\n", " S2I = pyro.sample(\"S2I_{}\".format(t),\n", " infection_dist(individual_rate=R0 / tau,\n", " num_susceptible=state[\"S\"],\n", " num_infectious=I_coupled,\n", " population=pop_coupled))\n", " I2R = pyro.sample(\"I2R_{}\".format(t),\n", " binomial_dist(state[\"I\"], 1 / tau))\n", "\n", " # Update compartments with flows.\n", " state[\"S\"] = state[\"S\"] - S2I\n", " state[\"I\"] = state[\"I\"] + S2I - I2R\n", "\n", " # Condition on observations.\n", " t_is_observed = isinstance(t, slice) or t < self.duration\n", " pyro.sample(\"obs_{}\".format(t),\n", " binomial_dist(S2I, rho),\n", " obs=self.data[t] if t_is_observed else None)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "The main differences from the earlier univariate model are that: we assume `population` is a vector of length `num_regions`, we sample all compartmental variables and some global variables inside the `region_plate`, and we compute coupled vectors `I_coupled` and `pop_coupled` of the effective number of infected individuals and population accounting for both intra-region and inter-region infections. Among global variables we have chosen for demonstration purposes to make `tau` a fixed single number, `R0` a single latent variable shared among all regions, and `rho` a local latent variable that can take a different value for each region. Note that while `rho` is not shared among regions, we have created a hierarchical model whereby `rho`'s parent variables are shared among regions. While some of our variables are region-global and some region-local, only the compartmental variables are both region-local and time-dependent; all other parameters are fixed for all time. See the [heterogeneous models](#Heterogeneous-models) section below for time-dependent latent variables.\n", "\n", "Note that Pyro's enumerated MCMC strategy (`.fit_mcmc()` with `num_quant_bins > 1`) requires extra logic to use a mean-field approximation across compartments: we pass `approximate=(\"I\",)` to the constructor and force compartements to iteract via `state[\"I_approx\"]` rather than `state[\"I\"]`. This code is not required for SVI inference or for moment-matched MCMC inference (`.fit_mcmc()` with the default `num_quant_bins=0`).\n", "\n", "See the [Epidemiology: regional models](http://pyro.ai/examples/epi_regional.html) example for a demonstration of how to generate data, train, predict, and forecast with regional models. " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Phylogenetic likelihoods \n", "\n", "Epidemiological parameters can be difficult to identify from aggregate observations alone. However some parameters like the superspreading parameter `k` can be more accurately identified by combining aggregate count data with viral phylogenetic trees reconstructed from viral genetic sequencing data [(Li et al. 2017)](#1). Pyro implements a [CoalescentRateLikelihood](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.distributions.CoalescentRateLikelihood) class to compute a population likelihood `p(I|phylogeny)` given statistics of a phylogenetic tree (or a batch of tree samples). The statistics needed are exactly the times of each sampling event (i.e. when a viral genome was sequenced) and the times of genetic coalescent events in a binary phylogenetic tree; let us call these two vectors ``leaf_times`` and ``coal_times``, respectively, where ``len(leaf_times) == 1 + len(coal_times)`` for binary trees. Pyro provides a helper [bio_phylo_to_times()](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.distributions.coalescent.bio_phylo_to_times) to extract these statistics from a [Bio.Phylo tree objects](https://biopython.readthedocs.io/en/latest/api/Bio.Phylo.BaseTree.html#Bio.Phylo.BaseTree.Clade); in turn [Bio.Phylo](https://biopython-tutorial.readthedocs.io/en/latest/notebooks/13%20-%20Phylogenetics%20with%20Bio.Phylo.html) can parse many file formats of phylogenetic trees.\n", "\n", "Let's take a look at the [SuperspreadingSEIRModel](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.models.SuperspreadingSEIRModel) which includes a phylogenetic likelihood. We'll focus on the phylogenetic parts of the model:\n", "\n", "```python\n", "class SuperspreadingSEIRModel(CompartmentalModel):\n", " def __init__(self, population, incubation_time, recovery_time, data, *,\n", " leaf_times=None, coal_times=None):\n", " compartments = (\"S\", \"E\", \"I\") # R is implicit.\n", " duration = len(data)\n", " super().__init__(compartments, duration, population)\n", " ...\n", " self.coal_likelihood = dist.CoalescentRateLikelihood(\n", " leaf_times, coal_times, duration)\n", " ...\n", " \n", " def transition(self, params, state, t):\n", " ...\n", " # Condition on observations.\n", " t_is_observed = isinstance(t, slice) or t < self.duration\n", " R = R0 * state[\"S\"] / self.population\n", " coal_rate = R * (1. + 1. / k) / (tau_i * state[\"I\"] + 1e-8)\n", " pyro.factor(\"coalescent_{}\".format(t),\n", " self.coal_likelihood(coal_rate, t)\n", " if t_is_observed else torch.tensor(0.))\n", "```\n", "We first constructed a ``CoalescentRateLikelihood`` object to be used throughout inference and prediction; this performs preprocessing work once so that it is cheap to evaluate ``self.coal_likelihood(...)``. Note that ``(leaf_times, coal_times)`` should be in units of time steps, the same time steps as the time index `t` and `duration`. Typically ``leaf_times`` are in ``[0, duration)``, but ``coal_times`` precede ``leaf_times`` (as points of common ancestry), and may be negative. The likelihood involves the coalescent rate ``coal_rate`` in a coalescent process; we can compute this from an epidemiological model. In this superspreading model ``coal_rate`` depends on the reproductive number ``R``, the superspreading parameter ``k``, the incubation time ``tau_i``, and the current number of infected individuals ``state[\"I\"]`` [(Li et al. 2017)](#1)." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Heterogeneous models \n", "\n", "Epidemiological parameters often vary in time, due to human interventions, changes in weather, and other external factors. We can model real-valued time-varying latent variables in ``CompartmentalModel`` by moving static latent variables from ``.global_model()`` to ``.initialize()`` and ``.transition()``. For example we can model a reproductive number under Brownian drift in log-space by initializing at a random ``R0`` and multiplying by a drifting factor, as in the [HeterogeneousSIRModel](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.models.HeterogeneousSIRModel) example:\n", "```python\n", "class HeterogeneousSIRModel(CompartmentalModel):\n", " ...\n", " def global_model(self):\n", " tau = self.recovery_time\n", " R0 = pyro.sample(\"R0\", dist.LogNormal(0., 1.))\n", " rho = ...\n", " return R0, tau, rho\n", "\n", " def initialize(self, params):\n", " # Start with a single infection.\n", " # We also store the initial beta value in the state dict.\n", " return {\"S\": self.population - 1, \"I\": 1, \"beta\": torch.tensor(1.)}\n", "\n", " def transition(self, params, state, t):\n", " R0, tau, rho = params\n", " # Sample heterogeneous variables.\n", " # This assumes beta slowly drifts via Brownian motion in log space.\n", " beta = pyro.sample(\"beta_{}\".format(t),\n", " dist.LogNormal(state[\"beta\"].log(), 0.1))\n", " Rt = pyro.deterministic(\"Rt_{}\".format(t), R0 * beta)\n", "\n", " # Sample flows between compartments.\n", " S2I = pyro.sample(\"S2I_{}\".format(t),\n", " infection_dist(individual_rate=Rt / tau,\n", " num_susceptible=state[\"S\"],\n", " num_infectious=state[\"I\"],\n", " population=self.population))\n", " ...\n", " # Update compartments and heterogeneous variables.\n", " state[\"S\"] = state[\"S\"] - S2I\n", " state[\"I\"] = state[\"I\"] + S2I - I2R\n", " state[\"beta\"] = beta # We store the latest beta value in the state dict.\n", " ...\n", "```\n", "Here we deterministically initialize a scale factor ``beta = 1`` in ``.initialize()`` then let it drift via log-Brownian motion. We also need to update ``state[\"beta\"]`` just as we update the compartmental variables. Now ``beta`` will be provided as a time series when we ``.predict()``. While we could have written ``Rt = R0 * beta``, we instead wrapped this computation in a ``pyro.deterministic`` thereby exposing ``Rt`` as another time series provided by ``.predict()``. Note that we could have instead sampled ``R0`` in ``.initialize()`` and let ``Rt`` drift directly, rather than introducing a scale factor ``beta``. However separating the two into a non-centered form improves geometry [(Betancourt and Girolami 2013)](#2).\n", "\n", "It is also easy to pass in time-varying covariates as tensors, in the same way we have passed in ``data`` to the constructors of all example models. To predict the effects of different causal interventions, you can pass in a covariate that is longer than ``duration``, run inference (looking only at the first ``[0,duration)`` entries), then mutate entries of the covariate after ``duration`` and generate different ``.predict()``ions." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Complex compartment flow \n", "\n", "The [CompartmentalModel](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel) class assumes by default that the compartments are arranged linearly and terminate in an implicit terminal compartment named \"R\", for example S-I-R, S-E-I-R or boxcar models like S-E1-E2-I1-I2-I3-R. To describe other more complex flows between compartments, you can override the [.compute_flows()](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.compartmental.CompartmentalModel.compute_flows) method. However currently there is no support for flows with undirected loops (e.g. S-I-S).\n", "\n", "Let's create a branching SIRD model with possible flows\n", "```\n", "S → I → R\n", " ↘\n", " D\n", "```\n", "As with other models, we'll keep the \"R\" state implicit (although we could equally keep the \"D\" state implicit and the \"R\" state explicit). In the ``.compute_flows()`` method, we'll input a pair of states and we'll need to compute three flow values: ``S2I``, ``I2R``, and ``I2D``.\n", "```python\n", "class SIRDModel(CompartmentalModel):\n", " def __init__(self, population, data):\n", " compartments = (\"S\", \"I\", \"D\")\n", " duration = len(data)\n", " super().__init__(compartments, duration, population)\n", " self.data = data\n", "\n", " def compute_flows(self, prev, curr, t):\n", " S2I = prev[\"S\"] - curr[\"S\"] # S can only go in one direction.\n", " I2D = curr[\"D\"] - prev[\"D\"] # D can only have come from one direction.\n", " # Now by conservation at I, change + inflows + outflows = 0,\n", " # so we can solve for the single unknown I2R.\n", " I2R = prev[\"I\"] - curr[\"I\"] + S2I - I2D\n", " return {\n", " \"S2I_{}\".format(t): S2I,\n", " \"I2D_{}\".format(t): I2D,\n", " \"I2R_{}\".format(t): I2R,\n", " }\n", " ...\n", " def transition(self, params, state, t):\n", " ...\n", " # Sample flows between compartments.\n", " S2I = pyro.sample(\"S2I_{}\".format(t), ...)\n", " I2D = pyro.sample(\"I2D_{}\".format(t), ...)\n", " I2R = pyro.sample(\"I2R_{}\".format(t), ...)\n", "\n", " # Update compartments with flows.\n", " state[\"S\"] = state[\"S\"] - S2I\n", " state[\"I\"] = state[\"I\"] + S2I - I2D - I2R\n", " state[\"D\"] = state[\"D\"] + I2D\n", " ...\n", "```\n", "Note you can name the dict keys anything you want, as long as they match your sample statements in ``.transition()`` and you correctly reverse the flow computation in ``.transition()``. During inference Pyro will check that the ``.compute_flows()`` and ``.transition()`` computations agree. Take care to avoid in-place PyTorch operations, since these can modify the tensors rather than the dictionary:\n", "```diff\n", "+ state[\"S\"] = state[\"S\"] - S2I # Correct\n", "- state[\"S\"] -= S2I # AVOID: may corrupt tensors\n", "```\n", "\n", "For a slightly more complex example, take a look at the [SimpleSEIRDModel](http://docs.pyro.ai/en/latest/contrib.epidemiology.html#pyro.contrib.epidemiology.models.SimpleSEIRDModel)." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## References\n", "\n", "1. \n", " Lucy M. Li, Nicholas C. Grassly, Christophe Fraser (2017)\n", " \"Quantifying Transmission Heterogeneity Using Both Pathogen Phylogenies\n", " and Incidence Time Series\"\n", " https://academic.oup.com/mbe/article/34/11/2982/3952784\n", "2. \n", " M. J. Betancourt, Mark Girolami (2013)\n", " \"Hamiltonian Monte Carlo for Hierarchical Models\"\n", " https://arxiv.org/abs/1312.0906" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] } ], "metadata": { "kernelspec": { "display_name": "Python 3", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.7.7" } }, "nbformat": 4, "nbformat_minor": 4 }