{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "# Gaussian process regression with varying output noise" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "This notebook shows how to construct a Gaussian process model where different noise is assumed for different data points. The model is:\n", "\n", "$$f(\\cdot) \\sim \\mathcal{GP}\\big(0, k(\\cdot, \\cdot)\\big)$$\n", "$$y_i | f, x_i \\sim \\mathcal N\\big(y_i; f(x_i), \\sigma^2_i\\big)$$\n", "\n", "We'll demonstrate two methods. In the first demonstration, we'll assume that the noise variance is known for every data point. We'll incorporate the known noise variances $\\sigma^2_i$ into the data matrix $\\mathbf Y$, make a likelihood that can deal with this structure, and implement inference using Variational GPs with natural gradients. \n", "\n", "In the second demonstration, we'll assume that the noise variance is not known, but we'd like to estimate it for different groups of data. We'll show how to construct an appropriate likelihood for this task and set up inference similarly to the first demonstration, with optimisation over the noise variances. \n" ] }, { "cell_type": "code", "execution_count": 1, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:55.216874Z", "start_time": "2019-04-05T16:22:53.892098Z" } }, "outputs": [], "source": [ "import numpy as np\n", "import tensorflow as tf\n", "import gpflow\n", "from gpflow.test_util import notebook_range\n", "import matplotlib\n", "import matplotlib.pyplot as plt\n", "%matplotlib inline" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Demo 1: known noise variances\n", "### Generate synthetic data\n", "We create a utility function to generate synthetic data, including noise that varies amongst the data:" ] }, { "cell_type": "code", "execution_count": 2, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:55.222487Z", "start_time": "2019-04-05T16:22:55.218562Z" } }, "outputs": [], "source": [ "np.random.seed(1) # for reproducibility\n", "\n", "def generate_data(N=80):\n", " X = np.random.rand(N)[:, None] * 10 - 5 # Inputs, shape N x 1\n", " F = 2.5 * np.sin(6 * X) + np.cos(3 * X) # Mean function values\n", " NoiseVar = 2 * np.exp(-(X - 2)**2 / 4) + 0.3 # Noise variances\n", " Y = F + np.random.randn(N, 1) * np.sqrt(NoiseVar) # Noisy data\n", " return X, Y, NoiseVar\n", "\n", "X, Y, NoiseVar = generate_data()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "Here's a plot of the data, with error bars representing two standard deviations:" ] }, { "cell_type": "code", "execution_count": 3, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:55.421858Z", "start_time": "2019-04-05T16:22:55.223920Z" } }, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "fig, ax = plt.subplots(1, 1, figsize=(12, 6))\n", "_ = ax.errorbar(X, Y, yerr=2*(np.sqrt(NoiseVar)), marker='x', lw=0, elinewidth=1., color='C1')" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Make a Y matrix that includes the variances\n", "We need to tell the GP model what the variance is for each data point. To do this, we'll concatenate the observations with the variances into a single data matrix:" ] }, { "cell_type": "code", "execution_count": 4, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:55.427016Z", "start_time": "2019-04-05T16:22:55.424378Z" } }, "outputs": [], "source": [ "Y_data = np.hstack([Y, NoiseVar])" ] }, { "cell_type": "markdown", "metadata": { "ExecuteTime": { "end_time": "2018-06-19T10:58:59.599039Z", "start_time": "2018-06-19T10:58:59.546311Z" } }, "source": [ "### Make a new likelihood\n", "\n", "To cope with this data structure, we'll build a new likelihood. Note how the code extracts the observations `Y` and the variances `NoiseVar` from the data. For more information on creating new likelihoods, see [Likelihood design](../tailor/likelihood_design.ipynb). Here, we're implementing the `logp` function (which computes the log-probability of the data given the latent function) and `variational_expectations`, which computes the expected log-probability under a Gaussian distribution on the function, and is needed in the evaluation of the evidence lower bound (ELBO). Check out the docstring for the `Likelihood` object for more information on what these functions do. " ] }, { "cell_type": "code", "execution_count": 5, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:55.440109Z", "start_time": "2019-04-05T16:22:55.429578Z" } }, "outputs": [], "source": [ "class HetGaussian(gpflow.likelihoods.Likelihood):\n", " def logp(self, F, Y):\n", " # logp is used by the quadrature fallback of variational_expectations and predict_density.\n", " # Because variational_expectations is implemented analytically below, this is not actually needed,\n", " # but is included for pedagogical purposes.\n", " # Note that currently relying on the quadrature would fail due to https://github.com/GPflow/GPflow/issues/966\n", " Y, NoiseVar = Y[:, 0:1], Y[:, 1:2]\n", " return gpflow.logdensities.gaussian(Y, F, NoiseVar)\n", "\n", " def conditional_mean(self, F):\n", " raise NotImplementedError\n", "\n", " def conditional_variance(self, F):\n", " raise NotImplementedError\n", "\n", " def variational_expectations(self, Fmu, Fvar, Y):\n", " Y, NoiseVar = Y[:, 0:1], Y[:, 1:2]\n", " return -0.5 * np.log(2 * np.pi) - 0.5 * tf.log(NoiseVar) \\\n", " - 0.5 * (tf.square(Y - Fmu) + Fvar) / NoiseVar" ] }, { "cell_type": "markdown", "metadata": { "ExecuteTime": { "end_time": "2018-06-19T10:59:02.113396Z", "start_time": "2018-06-19T10:59:01.860960Z" } }, "source": [ "### Put it together with Variational Gaussian Process (VGP)\n", "Here we'll build a variational GP model with the previous likelihood on the dataset that we generated. We'll use the natural gradient optimiser (see [Natural gradients](natural_gradients.ipynb) for more information).\n", "\n", "The variational GP object is capable of variational inference with any GPflow-derived likelihood. Usually, the inference is an inexact (but pretty good) approximation, but in the special case considered here, where the noise is Gaussian, it will achieve exact inference. Optimising over the variational parameters is easy using the natural gradients method, which provably converges in a single step. " ] }, { "cell_type": "code", "execution_count": 6, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:55.623449Z", "start_time": "2019-04-05T16:22:55.441186Z" } }, "outputs": [], "source": [ "# model construction (notice that num_latent is 1)\n", "likelihood = HetGaussian()\n", "kern = gpflow.kernels.Matern52(input_dim=1, lengthscales=0.5)\n", "model = gpflow.models.VGP(X, Y_data, kern=kern, likelihood=likelihood, num_latent=1)" ] }, { "cell_type": "code", "execution_count": 7, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:56.089503Z", "start_time": "2019-04-05T16:22:55.624821Z" } }, "outputs": [], "source": [ "# build the natural gradients optimiser\n", "natgrad_optimizer = gpflow.training.NatGradOptimizer(gamma=1.)\n", "natgrad_tensor = natgrad_optimizer.make_optimize_tensor(model, var_list=[(model.q_mu, model.q_sqrt)])\n", "\n", "# A single natural gradient step solves the problem:\n", "session = model.enquire_session()\n", "session.run(natgrad_tensor)\n", "\n", "# update the cache of the variational parameters in the current session\n", "model.anchor(session)" ] }, { "cell_type": "code", "execution_count": 8, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:56.465100Z", "start_time": "2019-04-05T16:22:56.090741Z" } }, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "# let's do some plotting!\n", "xx = np.linspace(-5, 5, 200)[:, None]\n", "\n", "mu, var = model.predict_f(xx)\n", "\n", "plt.figure(figsize=(12, 6))\n", "plt.plot(xx, mu, 'C0')\n", "plt.plot(xx, mu + 2*np.sqrt(var), 'C0', lw=0.5)\n", "plt.plot(xx, mu - 2*np.sqrt(var), 'C0', lw=0.5)\n", "\n", "plt.errorbar(X, Y, yerr=2*(np.sqrt(NoiseVar)), marker='x', lw=0, elinewidth=1., color='C1')\n", "_ = plt.xlim(-5, 5)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Questions for the reader\n", "1) What is the difference in meaning between the orange vertical bars and the blue regions in the prediction?\n", "\n", "2) Why did we not implement `conditional_mean` and `conditional_var` in the HetGaussian likelihood? What could be done here?\n", "\n", "2) What are some better kernel settings for this dataset? How could they be estimated?" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Demo 2: grouped noise variances\n", "\n", "In this demo, we won't assume that the noise variances are known, but we will assume that they're known in two groups. This example represents a case where we might know that an instrument has varying fidelity for different regions, but we do not know what those fidelities are. \n", "\n", "Of course it would be straightforward to add more groups, or even one group per data point. We'll stick with two for simplicity. " ] }, { "cell_type": "code", "execution_count": 9, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:56.469974Z", "start_time": "2019-04-05T16:22:56.466581Z" } }, "outputs": [], "source": [ "# This is a completely independent demo. Reset the TensorFlow default graph and session\n", "# to make the TensorFlow graph computations run faster:\n", "gpflow.reset_default_graph_and_session()\n", "\n", "np.random.seed(1) # for reproducibility and to make it independent from demo 1" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Generate data" ] }, { "cell_type": "code", "execution_count": 10, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:56.482574Z", "start_time": "2019-04-05T16:22:56.471315Z" } }, "outputs": [], "source": [ "def generate_data(N=100):\n", " X = np.random.rand(N)[:, None] * 10 - 5 # Inputs, shape N x 1\n", " F = 2.5 * np.sin(6 * X) + np.cos(3 * X) # Mean function values\n", " groups = np.where(X>0, 0, 1) \n", " NoiseVar = np.array([0.02, 0.5])[groups] # Different variances for the two groups\n", " Y = F + np.random.randn(N, 1) * np.sqrt(NoiseVar) # Noisy data\n", " return X, Y, groups\n", "\n", "X, Y, groups = generate_data()" ] }, { "cell_type": "code", "execution_count": 11, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:56.647640Z", "start_time": "2019-04-05T16:22:56.483739Z" } }, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "# here's a plot of the raw data.\n", "fig, ax = plt.subplots(1, 1, figsize=(12, 6))\n", "_ = ax.plot(X, Y, 'kx')" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Data structure\n", "\n", "In this case, we need to let the model know which group each data point belongs to. We'll use a similar trick to the above, stacking the group identifier with the data:" ] }, { "cell_type": "code", "execution_count": 12, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:56.651242Z", "start_time": "2019-04-05T16:22:56.648903Z" } }, "outputs": [], "source": [ "Y_data = np.hstack([Y, groups])" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Build a likelihood\n", "\n", "This time, we'll use a built-in likelihood, `SwitchedLikelihood`, which is a container for other likelihoods, and applies them to the first `Y_data` column depending on the index in the second. We're able to access and optimise the parameters of those likelihoods. Here, we'll (incorrectly) initialise the variances of our likelihoods to 1, to demonstrate how we can recover reasonable values for these through maximum-likelihood estimation. " ] }, { "cell_type": "code", "execution_count": 13, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:56.686784Z", "start_time": "2019-04-05T16:22:56.652610Z" } }, "outputs": [], "source": [ "likelihood = gpflow.likelihoods.SwitchedLikelihood([gpflow.likelihoods.Gaussian(variance=1.0),\n", " gpflow.likelihoods.Gaussian(variance=1.0)])" ] }, { "cell_type": "code", "execution_count": 14, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:56.849105Z", "start_time": "2019-04-05T16:22:56.687804Z" } }, "outputs": [], "source": [ "# model construction (notice that num_latent is 1)\n", "kern = gpflow.kernels.Matern52(input_dim=1, lengthscales=0.5)\n", "model = gpflow.models.VGP(X, Y_data, kern=kern, likelihood=likelihood, num_latent=1)" ] }, { "cell_type": "code", "execution_count": 15, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:57.450347Z", "start_time": "2019-04-05T16:22:56.850434Z" } }, "outputs": [], "source": [ "# build the natural gradients optimiser\n", "natgrad_optimizer = gpflow.training.NatGradOptimizer(gamma=1.)\n", "natgrad_tensor = natgrad_optimizer.make_optimize_tensor(model, var_list=[(model.q_mu, model.q_sqrt)])\n", "\n", "# A single natural gradient step solves the problem:\n", "session = model.enquire_session()\n", "session.run(natgrad_tensor)\n", "\n", "# update the cache of the variational parameters in the current session\n", "model.anchor(session)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "We've now fitted the VGP model to the data, but without optimising over the hyperparameters. Plotting the data, we see that the fit is not terrible, but hasn't made use of our knowledge of the varying noise. " ] }, { "cell_type": "code", "execution_count": 16, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:57.827067Z", "start_time": "2019-04-05T16:22:57.451629Z" } }, "outputs": [ { "data": { "image/png": "iVBORw0KGgoAAAANSUhEUgAAAsUAAAFpCAYAAAB0/VUQAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAIABJREFUeJzs3Xd4W+XZP/DvkSxZ3lPee9txtuPsPUlCCBQKlIYdWtpCC6Xt277w69uWTijQAS2EFTYUQtIssghk24njkcR7b1veU7bG+f2hCDu2xpF0tO/PdeUiONLRSSzr3Od57sGwLAtCCCGEEELcmcDeJ0AIIYQQQoi9UVBMCCGEEELcHgXFhBBCCCHE7VFQTAghhBBC3B4FxYQQQgghxO1RUEwIIYQQQtweBcWEEEIIIcTtUVBMCCGEEELcHgXFhBBCCCHE7VFQTAghhBBC3J6HPV40NDSUTUhIsMdLE0IIIYQQN1JQUNDFsqzU2OPsEhQnJCTg0qVL9nhpQgghhBDiRhiGaeDyOEqfIIQQQgghbo+CYkIIIYQQ4vYoKCaEEEIIIW6PgmJCCCGEEOL2KCgmhBBCCCFuj4JiQgghhBDi9igoJoQQQgghbo+CYkIIIYQQ4vYoKLan/F3AkGzi/4dkmq8RQgghhBCb4m2iHcMwQgCXALSwLLuVr+O6rPxdwKGngIuvA/cd0Hxt91ZAVq75fe5O+50bIYQQQoib4XPM848BlAHw5/GYritruyYglpUDryzSfG2kC5BmaP6MEEIIIYTYDC/pEwzDxADYAuB1Po7nFnylmhVi71BNMDzSpfn9fQc0f0YIIYQQQmyGr5zilwD8HICap+MRQgghhBBiMxYHxQzDbAXQybJsgZHHPcIwzCWGYS7JZDJDD3UPQzJNDrF2hVi7Yrx7643Fd4QQQgghxOr4WCleCmAbwzD1AD4CsIZhmPemPohl2ddYls1hWTZHKqX0AJTu1eQTSzOAH1zQ/JJmaL5WutfeZ0cIIYQQ4lYsLrRjWfaXAH4JAAzDrALwFMuy37X0uC5P210ia/tEDvF9BzQBMXWeIIQQQgixKT67TxBTTQ1+faUUEBNCCCGE2AGvQTHLsl8B+IrPYxJCCCGEEGJtNNHOXmiaHSGEEEKIw6D0CXugaXaEEEIIIQ6FgmJ7oGl2hBBCCCEOhdIn7IGm2RFCCCGEOBQKigkhhBBCiNujoNgeaJodIYQQQohDoaDYHmiaHSGEEEKIQ7FPoZ1aOfH7IZn7TXGjaXaEEEIIIQ7FPkFxV9VEmoC7tiKjaXaEEEIIIQ7DPkGxUk6tyAghhBBCiMOwT1As8NAEwwC1IiOEEEIIIXZHhXaEEEIIIcTt2ScoViupFRkhhBBCCHEY9gmKPSTUiowQQgghhDgM++QUh6ZSKzJCCCGEEOIw7LNSLJgUi1MrMkIIIYQQYmdUaEcIIYQQQtweBcWEEEIIIcQ8+btubJYwJNN8zQnZJ6eYEEIIIYQ4t/xdwKGngIuva2rEAKeeVEwrxYQQQog7cqEVPmInWdsnuoi9skjzS1butJOKKSgmhBBC3I12hU87J2BIpvn9oacoMCbc+Uo1K8TauRMjXU49qZiCYkKcDa3uEEIs5WIrfITwgYJiQpwJre4QQvjgYit8xE601yDt+8fJJxVTUEyIM6HVHUIIIY6idO/ENcgFJhVT9wlCnIl2deeVRZq7cYBWdwghppu6wgdMrPDR5wnhSttdImu7S0wqppViR0a5o+6HvueEEFtwsRU+Yke5O2+8iXLiScW0UuyoXKz3H+GAy/ecVncIIXxwsRU+QvhAQbGjytquCY60uaOAJvih3FHXxeV7Pnl1Z2rgTBczQogppn5eOPEKHyF8YFiWtfmL5uTksJcuXbL56zqdIdn03NEfXKDVQFfG5Xuev+vG1Z0hGQXEhBBCiB4MwxSwLJtj7HGUU2xtlCNK+OZC+VuEEEKIo6Cg2Jos6SnrYr3/CAf0PSeEEELsxuKgmGEYCcMw+QzDFDMMc41hmN/wcWIuwZKeslQZ7H64fM9p54EQQogjc+LrFB+FdmMA1rAsO8QwjAjAGYZhDrMse4GHYzs3S3rKUmWw+zH2PaeOJIQQQhyZk1+nLA6KWU2l3tD1/xVd/2X76j1XRJXB7ldUZuh7Th1JCCGEODInv07xklPMMIyQYZgiAJ0AjrEsm6fjMY8wDHOJYZhLMpmb5EdSjqhlLMnJdkXanQft+0j7vqL+xIQQQhyBk1+neAmKWZZVsSw7B0AMgFyGYbJ1POY1lmVzWJbNkUod/x+GF5QXbBlLcrIJIYQQQkzAa/cJlmX7AJwEsInP4zqt3J3A5ucn7pC0d1Cbn3fd7X8+OfkdJ+9o54EQQogjc/LrFB/dJ6QMwwRe/70XgPUAyi09rsugnrKEL7TzQAxx4opvQoiLcPLrFB/dJyIB7GYYRghNkP0Jy7IHeDgucXdT7ziBiTtOd1wtpo4kRB8nr/gmhLgIJ79O0ZhnG2NZFr0jCrT0jqK5dwRh/hLMjw+y92k5Ju2FXpox/UJPKSiETNDeQMrKb7yB1P7suNsNJCGETMJ1zDMfK8WEI6VKjT8eLkdMkBeiA70QH+KDi/U96Boaw8YZEfY+Pcfj5HechNiMJT3RCSGEAKCg2GbUahbPHa3APQvjkCT1/ebrWVH++PhiIw6UtGLrrCg7nqGDol7NhBBCCLEBXrtPEN1YlsXfTlRh2+yoGwJirTsXxGFcqcbnhc12ODtCiNNz8opvQghxBBQU28CbZ+uxPDUUM6IC9D7mtnkx6BgYQ3u/3IZnRghxCeZUfFO3CkIIuQGlT1hZfdcwGAA5CcFGH3t3bhx2n6vH42tTrX9ihBDXYWr+PXWrIISQaSgotrIP8xvxxPo0To8N8BIBAPpHFAjwFlnztAghrsaU/Pus7ZqAWDstEpjoVkHTIgkhborSJ6zoSnM/UsJ8IREJOT/nzgWx+PhSoxXPihDi9mhaJCGETENBsRV9XtiC2+bFmPSccH8JBkaVGB1XWemsCCGEEELIVBQUW8npKhkWJ4dAKGBMfu5t86Lx6WXqREEIsZLJ3SpEPoBX8ES3io5yKrgjhLglCoqtQK1m8WV5J9Zlhpn1/CSpL5p7RqBUqXk+M0IIwUS3Ct9wQDEMeAcDISmar727TVOER4ExIcTNUKGdFZyp7sKajDAwjOmrxFrrssLxVYUM67LCeTwzQgjBRAFe/HLg0/s0wbBXMCD2AYY6qOCOEOKWaKXYCs7WdGFpcqhFx5gXF4Sipj6ezogQF0b9ds2TuxMIz5gouBvtAcaHqeCOEOK2aKWYZx0Dckh9PSEwI5d4MqGAAQsWajVr8bEIcVnUb5cQQghPaKWYZ3sLW7B9bjQvx5oVE4iSln5ejkWIS8raPjG57ZVFml/ayW60/W8cjYcmhJBvUFDMI5WaRe+IAqG+nrwcb3lqKE5X0oWJEL2o365lzBkPTQghLorSJ3h0qlKGlWn8XYi9xR4YVVC/YkKIlZg6HpoQQlwYrRTz6EJtNxYlBfN6zKhALzT3jvB6TEKckq6CutMv0Pa/pXJ33riqbmg8NCGEuDAKinnS2jeKiACJRW3YdFmbGYYvyzt5PSYhTkdbUKcNdrW5sCd+Q9v/UyhVapQ092HP5WYMjyntfTqEEOI0KH2CJ/8tbsWdObG8HzcywAvt/XLej0uIU8narukwoS2oAzQrwtIMYNadwNwdbr/9X9TUh8NX2+AlEiIr0h9ZUf545atqhPtLcOeCWHh6CO19ioQQ4tAoKOYBy7LoHRlHkI/YKsf38fTAoFwBP4nIKscnxOFpC+peWaQJhgH9BXVuuP2/v7gVHQNy/M+mjBt2qzIi/FEjG8KfDpfjrgVxSI/ws+NZEkLcVv6uG2sXhmQOuXhB6RM8KG8fRFakv9WOvzJNilOVXVY7PiHEOanVLF75qhpiDwEeXp6kM30rWeqLp7dk4YO8BsgNFe7SEBSig1yhwtWWfnxZ3gG1mrX36RAHplazOHKtHfVdwzf+gb70NwccJ09BMQ9OlHVgXSbHccxmXHhmRPmjrG3AgjMkxMlRP91pWJbFX49VYGWaFBtnRBh8rFDA4Hsrk/HKVzW6H+BEFy1iG1db+vHb/aV440wdmntHIBYK8dsDpegdHrf3qREH1NQzgt/svwY/iQfy6rrxly/K8dLxSvQMjztVP3lKn7AQy7IYHlfBx5PDP6WZ07cYhgHP9XuEOJfJ/XSn/uw44BacLXyQ34jlqVLMiArg9PioQC9kRfrhRFkH1k69iTeUs+1gFy1ife39cuwrasEzWzNv2H2YFRuAvx2vwrbZUZgdG2jHMySOgmVZfFrQjPZ+OX65ORMSkRBI1vzZoFyBv3xRgf/dkgkJ1/Q3O7PLSvGQC1VElzT3Y1Y0t4uSJXdLwT5idA2N8XDGhDih3J3A5ucnPkS1Ocabn3fLgPhcTRcEDINFSSEmPW9TdiQuN/aic2BK8S4NQXEPHHYqR8aVePlkNX66IX1aOo6/RISnt2TiYn0PztVQSh8BDl9tR5C3GI+tTdUExJP4SUT4/qpkvHCsEiycI/XGLkHxoNx1guIvyzuxOiOM24MtuPDMjw9CQUMvD2dMiJOifroANNuUZ6u7cNcC87rd/HB1Ct4+V8/vSRHHxyFFRq1m8fyRSjy2NmVagKPFMAweWpaI46WdhnPUiXPjeANV1NSHdVn600ejA71wS6oIva9sdIr0N7sExRKRwCVyZNVqFuMqtd4PDz5lRlJeMSHuTq5Q4bVTtfjx2jSze6J7iz0Q4CWCbHDSzhPlbLs+DjuV7+U14I6cGIT5SQweimEYPLgsAW+cqbPFmRNb41hj8NbZety/JMHo4Wb0nkTwSC0G/VIcvp+8XYLiQC8x9he32uOleXWpoRcLEoK4P8GCC49IKICKKn8JcWuvfFWD769KhtjDso/u2+fH4NOC5okvTM7ZdvCLFjGTkZ3KcaUarX1yZHLspBQT5A0/iQct1rgiDjdQDd3DEAsFiAr0Mn686+lvb6b+HaPiYIdOf7NLUMwwmhxZZx9KcaZKhmUpJuTbWXjhEQsFtF1FiJs6Wd6J9HA/RHO5CBkR4uuJQbkCY8rrnyeUs+329he3YtvsKJOec8/CeHyU30gLNq6GQ6rnO+cbcO+SeO7HzN2J25bNxft5DROv4YCfLXZryXbH/Fj851KTvV7eYio1CxXLmrZic/3Co7p3P861M/jl0Tb80v+PGFzzR05vjpkxAbjS0m/BWRNnNKZU4fPCZrz6dQ1eOFqBvx6tQP+owt6nRWxINjiG/PoebJkVydsxt86KwoHitokvUM62azOwU8kOdaK8fQBZUab12xcKGHx7QSw+vui813Jiuq8qOrEkOcTkKZmxwd4YlCsduq2f3YLiAG8RlGrWaTtRXG7sxbw4E1InrntNvgaL/16C77yeh31Frfi8cgxbLmSibmqzax3mxgXhMhXbuRWFSo0/HS5Hapgf7lkUjyfWp+HhZUl4/kjF9A4CxCWp1SxePlmNH61O4fW4WVH+KG8fAMvSKp9bMLBTWXXyPaxIM7HLyPVCrBlRAWjsGYFioIN6WrsKI6me52u6sYZrg4Ep7luSgHfON/B8wvyxOChmGCaWYZiTDMOUMgxzjWGYH3N97u3zY/DZ5Lw2J3KmqgtLU0JNes7Ra+34w6FypIb74p/fmYuCp9fjw52LMDSmxLf+dQ5FTX0Gnx/gJdLduYMmUbkklZrFX74ox32LE5AdHQBfTw8wDIMAbxF+tTkTr3xVw+lmiji39/MacOvcaG690E2UkxCMS3Sj7R4MpMh8wmzCMlOuZ5MLsU6/gO0xAxjddZPma6dfoOuPszNwA9VX8AkiAiRmF/oG+4jhJRaguXeE55PmBx8rxUoAP2VZNgvAIgA/ZBgmi8sTY4O90do/6nSjI1nW9K4TXUNj+OWeK5gR5Y+37s/F1llR8BILMTcuCJ9+fzF8PIW4+7ULuNJsPD3ihpUdmkTlktRqFs8frcC3c2KREOoz7c+9xEI8vSUTH+U3oqGbAmNXVdo6gHEVa7VBCesyw3G8rMMqxyYOSEeKzNXob2NGtL9pQc7kQqwTv0HGnk3wH6wBG5ICFL1P1x9nZ+AG6mNsMjn3fKrvLorH+3mNPJ0svywOilmWbWNZ9vL13w8CKAMQzfX5S5JDca6m29LTsKmytkHOFbqAJoj91Z4rGBxT4sU750zLQ06S+mLPo0shEQnw0vFKg8dKkvqgRjYpCHKi8YmEu93n67FlZiRSw/30PsZDKMAT69PwgYN+uBDLjI6r8NHFRk4tj8wlFDAI8BJhQE456u7q4JU2bJ1lYpCjDZK8gjX/z6rAgoFyqBvorqbrjyvQcQOlznkYPSPjCPH1tOjQ3mIPhPl5oqVv1MKT5B+vOcUMwyQAmAsgj+tzlqeE4ky1c03GOVnRidXp3POvPi1oxtHSDvxsQzrS9AQ5Uj9P3LckASfKO1HZMaj3WDnxwSho6Jn4Ak2icjm9w+PoH1Ugm8OkRIlIiPQIPxQbSb0hzuflk9V4dFUyhALrznhfnR6Gk+WdVn0N4phGxpXwFgkhEpoZCjATz2PAQjTWS9cfF5Zf34NFiaZN0dTntnkx2OOA6bO8BcUMw/gC+AzAT1iWnda4kGGYRxiGucQwzCWZbCL/VSBgEOorRocTFQ0NjynhJxFxemzHgBy/2V+KhYnBeGhZosHH3rc4AV4iIV79ulbvY2KDvdDU43h3V4Q/75xvwL2LEzg//pY50dhb1EIFU4DL5NcfvdaO7OgARAZY3n7NmIwIP5S3678RJ67rqwoZVpqwwPONyYVYXsEAJm7cVGpqG+qqTlXKsDzVtFoqfQK8RFCo1Bgdd6z3Cy9BMcMwImgC4vdZlt2j6zEsy77GsmwOy7I5UumNP4S3zo3G54UtfJyK1TV2jyA22Jvz43efq8fIuBJ//tYsCIys+AT5iHHngljsK2pBq55thWl5XzSJyqW09I3CSyxAsI+Y83OEAgYr0qT4qtLNv98ukl9f2TGIsrZBbMqOsMnrMQwDsVAw0bOYuI0rLf2YyWFHahptIVZICuAdDIAFGE2NjVDeS9cfFzQoV8BLJISHubsKOmybE419RY4V+/HRfYIB8AaAMpZlXzDnGCG+nugfVUCpUlt6OlZ3rKwD6zL1z/meTK5Q4cP8RqzPCtdZLKXLQ8sSwQJ408D4zAAv0USfWppE5VLev9CA7y4yoSH6davSpDhVKXPvJvpjg5qLtDa//uVcp8uv7x4awwd5jfjRGn7brxmzNMX5ajuIZeQKFcRCgXldBLSFWHPumcgh/v45YO2v0e2dRNcfF3SwpI3XPukAkBLmi9quYYfa5eQj5F8KYAeANQzDFF3/tdnUg6zJCMOXTpDX1j00BqkftyTz/xa3ondEgftMKJSJDfbGzbMi8WF+I/pHdBe/ZEX5o7T1eoYKTaJyGZUdg4gO8oK32PTWWwzDYLsD3nXbRP4uTRuoE7/R/L8kSLNbMno9937WnU6R3zimVOGl41V4amO61fOIp5ofH4SCemrN5k7OVndZthWeuxPw9APSbgJu3w2EZwBzd0CdvhnNURvp+uNi6rtHkCT15f24i5NDcN6BbsgtbnzJsuwZTE4oMlNOfBD+dLgcG2bYZsvQHJ2Dcs5VlyzL4u2z9UgP98PiJNMS07+3Mhl7i1rx7oV6/GhN6rQ/nxHlj08LmrE4+fpxp3740CQqp/RpQTN+tjHd7OfPjg3Eoattxh/oSrQpEyEpml/d1Tf+OSME0jajvH0AX5Z3Qn49f027oB7oLUJOQjBmxwSY3XeTD2o1ixeOVeKRFUnwtUI/YmOEAgYsWKjUrM0DcmIflxt78dP15n/efPOzBwDdVcC33wc+uQdS7c9g/i66DrmI7qExhPpyT+kzxcpUKf78RTmWmDj3wVps/+mrB8MwiAn2RmP3COJCuOfs2tLx0k5syOKWOnGxvhelbQP4420zTb7YZkb6Y0lyCD673KIzKA70FtOYXxfT1DOC6EAv86vAr8uI8ENZ24BJLQOdWtZ24OLrmu1aydRevgzAqtD19l24uvhNPLBkNrzEN/YW7xkeR35dN/YXt2JBQjA2ZIUbzf3nm1KlxvNHK3HzrCiT6hX4Njc2CEVNvZgfH2y3cyC2oVCpIWAYy97rWduBvH9rbkS7q4F/LQKgudvs846HMv4mOEaYQyz1VYUMq9LNm2BnjEDAICbIy2FiP7uNedZl+5wo7NVu/zpgFXlzL/ciu93n6hHgJcL2OZxbNt9gU3YE6rqGUSMbMuv5xLnsK2ox+70y2aYZkfjiajsPZ+QkJrcklN/Ylk4uCkCPJA6ho3W4XVIwLSAGNNOVNmVH4pmtWQjyFuEPh8pw9Fq7zXLcxpQq/OFQOe7IieHUgs+alqWG4kyV42xjEuvJq+3BIhN3MKfxlQIPfDHRq/h6QAyvYMi/ewAHa3RMXyVOqUY2hGQpt7ooc2yfG43/FjtG6p9DBcV+EhHGlCooLrzqcFXkA3IFfCXcFtZb+0bxxbV23LUgVueFmAvtXPETeqZNeYmFGBmnDx1XoFCpMTKuQoA3tzZ/hniJhVCpWYwrHb9olVfsxN+XBTAqCoRE0afp4rH215y2cRcmheDprVmQiIT47YFSVHdat03Z8JgSfzhYhoeWJyLZCrl6ppKIhJArVQ5V9EKsI6+uG7mJ1tsRiPD3Qlu/87RZJfopVGp4CBirppf5SUQYVaigcIBmCw4VFAPATdmROM4sdrgpbSfLO78JVI35IK8RLMua1UVAKybIGxkRfjheprv4MDPSH2Vt1FvUFRwr7eA1l35dVrjemymXo71hvl5UxzJCMABYSdBEjrGn/qmAuqxIk+JXmzNxtrobLx6rRM/wOO+nfb6mG389WonH1qYiOtD6vYi5Sgv3RXUn7U65MpWahZplLU7VwpAMeGvTREGrtrRotAd4axMy/EZRSzudTu9SfS9yEqyfUrUmwzGGCDlcUJwdHYDCbpHDTWmraB9EuoGRu1osy2JfcQuWpUotzg9clxmOgoZe9I1MvyhnRwXgWmu/RccnjqGoqQ+zY/jbOp8dE4CiZjeZcDepJSG79td4e9b7kAemwnuwTtMuyswuLCKhAPctScB9SxLw7vkG/PvrGgyNWb4zMzymxAtHK9AxIMczWzMRauG4VL4tS5HirJNNGCWmudLSj9kxU/PvzVC6d6KwNSQFePSC5r8A0F2NjUweDrtTKpeLyqvrxsIk6wfF8+KCUNBg/w44DlNoN1lqmC+qZUOwbadO/eQKFcQe3Po5ljT3o6lnFI/pKJAz1drMMPzzZDW+qpBh+9wb802lfp6QDY5Z/BrEvmplQ0gI8eF1a4phGIT7SdDeL0dEgIS34zokbcCbtR1vFA4id0EwJOsPay7YPFS+B/uI8eN1qWjtG8WrX9dALBRgy6xIk1sTdQ7KsedyCwZGFbh/SQLC/B3z+yL180S3FVbGieO4UNuN7yyMs/xA2p+vsUFg7o6JHOPCdwFPP3jl7sTA4TKwLGvXzi7EMuNKNTw9zEsDNQXDMAj3l6Ctf9Qmkzz1ccig+OYUEYZevRkYvb5CDExMabPDavGZKu79HA+UtEIkZLAxy/Lt8NkxgQj19cTxso5pQTFxDf8tbsXDy5N4P+62OVHYc7kZj6xI5v3YDid3J74s74DUzxOztCtgPLeCigr0wk83pGNQrsChK234+FITpL6eyI4OQFaUP/ynjH1XqtSo6hzClZZ+1MqGEeIjxu3zYxxuZVgftZq1eRcOYhtDcuW096vZdLUDXf7kN/87NzYIhU19mBcXxM/rEZtq7B5BnA074myfG41PLjXh+yvtd91yyKBYUrkfktE6qEPTIbj/oOaLu7dOTMmxce/DwiZu/RxZlsXBkjYsT5XyUjQlEDBYkyHF4avtUKjU03LARNdHs9riLo7wT65QQalirdKXNtRXs+LnDqs0NbIhlDT34yfr0qz+Wn4SEe5coFll6x0eR2nbAD7Kb8SQXPnNvzMLQMAAqWF+WJQYgjvmxzjV9yA9wg8VHYPu09bPjcgVKnh62C5rclW6FK9+XUtBsZM6WdFps3HzgGZnrm9EYdebcocMipG7E7KhMZwWLcVt2lXh+w7YJSBWmtDP8XJjH1r75XjKggEMU63NDMcnl5pxsa5nWnPr9Ag/VLYPYSaP+ajEdk6UdWI9x77X5pgRFYDydtcObgblCrxzrh7PbM2y+WsH+YixNCUUSx2k6TxfFiaG4EBJq0u/b9xVQUMv5sfbLkDVdjRxavm7NAX+2lhkSGaXWMQeOgbkCLdxqteylFCcrenC8lT71I85XKGdlnTNj1A24DnRHshOU9ou1vdiAcfKy4MlbRB7CHgNdJanhkLsIcAJHVWZ2dFUbOfMrrT0Y5YVb2hWp0udYnS6Jf55sho/XpcGD0sr6ck3pH6e6B6ivGJXVNDQi3k2DIoBIDHUx3m7UGin9jlQe1hbGR5TwtvMlrKWWJIcgrPV9uuX7tBXknlxmnwkezpf282pyblazeLQlTasSpPCj698LQDeYg8sSQ7BibKOaf1DowIkaO0b5e21iO30jYwjwEtk9d6Pwzx0THBUh6+0YVFiiKYXMeEVw2g+04hrGVeqIRHZNtBZkxHmvDfnWdsdrj2sreTX8TDgxQwCAYMgbxG6h+zTSMB+QTGHCXXrssJxrNR+/VZVahYqtRpiDjlYlxp60T4gx5ZZkbyfx5qMMNR3j6CxZ+SGrzMMA7psOadDV9qxeab1c7USQnxQ1zVs9dexNdngGK629mM1x97hxDTavGLiOvpHFfDjOICKT6G+nlbp9W0TkydmOkh7WFspauqbKFy2sVvmRGNfUatdXts+QbFayWkLQiQUwE/iYbcfqLy6bixM5HandLCkFRKRAOsy+c8R1Z7DpfrpPfwEDAOlA0yBIRxMGl3e2DOCeM8Rq2/BrcsKx3E73lhaA8uy+NdXNXh0laM0bXQ9i5JCcKGWRj67krzabiw0c+WvuXcEB0pa0T+qMOv5Qd5i5w2M3ZSS44KgNUQESNAxKLfLdE37/I07yzhvQdw2NwZ7Ljfb6MRudK66G0sWouXAAAAgAElEQVSSOaZOXG3Hmoww+Fihk0BqmC/8JR641NAz7c9SwnxRI3O9lUCXMyk3rbWlEclewzbJTQv2EaNHx/AXZ/Z5YQs2zAi3StcOouHUq3tEp6utA5gZzb2GoXtoDH87XoUtfz+NZX8+iR99UIhNL53C6SqZya+9JtNJUyi0OcTaFWLtirE2x9hFDcgVZqWBnq6SYc1fv8Kml07h/rfy8cs9Jcivmx63cJETH4xLdhjmYb+VYo5bEBEBEsiGxmye36ZQqcEw4FTAU9zcB9ngGDbyOKp3MoGAwfz4IJ0rxTOi/KnYzhlMyk0LensFvpX3bZvlpkW6UO5519AY6ruG7ZLr5m4YUF6xK1Gp1RBybHM1plThgbcv4qUTlfASCfGrzRl48/4ceIuF2PFGPp7ZexUj49zrFZKlvs5ZbDdpYiZ+cEHzS5tjXLrX3mdnNfm1PchNNG2K3b6iFjz49kUwAGKDvdE9NI6DJW3Y8UaeWYHx6nSpXcY+O8VSy/IUKU5Xd2Flmu1yeM5Ud3FutXSirBNCAYNVadbLb8xJCMbJigr0jYwj0HuisCg+xAd77ZR7Q0xwPTeNfWURvEa6AAVslpu2ISsCR661474lCVZ9HVt4/XQdfrDaDQaSWOhKcz/+9XU1BAyDIG8xgnzE2DQjAllR3NuspUf4u16/Yjdtr9UxIEeYH/fWWr/dX4qS5n68tmM+Nkxa7FmSHIrnjlTgjTN1qOsaxrsP5XIuFhZ7CCBXqGxe6GeRSRMzYef2sLZU0tKPx9ZwT097/XQtnj1YhkVJwXjt3pxvhsN0D43hjlfP46HdF/HxI4tN+vzxEArgLRaif1SBAC/+mhcYY5+VYoGHSVsQS5JDcMaMLRtLXKzr4dyK7XhZB3Lig3gZ2KFPzvU2OlNngwsFjF3yboh5VHZYeYsIkKBjQG7z1+XbhdpundPjjJqUyw2AU5Gvs1Kq1Pjnl1W49ZWzuFDbg2utA9hf0op/fFmFO/59Dhfrua/YLEwKdq28Yjdur3WupgtLU7jtrnxe2Iz38xrxvZVJNwTEgKbv8DNbs/B/N2fhTHUX9pe0cT6HpSmhOFfTZdJ5O4TcnTcuXNipPawtqdTTh4Xp88aZOjx7sAybZ0bg7Qdyb/h8DvH1xLsPLYSP2AP3vZWPhm7TUj23zY7G/mLbLvrZJygOyzRpC0IgYJAY6ovqTttsv8gVKoiEAk5bTc29IyhvH7RKgd1ks2MDIRIyuKgjhcJ5ZmW5sesXYA95N1g75KYF+4jt1uKGD+NKNb642o6bTe3u4kaBUHPvCO587QKeP1qJTdkR+PKnK3HyqVUo+n8bcOGXaxHuL8F9b+ZzDoxdLq/Yjdtr1cqGkSz1Nfq4ivZB/HLPFSxMDMbPNugfQrVjcQJmRgfg2QOlGJRzK76bFxeEyw32bbFKjOsf4b4y2z00hhePVWJ1uhT/uHuezl2A6EAvvPtQLhQqNe57Mx9yBfdhLnEh3mia0nXL2uy3UnzfAWDz85zvuG6dG429hS1WPjGNrytlWJXObUtbWzywNtP01AmlSg3ZILdARSISIjs6AAU6iu2CfcTocuKAxy1cz03r8koEY4fctLWZ4ThR5oSFLte9c74eOxbHm97X2U0CoUG5Ave+mY/K9kG8dOcc/OPuuTekWYX7S/DhI4sQEaAJjLnm+LnUDbcbt9diWRj92VGrWTz+YSH8JCL84ztzDdbTCAUMfrc9G7KhMbx0vIrTOQgFDFiwtLPp4PLqupHLsevWP76sxsi4Ev+7JcvgImJquB/+cfdc1HeP4P28RpPOZ1ZMIIpsOK/Cfn2KTdyC8BIL4e0ptEnwV9TUhzmx3PrznSjrRGKoD5I43IVPdqaqC88eLMObZ+vwVQW3YCUnPgjFzf0YmzI2MyPSHxXt1FPUoeXuRG3u/6Fi04ea9772Am3CjaElEkN90NDjnF1KWvpGMaZUc1rpmsYNAiGWZfHzT0vQ0D2C1+7Nwfa50ToDoHB/CT7auQiRARI8+PZFTqvAUn/XSL1xZ9pBQcacKO9ERccgntmaxSn/eE5sIO5aEIe3z9WjvH2A07kkhVK3JEd3taUf2Rxyfxu7R/B+XgPuXBCLlDDjn83LU6VYkhyCf31VbVKR5oYZ4ThyrZ3z4y3l0BPtpvp2Tiw+udRk1dfoH1XARyzktCI1PKbE+ZpurDVhgEBb/yiePVCK4XElfn1zFn6xKQPdQ+PYfa7e6B10TkIwxpVqXG25sdtERoQfytq4fSgR+/lMeBMWZk/akuR4Y9jaN4o9l5txsb4H3UNjZq+0eHoITdq6chS7z9XjwaWJ9j4Nh7XrdC0OX23HLzalY7GRFpJh/hK8umM+hseV2HW61uix58YGorDRRba83bS91uVGbqOdd52uRXSgFzZnc++i9PON6fCXeOCZvVc5fS4tTws1q6UbsR0Vy3LquvXXYxUQChj8ZF0a52P/dEMauobGsftcA+fniIQC+EtsN+HOqYLiUF9PDI8prXph31vYgu1zozk99nRVF8ZVaqzV5hNzKOh543QdntqYjo0zIr4JvL81PwYzovzxwrFKg683//oH29S84kBvMQbMbKpObEOuUEEoEHD6sJnsYn0Ptvz9NJ78pBh3/Ps85j97HAt+fxxfV5p+YVmaEuJ0hS4XarsxMzoAXmIzK9ZdPBA6V9OFPx0ux+aZEdi5PInTc1LC/HDzrCjsPldvdLU4PcKP8yqgw3PT9lpXWwaQHW145a+kuQ/5dT14YGmCSZ9RQT5iPLkhHRfre3XWu0wV5iehVD8H1js8jkAvsdHHXW3px76iVjy0LBHh/ty7msyPD8bqdClePVXDORcdAL41Lxqf2yh91i5BcbsF23G3zo3BnsvW+cdhWRatfaOICfLm9PgTZR3wl3ggJyGIU0HP+Zpu5CQE6UxGz0kIRlq4Hwob9X+whPp6IinUR2e/YuLYTpZ3Yo2JI4n3Frbgnl15CPIW47NHF+OtBxbgma1ZCPHxxA/fv2xysDInNsipVv1UahZfXG3HVktGp7twINQ/osDjHxYhMdQHf7l9tkn51o+vTcGoQmV0tVgkFNilYwrvtK3YNj+vSZ0BNN9/G6Yw2YtCpYanh+Gbyl2n6+Dn6YE7F8SafPxvzYuGn6cHPsrnlisqFgqnpQASx5BX142FSca7bv35i3IEeovwvZWmt8d8cn06+kYUePNMPefnhPlL0DU0bpPPIrsExbLBMbNHz6aE+aKua8gq/zjna7qxiMMEO0BTlHCyohOr0sM0rUuMFPSo1SyOlrYbHPBxU3YEDl81nDszPz4IBQ0907aqBAIa9+zIipv7MTuG+zSpl09W4ycfF2FefCD2/GDJ9TvsMDy0LBFvP7gAPp5CPPjWRXSacIOpad/nPAMZ/nOpCbfPjzG9uG6y3J0TgZAdcrmt6bmj5egZHsPf7ppr8nQ/U1aLBQzj3IHx5AULbXGldsHCxfvNjivV8BAYvsy39I3i0JU23JUba9YUM2+xB7bNicLBK23oHzG++rcgQfcgKmJ/pa0DmBFl+DpVIxvC6aou7FyeZHp7TAAzYwKwcUY4Xj9diz4Tpq2uyQjDiTLz4kZT2CUo9hIJ8cQnRWjsNq/VxpZZUdhXxP9q8deVMqxM5VZ8U9Tch66h8YmuE0YKeg5eacOWmZEGL/AeQgFig7xQ16W/EGFBQjB6RxTTihUSQ31Qb2IPQGIb/SMK+Ht5cA7uipv68NyRCtw8OwrvPLjwhi4CABAZ4IU37luAvlEFHn7nkklFCzNjAnClxfEnIPaPKNDcO4psE8bS6uWCfUaLm/rwfl4j7luSYPa/EdfV4rRwP1R2OHEhr5t0INHlWms/Zhgpmnr7bB0A4H4L8vbvzo3DmFKNzwubjT52fkKQSf2yie2oWRhtRftZQTMEDHDH/BizX+eJ9WkYHFOa1IliQUKQTcY+2yUojgv2BgPg0fcLzMoPnhMbiPL2QV63YNr75ZD6eULAcQzmibIOzlPsxpQqFDf1IYfDMJBvzY/Bnsv6P1jmJ2jyii9N+VDJjPRHWZsTX7hc2OGrbbgpm1sKAMuy+L/91xDq64k/3jYTYg/dP6LZ0QH4+11zcbWlH7/5bynnc1me6hyFLm+ercMDSxN4O55SpXaZVlAqNYun916F1NcTT67nXuQyVUqYH7ZyWC2eE+fkxXZu0IFEn8uNfQaL7AblCnyU34QtMyMRHehl9utkRwcgO9ofH11sMvpz5ukhhIJ2NR3OoFwBHyM7Tio1iz2XW7AyTYowE3KJp8qI8EduYjA+LWjm/LnMMAwSQ31QY+Vx4XYJisUeArx45xxcax3Ar/ddM+sYdy2IxUf5/HWi2FPYjNvmcb/zOVHWeeMUOwMFPfvOFOGu3DhOx/UWe8BLrL/1XFKoDwK8RChuvnG1LzHUB7XU6sYh1XUPIzHUh9Nj9xW1orCxDz/flG50S3xdVjgeWJqITy83c9518RZ7YNTBO1BUdgwi1FeMEF9Ps49xobYbt/zzDBb+4Tgyn/kCKf97GGv/+jX2XG52+jSj9/MacKWlH09vzTJru3uyx9ekYGRchU8L9H+WRgVI0NY/atHrEPvoGxlHsI/+wqnDV9oxOKbk5Qb07tw4lLcPcuopK/X1NCn1i1hfYWMf5sUZbkV7troL7QNy3D7f9Nzzqb6dE4u6rmGTVn+3z4k2uGjIB7t1n1ibGY4frErGx5eacMqMSvokqS9kg2MYMKGCUR+FSo1BudLgh8dkOqfYGSjoCaw7xKmPn9adBlrPMQyD7Gj/aW3ZNAUxzn2xd0UN3cOI5Vi4OTKuxJ8Ol2NmdABu53iD9siKJAgZBv8+VcP5nKIDbT8liCuWZfFhfiPnm8ipxpVq/OlwOe7edQF9owqsSgvDPQvj8PjaVHiKhHjyk2Kse+Fr7C1sccqVY9ngGJ47UoGlKSGmT/fTITXcDzOjA3Dwiv5aBotyuh2Bi3cg0YfL+/vglTbEBntx7stvyLbZUfASCTktVi1Pk+J0lXN1wnF1xU19mG3kffBpQTMCvERmDSubavPMCPiIhfjkIvfFTS+xEEHeYrT0We8m3a4t2X68LhXxId743YFSs7ZTdiyOx7vnufe70+dASSs2GSiAm0rnFDs9BT11uf8H0eLvmXQ+Ib6ekI+r9OaKZkcHoKJ9EONKCoId3f7iVtw8O4rTY//9VQ3aB+T49c1ZnNN4wv0luCMnBp9eakZ7P7eVl7WZYd+8hx3NkWvtWJcZrileNVFj9whu+9dZ/PvrGty1IBaHHl+OP98+C09vzcKT69Nw8LFleHXHfHiLPfCTj4vwuwNlTlN0qPX8kQrIFSr89pZs3oLVzTMjUdzUh+Ze/TdKvp4eJrVQcigu3IHEkKaeUYM35P0jCpyt7sJmI7UuXPlJRLh5diT2l7RiaMxwnUNSqI/B2hlie2NKtc7OWFr9owocudaObbOjDD6OK2+xB26erSnQNPZ+meyu3Dh8zLHTiTnsGhR7egjx9JYsVHUO4b0Lpge32v54lmzDDI8pUd4+aPQOabLj+qbY6Sjo+Uy4GctSQ00+r82zInFMT4eOmdEBGFeppxW/+ElEnKp/nV1D9zBOVnTi88JmvHmmDnm13fY+JZ2UKjWGx1Wcpkm19o3i1VO12DY7ilPu+WTfX5kMFctyGsYAaH5uHHFKmVyhQkFDL5ammP7zIleosPOdS2jqGcWrO+bjj7fNmpYfJxAw2DgjAgceW4YHlybizbN1ePKTIqfJb7zW2o9PCppw7+IE86b76bF5pmZB4LCB1eKZMQG40uz4BZo6uXAHEkMuNfR8U4Oiy9HSdijVLLbMtHzHQeuu3DiMjKtwsKTV4OMYhgHDOE8nHFc3rlTDQ2j4xuhgSRvGlGrcbkGB3VR35MRiZFyFQyVtnJ/j6+kBb08Pq13D7D68Y11mGJanhuLFY5Wcxo5OtWNxPF49VWv2Vujb5+rxwBLuVbdDY0pc4DjFbkypgoCBWate6eF+ekc3z7xebV4y5SKVEelCjfZ1YFkW715owLHSDgR6iTAnNgi3zo1G78g4fru/FAU2qEw1xakqGVamcSvkeed8AxQqNX6+Kd34g6eIDfbGLbOj8EFeI+efoSBvsc0mBHG1+1w97l2cYNZznz1YioqOQfz97rkG2x4CmuD4ma2Z+NnGdOwtasXDu03r4GEPLMvi2QNlCPQS4fE1qbweOz7EBzOi/HHwiv4L06yYwGl1DE7FBTuQGFPXNYwkA7UMh6+2IybI65vrCR/mxgYiJsgLR64Zb52VGemPMhe+XjmTq639Rt8HnxY0ITXMF7NMaC1qzLy4QCRLfUyeVHx3bhw+MKFzhSl4CYoZhnmTYZhOhmGumvFc/L+tWRgeV+GFYxUmv7a/RISbsiPwsQl5KVr1XcPw9BAgIoB7FeWZKtmNU+wM+KpChlUmDmzQYhgGHnp6D8cFe8Nf4jGttVZGhD/K9QTSzq5/VIFnD5YhO8ofDy9Pwty4ICSG+iDIR4xN2ZF4eksmWvtG8cdDZQ4zyjivrgcLE42v+o4pVfjkUhPWZ4VzHhwz1Q9WJ0OuVOHNM3WcHr8+K1zvToQ9NPeOQM1qAnxTHb7ShvcuNOJ7K5I434QwDIMfrk7Bn26bidNVMvzog0KH7sV7rLQD52u78ZN1aRPFvTzaPDMSRU19enP1fD09HP7GgdyIZfXng/ePKnC6SsZb6oQWwzBYnxWOM9VdRt8vi5NCcL7GMXf53E1Bfe83E3N1qZUN4XJjn+V946dgLr6Oe2f54FJDr6arhI4pwLoEeIkg9hBYZToiXyvFbwPYZO6TU8P9sGNRPD7Ia0RZm+l3jjkJwRiUK01eJX3nfAN2LI436TknyjonptgZUdjYh7kWFDDMTwjWWZnJMAxmxgRMK7aT+nm65AjNrqExPH+kAo+tScHcON3/7gIBg5tnR+GBpYn4w6Eyu+c/ygbHEOrjyekD5PCVdvQMj2PHogSzXy8lzA+bZkRg9/l6TsFLQqgPGh2o2O6d8w24f0mCaU/K34XWlkb84rMSzI4NxE+XBnP6QJ3srtw4/OaWbHxZ3onfHeDe2s6WxpVq/OFQGZKlPvjOQvMKEI3RbqEfNrBaDHAr3iL21z+qgJ9Ef/eaE2UdUKhY3JTNvZaGq/WZ4RhXqnGq0nAhXZCPGH1ukO7nDPpHFdP64U+mHSp2y5xo/l70+lCde8p/iDDBAA6cL542BdiQ71hptZiXoJhl2VMALOrG/cS6NPh6euDFY5VmPf/BZYn4MK+R82rGyfJOLE0JMTr+crJpU+wMGJBrPpQsuatamBiMC3ryZd2l2I5lWbxysga/uCnD4A+tVkSABD9Zl4Y/f1FuVjoOX/YVteCWOdwK7N690ICkUB8s4ThNUZ/7liRgUK7kvALsJxGhf9T+F6VzNV2YExsIL7EJxRvXP1BVb21BENuPl7fFQvzeNs4fqJPtWBSPh5cl4u1z9XjrLLeVdlt653w96rtH8PSWLLNSsbhICPVBVqQ/DhkIiiMCJGh3wFx0Ml1hY6/B/sSHrrQhKkDCS9eJqRYkBsNf4oHjHKaPiYQCGvlsZ2o1C2NhyvGyDsyKCTBpV92o60N1PLorcFzyP7j38l0mDdUJ8hHDx9MD1Z387o7bPadYK8BbhAeXJeJoaQdKW01fLRYKGPxgdQr+erTSaOHM5cZeXG7sxRoTUxumTbEz4Isr7dhk4V24RCTUG/TqK7Zj4MTFC/m7bmyRNCRD4WfP4aaZESaNsQ32EeMXmzLw4rFK9NohMGZZFrLBMU7Nza+19qOgoRf3LIrn3HFCn9yEYEQHemFvIbdpj+sybTM20xClSo2j1zpMX7HK2o4h/xTEKhtxxPMXiPlwtUVTyn65ORMbssLxuwOlZo+gtwbZ4Bj+drwKK9KkWJVu3UETW2ZF4nJjH1r1pFDMjHbiYjs3c6VZf47ooFyBU5VduInn1AktkVCA1RmaDjfGUpLmxwfhcoMTD4ZxAbVdQwYLd2WDYyhq6sPaDOMpoyaZNFTHX92HIAxAIQkxaajOvYvj8d6FRl5jHpsFxQzDPMIwzCWGYS7JZLp7Qz6wJBF+nh7458kqs14j3F+C7y6Kx+8OlOrtw3rkWjsK6nvx5Po0kz8QTJliV9c9zEuFeEKID+p1tK7RV2wXF+JY2+KcXV/5+6Z36JAM429uxryrv8eCzs9MPpyfRISfbUrHi8crbb6afqG2Bwu4dJDI34XPzxRDIhJo+hJzzKfSRyBgsG1OFE5VdXFKo0kN90N1p3WnAxnz0cUm3JUba/LPoto7FA8z/4deBEAy3mPxlDKhgMFLd81BdnQAfvxRIe+rD+b6yxflkCtV+PXNWVbvF6y9MdFulU6VHqG/+Jc4lnGV/vZaJ8o6Ma5SYzOPXSemWp8Vjp7hcVxuNFz8nJMQhIIGGvlsT/l1vQbTQU+Wd4JlgXVZlvcmNsbUa7VIKMBt86JNLtQzxGZBMcuyr7Esm8OybI5UqvuiFeAtwv1LE3D4avu0FVCuEkN98PSWLHxe2IL9xa2QK1ToH1Ggc0CO3efqNa2bViSZdYE5Xjplip0enYNySC2YxjXZqgypzp6y+ortMp21A8X1rRTIyoFXFoF9ZRHEPZVQh5q38gdoijB3Lk/Ci8crbZoLeaKsw/guxPWbgLuuPYrvzPBGgLrPpHwqfW6dGw2VmsWBYsMtkbR8PD1M6hHJp57hcXQOjiEjwt/k5x6+2o6qjkF4m5JyYYS32AOv7pgPL7EQj7xTwMtgIEtcbuzFfwqa8dCyJF5bsOmTJPVFRoQfjl7THRTTeF7noFCpITBwfTtyrR3h/p4W1bsYszJNCpGQMZrKJREJIVfQe8qeWvtGDRZ4Hy/rQFSABFmRpn9OGzRlqM6AIBA+yl6wJg7VmRUTiI6BMd5atDlM+oTWg0sT4S0S4p9fVpt9DLGHAI+vTUWQtxhvna3HZ5ebcaysA3NiA81OFK/vGkZFx6DRdk8A8HWFDKvN7DoxVZifROeqn75iu5QwX1R12Hf1zyyTtlIw0gVmpAtKSQgE95u38qcVG+yNdZlhePtcPX/nasCV5n7MiPY3ngqRtR29PklIYZrxq/oHgFcWWbT9r5UW7oesSH98XsQtKF6VLsVJOw3y2HW6Fg8t5d4OUUulZvHmkTx85v0HeI738DqlLDLACy9/Zx4ae0bw5MdFdktFUqlZ/HrfNYT7e+KxNSk2e90VaVIUNvbp7eDCgortHF1Z2wAy9QQwSpUaZ6q7sCotzOJ0LUP8JCIsSgrhlIrkK/Gw+w0o0U2uUOF0VRfWZobzv1M1ZajOyTX7UKmOBmPGUJ2HlyfidY59+o3hqyXbhwDOA0hnGKaZYZiHzD1WkI8YOxYnYH9Jq8Vbu8tSQ/HoqmQ8uCwR9yyMN2lAx1RHSzWrJ+uzjOfVNHSPICHEvNZauuj70NBVbOdKqzkePH1oz48PRrCPWO8KGJ8OlLTi5lkcCux8pfih6LfoYwLgIe+2ePt/slvnRqO4qQ+1MuM/P1mR/mZ1fLHU+ZpuZEb6m9VebF9RC7L6TiJB3WSVKWULk0LwzNYsHC/rxN+/NC+Vy1IfX2zClZZ+/Gpz5rQhJNa0OCkE4yq13p7fYf4SdA66XocbV1LQoL+9VnFzPwblSqzg2LrQEhuywlHbNWz0Or44KQQXqDWbXbT3yxFuoHjufE03RhUqXsY6TzNlqM6KuVn4rvIZHEt4yuQe4j6eHliXGc5LYMxX94m7WZaNZFlWxLJsDMuyb1hyvIeXJ8LTQ4BXTpq/Wsy3o9c6MCPK32gfVdX1Sk4+76pWpEpxWkd7G33Fdk5p0laKXBwMpSSEl5U/rVvmRKNGNjxtZZ1PVR2DiA/xgQeHDgHVnYOoaB+Epwf/mzXb5kSBYYC9HFaLGYaBl0iIYRumUIwpVTha2o6bZ5me06hQqfHS8SpcDr8d6pues9qUsnsXx+Nb82Lw0vEqfKEnx9ZaZINjeO5IOXITg7GN44hwvixIDIZQwOBcje52WlRs5/i6h8Yh9dOdvneqUgYBAyxNsazTDRfaXv7GulBkR0/f8SS2cbG+BwsM5BMfL+uAt1iIRUlWer9MGqoT5CNGWlISfi9bZtZu1MKkEKSF++FtCzsIOVz6BACE+nriO7nx2FfciuZe+xeNyQbHUNDYiw1ZxlMnSpr7MCuG31ytGVG6V/P0Fdt5iT1sGuTw4vpWChuagX/PeA8eP8rjbeVP63srkrC3sMVq4yH3FLbgtnnc0nOO5V/Fh+Jn4aXo5XX7H9AUnC5NDsXewhZOHy6bZ0XigJGxrHx6+2w97l+SYNaN44GSVjT2jOCJdWkQLHzEalPKGIbB72/NxuzYQDzxcRGutdrmos2yLP7nsxIMj6vw++3ZVi+um8rX0wMzowP0DlXIiHDSmgWORsaVDjP8xxzGft5PVckwKyaQU3tLS0UFeiE72t9oCoVQwMBZGyY5u6qOQaSF+en8M5ZlcaKsEytSpXqLNvl208wI1HePmD2EbEWaFPEhPnj3fL3Z5+CQQTEAPLQ8EQyAN8/U2/tUcLysAywLbJhhPHXiXE03FlvYb3YqfRdG/ZPt/FDhbKvH17dSDs57Detysnlf+QM03Rme2piOf35ZjdFxfi98TT0jCPPz5PThwbIs5MV7kCZoscr2PwBsnxuNxp4Ro9XfAJAs9UWtjg4n1lDXNQyhgEF8iP7xs4bsPteAJKmPye0UzSERCbFrx3wEeouwc/cldA5av0fvxxebcKK8E/+zKQOp4bovVta2JDkEJc39Om+sDbWJdGaDcgV2narFKydr8K+vavDnL8pxrLTD6QLklr5RRAd56fyz/hEFipv6bJI6obUyTYrCpj6jw5TC/T3R3k89sG1NzUJvbvm11gG0D8itkzqhx4asCAgY40OEDFmdEYbIAC+zW7I6bFAcHeiFbbOj8NHFRkWcp5MAACAASURBVPSN2G8IAwAcvdaO2GAvZEQYv0iNjCtN6qnLVaiveNpFWV+xXUakH8rbnCwoBqDKeRhFPSJka/tr8rjypyURCfHYmhT85Ug5r2N938trwLdzYjk9tqipD38bXIXL2f9rte3/jTPCIfYQ4NAVblv/WZH+Vl8NVatZ7D5Xj3sXJ5j1/OKmPhQ19eFeHno6cxXmL8Gue3PQO6LA994tsGqQ1NA9jN8eKMXSlBDTp/vxaHFyCJRqFhfrnbxVlo6+57o6u3yU34jXTtVi86xIPLUxHU+sT8PPNqQj2EeM3x0odaoiMEP5xGdruqBmgZVpoTY7n6UpoVCpWeTXGX4vLU0JxdlqwxPwCL8G5Ar4Gph6eKy0AwwD3poGcCH180RuYjAOWZiyti4rHN9dFI93LzTgHyeqTJr067BBMQDsXJGEkXEV3rvQYLdzGJQrcLa6GxuzIoxuZfYOjyPQyzrbUjkJwSion77qNyMqABUdg1BOKq6L8JegvV93A35HdvBKGzabkWdqqjB/Ce5ZGI+/Hq3gpZL+ZHkn5sYGci6I2lfUCrGHAClbn7Da9r+fRISlySE4WtrO6e+4cUaE1XNn389vxO3zYyA2M4969/l6+IiF+Nb8GH5PzIjs6AC8eOdsFDb24ScfFd3ws8YXpUqNJz4uglDA4LnbZ9ss6NclJz4YIiGD83qmaUr9PNHp6JPtdPQ919Xy8Oi1dgR6i/DTDemIDpxYYRUIGMyPD8LPNqbj+SMVTrNiXN05hBQ97ftOVcrgJ/HAbJ7T+wyZFxcETw8BzhgJeBNDfVBno90qonHZwA0UAHxdKcOc2ECE8tRelqvNMyNR3Tlkca94qZ8nHl+biu8uisc5Ewo5HToozoz0x8o0Kd6+3l/YHr6ulGFcpcYGDq3YzlR3YVmqde7CMyP9UaYjzyYz0g/jSvUN29+2zkPky5XmPsyL0/9DyqeUMF9syo7AyxYWc/YMj+NCXTc2ZXML5pUqNQ6UtGFdZhj8JaZ3XjDFhhkRaOoZ5ZSfJREJIWAYzmPSTVUrG8LwmHJiF8BEXUNjOFDchm/Nj4Gflf/ddNmUHYn/tzULX1xrxxOfFPO6y8CyLP54uByXG/vw7PZsRAXq3v62FS+xEHNiA/V2BMiODpiWsuVwpvQ919XysLF7BNdaBwz+7AZ6i/HoqmQ8d6SC1++5tbB6tsNZlsWpShmWJodyKgSeqn9UgU8Lmk1eNZeIhMhNDMa5asNBifaaRe3+bKekuR+zYnR/Hg/IFShp7sOyFNvtKmhpO3wd5WmyaJCP2KSCZYcOigHgeyuT0DU0jj2XuY2u5dvRax0I8REbvKPSqmgf5JRiYQ6hgNH5gZEVqXlT6xqN7UwfMFdb+jHThisYgKbp97z4ILxlZrUqy7J45WQ1frCKex/ZczXd6Boaw7bZ5vXLNsXazDAwDIw20Ne6eXYkDhSbn8ulj+p62sRDy0zvSaz18cUmjKvUZqde8OHBZYn4xaYM7C9uxc8/LeGth/G/vq7BG2fqcP+SBJt3m9BncVIIrrT06wyCMu3Uxs8kU/qeT215OKZUYdfpWvxwtfGf3cgAL9ydG4u/Ha+0wYmbb1Cu0LtbVSMbRmu/3OR84rb+UfzteBXePluPjAg/vHKyBnsuN5t0bVmaEoqKjkGjOfmp4b52n7DpTsaVanh66K6BuVDTDTWr+d6Z6mBJG146XokXj1XiuSPlyNOz46RPZIAXZkYHcL5u8c3hg+LFSSGYFROAXadrbX6nPqZU4WR5J9ZlhkNoZDtT+yFhzVVaL/H01llJUh+IhYJpF6moQC+0OlHhwpFr7djAoQc035YkhyIhxAcvn6w2Ocj5b3ErVmeEIcCL+8rlvqJW+Ek8sCrd+sUuYX4SzIsL+qbHtjEpYX6o6eL/ovTO+XrcvTAOIjNWqADN6vp7FxqwLCUUKWGTtoY55ozy6dFVyXhiXRo+u9yMX3xWYnHR2Uf5jfjLFxXYNjsK/2+r9Uc5c7UoOQRqFsivnZ4LKhEJMebkxXYvf1mNR1YkcU7lSQnzQ1aUP845cN5rYWMf5sbpXlg4Van5OVluwk7mxfoefJjXiPuXJuDH61KRHR2A/7kpA/Eh3vjN/lLOKTTa1UZjq8VLkimv2FbGlCqD7/1zNd2QiAR630/6fHyxEWIPAR5bk4on1qfhqQ3pKGjsNTkw3pAVjqKmPrukaTl8UMwwDL63Ihl1XcM4xvHizpeT5TIMjimxaabx1Iny9kGkW2mVWGtubBAKG/tu+JpIKEBahC9KpwTFGZH+KHf01ZzrRsaVEDCMzdq+TLU6IwxrM8Pwu4Pci2pOV8nQ1i836U5arlDhyLV2bJoRYbO/64ascFxtGUBLH7cc88wIf527DuYqae4DA5g1ylnrWGkH2vrluG9y8RnHnFFreHxtCh5fk4L/FDTj7l0XzK6aP1DSil99fgUr0qR4/g775hFPNS8uCGIPgd68Ysc5Uz2mjJCd3PKwuLwKyWG+RnvOT7VxRgSOlnY4bBpFcVMf5ugZUHWqSoakUB/Of+cBuQJfXG3HE+vTpt30z48Pxs83pePNs/WcjpUV6Y9Ab5HRvGKpnye6huxbVO8urrb0f9PSVZcz1V1YkBCsdyVZlyPX2iESCrA+a2IRkWEYPLoyGZcaeo0WW062fkY4WBY4Xmb7aasOHxQDwKbsCMQFe+NfX9faNCVgb2ELQn09sZxD4HOmqsvq+TdzYgNR1DS92E4byEz+t0kL9zW715+tHSxpwxYbFNgZkhHhj8fXpOKvRyoMbg2r1Cz+9VUNuobG8P2VySa9xteVMgyNKbFtju22yLW58Mc4TvO7aWYEPrvczMtr9wyPY29h643BrBnez2tEdKDXjW3YOOSMWgvDMHhyQzr+cfdclLUNYOs/TusddqHLyLgSz+y9ih99UIg5sYH493fnmV18aC0SkRDz44L09isO8fWEzJEn200ZITu55WHbuQ+xlcvUySkYhsEdOTH4z6UmK5yw5caUap0322NKFfJqe0xaJX75ZDV+uDpF786Ft9gDmZF+KGgwHugIBAyWJIfgbHWX0eu3QMBYpZCV3KigoRfz9KSEdgzIUd05ZFI8c6m+B009I7ht3vQiaIZh8INVybhY38Pp/QIA6eF+iA32svlCKOAkQbFQwGDniiQUN/WZdLdhif4RBb4s78S22VGcChP6RscR5GPdhuheYiHkiukfGFlR/ugeHr/hIuUt9nCqiuk0O/VknSzIR4xntmahpLkPvz9YisNX2qBQqcGyLLqGxlDU1Iff7r+GtZlhuHWu6R0QDpa0IchbhMXWmg6kQ2KoD1LDfDkXLXh6CLEkOQQnKyy7Q1epWfz9RBV+sj7VopSApp4RnKnuwrdzYm9MYTKSM2oLN8+Owr4fLkWAlwjffT0PP/6o0Ghbu6KmPmz9+xm8e6EBDy1LxAc7F8FbbLsxzqZYnByCsvYBnS0xHX4K2ZQRstr3S9vS32Fo9oNG0+H0mREVgKbeEYdr0zamVMFDqPvvVNzUj1GFCks4BjkHSlqxOCkEwUauZ9tmR2F/cRuntLOlKaFo65cb7TAxJzYAxTQx0eoG5Uq9aX/aG3yuu6CdA3KcKO80WDOiDYwPlrTrfr9MSYVjhrvwy5AzOFvTjSEbDyJziqAYAO6YH4MQHzFePWX5bGsuDl5pw7hKjVvnGi+IGhpT2uzCJhQwUEy5k86M1GxNT02hcAZVHYNIDtPdQsgePIQC3LkgDr/anInwAAn+caIKLx2vwqErbZANjuEXN2WYFcDLFSqcKOvApuxIs6q/LbFhRjjy6no49/temxmOM1VdGFOaf1P12qla3J0bZ3GHjf8UNINhgNtzbNuGjavUcD/s+9EyPLw8CcdLO7Dl72ew4408vHu+HkevtX9zI/+XL8qx6aVT2P7yWcgVKnzw8EI8szXLbilDXOTEB4FlNYH8VFmR/o7/eTNphCwAwFeKtxXrcYuFOzX3L0nEWw4wVGqyqy36OwlcqO0GwwCLEo3fjLf1j6KqYwir0o33pmUYBrfNi+a0s6RddTSWM5ybGIK8OtPyT4lpjN3EnKnqRqC3CFmR3FLe3s9rxPdXJhtd/GAYBhtmhE9foNGTCrf5/7N33uFN3Wf7v4+2bdmSLe+9t5nGgDFmz5C9J1lktX3Tpu2vzdt0zzRJd5I3s9mTNhA2BMzGC7Dx3ntb8pK1pfP7QxYY+5yjoy0TPteVP2JsSRjpnOf7PPdz312v4G7y4GU9vLuYM0WxiM/F9vx4HKsfRIMbZAG7LvYgKcQP2VHW3xjnWuTId3KKHR2ZFJHPdEWxiM/1+m7xvqo+bPOwdIIKgiCwKDYQz200m/k/tDweGzLD7D78HG8YwqTOiBty3P933ZgZDqOJxLF69t3fB5bF4aPiTrueb39VH6IDfRzW2BtNJHaWd6EgOfgqD1kAjJpRS8ehQz7pNIcIJsRCHv53awbOPr8O/29zGur7J/Dz3TV44sPzuPnVM7jrjXN442QrJD58PL8lHQe+X8i6a+cs2ocnraaKzWRejBQEgVl7DIB5aqX18mvLTJoHlYiV+dq98GkhxF8IsYiHdi/y1S1vH8HiuCDKPzvXIkdmRAAkvtYPqB8Xd+Lp1exlYfOipWhj8d6KDfJFdKCPVV2xWMiDSju33ldzjeYh5dULy9MgSRJnW4aRnyRjteMwOK6BiM9lvWy+LFGG0jbF1TIaGikcGZyO04ICt7tQzJmiGAAeXBYHHz4Xb7q4W9ylUKG0XYHbFkWzGv1WdY9inpvsxHLjAlE+I8RD4sNHlNQHdTNS7NLD/b3aOklrMMJoIr12fOxM9lX1IchPgGWJ1DcuJjrkk/iirMvukW1OlARhAUIcrmF/cUkI9oNGb7R5iWzvpV4otQbc6ARrsTPNw+gd0+DuJRRJgQya0fqjH+B3e2txsmkYfzpYjy/KuhzqerNF4sPHM6uTUfz8OpT+7zrs/V4B3tmeizcfXIyLv9iAz59cjidXJdnkVuIIY2o9PinpxB8P1OFcqxyvHW/Bros9rPcyxEIeUkP9KTvFAOCd62b0/OdCN26n0Dzaw315sU7T3jsDunG4Rm/Ehc4RLGMh2ZIrtfAT8myeXjy8Ih4fWgnYIggCK5KCcbZFbnVRkcpl6TrOo7hVTvt+aB2etGmB/JPSTtyXF2vT86/LCL26QUMjhSMe3ouFGSk4Vj84azruSuZUURzoJ8DdS2Kwu6IHfS5MbNtdYfZEZuMZSpIkjCRpt0bNVmRiIUYoxuCZkQGonaFnzI6SoNqJTgLO5mTjsFusyTyNWmeRToTbJJ0wmkh8eK4d+6v6sShOig/OtuPPB+txoXP2siUTHA6B9RlhONk0ZNPk4JEV8Xj9eDPrC9Luih5o9SbWcdfW+Ly8C1Jf/mUz96ug0IwO3r4Te6KegyJ7O352QwYeXBaH/92agfkxUvzlcKPb7LS4HAKhASJkR0mwLiMMG7PCXR7UMpMuhQqvHG7A8iQZnt+SgXvzYvGTzemIDvTBr/fUsvaDXRgrRUXXKGUhHeQnsCk+1ZN0KVQI9Rc6Ta7iI+DCV8Dzir+/yUSC7vZT0TUKrcHEao9h5/lu3GFHWmSovwhKjcHqVCY/WYYJjcGq7j4vIQilcz1i3IsZHNciLEBE+WeWa+SKJOtF8eCEBgIeh9UEYjrmpUs5q8P5xsxwjKn1KHPTLhkwx4piAHisIAEk4LJuMUmS+OpiD/Lig1jZ17TLVYiX+bnktTAx8w2VERGAtuFJqHVXip6wAJFXx7Fe7BzBwhj3JNh5kuMNg1DpjNhmg3SiU67Cb/bUYFFcIJ5enYTkUH98d20KntuQitI2BS51U3fv6FifEQaVzohiG/wifQU8PFaQiN/urcXIJL0e2WQi8UmJWWrhrAjmkUkdjtQM4JYFUfS2QNM0oyRJ4vWycazb/jPkJwVfNeFJC/fH81szUNKmsE165QEfZGcwptbjzZOteOGGTCQEX31tyo0Pwgs3ZOCj4g5Wh50FMVKMqfWUC1I5cyHZbor/Xuhx2mHNwl250fjCC5woGgYmkEojVTrXIgeHAJYkME+otAYjxtR6hPjbF+mblxCEEiuFi6UwL6Hwvp7OghgpKmmmE9dxDJOJBNPw+0yzHFFSH8TJrNc+n5TY3iUGzFODVWkhOGHRCjNI4QqjSAh5HKel27FhzhXFMUG+uG1hFD4u6XRJt7i6ZxwtQ5O4hcWCHWA2RV9lY0qQo1AFc2RGBMBEmi+QcwG1zgghj+tV3qyuYm9VH4LFAuRZuTFZMJpIvHO6FS9sy0RW5NXLMzwuB0+sTMR/L/RcdQCyxvIkGXz4XBy10fcxVuaLH29Kw9+PNqGR4r11pnkYv9lbi9QwMW5e4LyUvl0VPdAZ2Xed/3uhB1uyIxilOP+zLgWfl3WxOyh60AfZEXQGE1453IAfbUyjtXnjcTm4f2ksPi21rhlfOBW7TrlsR7Hf4I0YjCZoDUbatDd7kYmFUOuMNn0OXUFZuwJ58dTXluJWObIiJVYlO3sr++yyqbNQkGw9eCM0QISEYD+ri3R8Lset4/JvE7V947QLdEaTWU+8IllmVTY6NKEFn8uB1Nc+x63ClGCcapp6vzBI4Xyb9qIgORhHagfcZsc754piwHxzI0kS/zzW7PTH/uBcO0R8DuuFqMEJDUJpRhGuYl60BJdm3KSyIs1v9Jk3KQGX4xY9pa0cqx/EugzrG85zHbXOiGN1g9iUxV468VlZJ+5aEkO7EMThEHhqVRJeP87+/S/ic1GQEoyjdbZfXPxFfPx8WyaO1Q/iL0caL//32721UGoN+OWNmciluSnbA0mS+LysCzlREmRGWl907R1Vo3tEbfXQweUQ+PGmNPzzWLN1zaIHfZDthSRJ/PWbRjxekGh1pJkS5g/FpM7q+D85VAw/AZdy2c5XwIPGwwUhG043D6PABo9eW7htUbTHtcVDE1rKe5BGb8TFzlGrewwkSaJhYILVZ40OHpcDDgGr95qlCUEobVNY1RWHS3zQyzJw6DrsKW6VYymNlKamdwzjGgMrPfEX5V12dYktEASB1DAxWoeUtPaJ2PoykLcDG7PC0DOqdpvbzZwsimOCfHHPklh8UdaFTrnKaY/bP6bBrgrzmI2NTkajN0LAdb+lUmqYPxoHrtYERgf6wF/Im5VGlhruj8Z+78uTr+oZu1zIX8scqx+EWs/edWJgXIPhCd2sDvFMwiUipEcEoMgGR4kNGWHoHdPYdXHhThXiz21Ivfzfz7dlYlNWuNOjiat7xlHfP4G7qBbsZkCSJN482YonVyWyemwfARfPrk/BP441MX+jF/gg28r+qn4UJAcjlsXoEwAeLUjAv8+0MX4Pl0NgfoyUdtluLlDSpmBlR2YPCcF+6FKoPJZyR5Ik6M64FzpHoDOasNyKM1JxqwJLWU6xmFidHorjDcz2WUsTgzCuMaC+n/katDo1xOpjXcd2RlV6Wv9pizNIvhU9sclEQqk1OJzLsCkrHIcsy98U9onI2wEAWJseBoKA21wo5mRRDADfXZsMLofA3442Ou0x3z3TBqOJxI6V7G6wZe0K1iNxZ8LncmAwXT1eIggCGRGzx5neqPub0OghFnKdXkx5I19X9iDEX0h7Op/J26da8fhKehP06WzNiUBxm5xR7zudNemhIAjYLKFwN5+Xd0LI47BadN1d0Ysb50fatEAVLBZicWygwwEl3oRGb0RF14hNseMBIj7iZX5W9ZsLYqSo6xunXNKU+gqgYPn+8wQTGj38BK6VaW3MCsNhlomRzqZToUIsze5LsUVPbGWKc7xxEGtY+BJbY2GMlHKiMJ2lCex0xTFBvugZdV7D6zqA3mhi/BycbZYjLczfqq68pM05dY/UV4BRtc7q5DLEX4hFsYHXi2JrhAWIsD0/Hrsu9qB50HEdrcW+6IZ5kazz4c3ekJ5ZFCMw24Q7I8JswTb96xESkUudOuzhm7oBc/zwHF1kYsuYWo+i+iHcOC+SlTvJ0boBLEuU2aR9fKwgAV+eZ7fsE+IvxPxoKY7Wudf30RY0eiN2V/RiS3a4VR0kSZKo7Ru36zO4MSscp5uG6fWgLHyQvYkPzrXjoeXxNv/c7Yuisbuil/F7FsRIYTCRlAl2OdHed+iezv6qPmxxsTf44rggmx1hnEVpm4J2ia64VYGcKAn8GZxPxlR6SH0ETjk0EAQBfxGP0ToyUuqD6EAfVgEdfC+V/s1VqnrGMJ8m4EWjN6KsXYH8ZOvNm1NNQ1jpJJ/1jPAAVntQGzLDUNM7ju4R1x+U5mxRDABPrUqCD5+LvxxxvFv8cUkHlFoDnixk1yUGAIPJRLvM4mriZH7oUFz9BsmMDMCkzojOaV/3xm5sQ78SqR2fzclFJls4VN0PndHEKkHLZCJxtkWOdRkU9mMMhPqLoJjUs15MWZ8RisruMa91JTlY3Y8JjYGVdKKiaxQLYuz3B3+sIAHv0skH9v/46uWP7fsAcZj5a7W77H5OV9A7qobBRLI+zE+HwyGQFRmAJoYb04JY8++YNtnOi20fW4cnkRTi+sTMrEjPxF63yycRTyGXUeuMuNg1gmVWpBOHavqxMcu2aw4Tm7LCcbCauWu+NIEiwIGCvIQglLV55rDBBp3BhJFJHboUKq+ellgoZejwXugYgdZgupw8SIdaZ4SAx3FaKuv6zDAcYeGfv3HKlvMbN3SLPVLRDU1o7RfRT+suBvkJ8OyyQMhqP8Cxevt/WRq9Ef8+046VKcHIjmLWclroHVUjXOJj/RtdxLxoySxbLkuy3UwJBVU0tKcYVenMHcA5uMhkK7srexAv86WNX53OiaYhrEu3b4S5Jdv6jciCpei2Jd3OnXxe1oXYIF9WGtDDtQPUHsYsiZT6wF/Em61vLH0LqP3KXATf8b75azu3A8oBIPPWy1o3b+H9c+14OD/e7p+njF6dRqi/CFFSH8rRuJ+QB5XOO4MWuhQqxATaflCwhy054ThQ3eeW57JgKSqpGh/nO0agN5JWQzucfWhIDhWjZYh5h2VpYhBGVHo0WfHKzo0LQpkX+hVf6h7Fr76uwZsnW7C7ogclbQp8VtaJvx5p9NpmAwAoNQbaqcGZlmFwOYRVWcTh2n5sygp32msSC3lQ6Y1WD0iJIWIkhfjhiBumnB4pigP9BDhc048/7K+zauNyFRQ2SY+3fg+/5b+H81++ZHfi11cXezA0ocVTq9jHW55sHMKqFM8t3CSGiNEydLV3aGqYP7gcYtYiVWqYP5oGvGPZ7mB1PzZnh8/JRSZbGBzX4GyLHDctiGLVrT/XIre6EEPH/Bgp6xF2erg/oqQ++MYLdcUd8kmca5XjzsXRVse5ikkd/EU8hyN778uLxaclnVdLkSwHNuUA8P4NVx/Ytr7k0PM5m/J2BTIjAhxKhfQX8aHUGhhvTJYQj7nE3kvui5AX8sxhHqMUwUquonFAidQwan/i4lY5uByCUU88MqlDoI3BC2yQ+ggwpqa/Fy+7rCtmllAIeN5lzTam0uOVww2o75vAL7Zl4rtrU/DwigTcsTgaz6xOxqMFCdh5oRv/OtYEgxe9bmDKFIBhqn26WY4FMVJGqQ0A1PVNXG6+OYv50RJc6rZ+/9qQGY7iVgXGVPbVeWzxSFHM4xB4eEUCnt+SjtbhSfZdXoruIme4ARppCr5QLcIf9tXZ/FrGVHr842gTcqIkyLehKOkaUbHe8nYFXA4x6yYm4nORGOxHuWznidEeFR0K1axAgWuRPZf6QJLsUhHbhycRG+TrkNQlO2r25IAKgiCwLiMUp5uHPO6vOpOd57vBIYA7cq0HgPz3QjduW+h4UAiPy8EN8yKxf3qXbw4d2A5U97N6j1ljfjRz0bsgRoqeUTVlJ0ziw2e97OlOxjV6u31U7eG2RVH4z4Uetz3fqaYhrKRpzJxrlWNetARihv2EgzVTDQonk58kw7kW+oI3JsgHERIRilmklMXJfNFOERzjbnpG1XjlSAMeWZGAu5bEUB7aLTHvN8yLxIsH672qML7YOYpFsdS7F2NqPaq6R7HCSv0zMK5BqJ3hLkysTgtltfS8MSsMRhPp8gVpj2qKCYLAg8vi0D6swukmFh1jmpuV6PEDuK1wIT4r68KpJvZLMCRJ4n+/qsLQhBa/uyWbdVGiN5rA9QKtLkEQs6yAzHHPVxfF0YE+bhGoW2NoQguZxcZlji0y2crXFT3IjgpAcqj10eSuih7cyjIsho4t2eE4wFJCsSEzDBq96bIFjzdgNJHYeb4bhakhiLAiSzKZSAwptQiXOMcfPC8hCBWdo9AZvOcmxoaydgVy4wKdsjewOi0ERQwWWJYQj4sUhXNOlATVVqJ73U3b8KTbD98REh8MTmjcZs82otJR2mupdAZUdo1alU50KlSIc0Eaa3aUhDHKmSAILE0IQkmrdV3x6rTQK8lnHmJkUoc3TrTgZzdk0NqZTSch2A/35sV6VWF8vkNBu5Bc0iqHiYRV55o9lb3YNt/5kxcRnwudwWQ1JnxBtBQh/kIcrnWt04tXLNo9WpCAqp4xh/RDP1ifisQQP/z0P1WMo5vpfHm+G/uq+vDcxlTMt2Fhp6JrFAs95DoxneRQMZpn6LIyIgLQO6a5aoxHEAQ846J5NQer+650JhhSbLxtkclW2oYnUdk9hpvnWy90VToDTCQcTtviczmQ+vAxOGFd07YsUYYAEY+1DtkdnGwaQt+YhlWC3Ykm56dI3r0k5kpk7xw5sB2qdp6+T8Tnwmgy0d7EsyIDwOMQlPZtWVES1HjZst2x+kGstVOj7whr0kJxotH10iSVzgARTfx5efsIDCbycqwyFcPKaQ0KJ8PlEDBZKXaXJsowrNSi1UoXOCxAxOqa5ipUOgP+cqQRP96URh83T0FiiBh3L4nFnw81eMzDejqTOiN8BNSv/0zzMHz43MsHXzrkkzqE+rsmqCw3PhDnmtFxBAAAIABJREFUrTi4cDgE1meE4XiDa6ecXlEUA8BTqxJRVD/IvIDHcLMSaRV46Y75GJzQ4N43iyG3ktTUNjyJX31dg2WJQXiykL2WGDD7P7rKDN4W5kdLUDljZG6JcJypK+YQhMdPrT2jGkRbFl+spNjMhCRJnO8YwS93V2PVS0XY8UE5vrrYbbeO3JV8XdELggCrU/Xuil5W7hRsuGNxNL5iMb7lczlYnxmGb+oGvEaz90lJJ2R+Aqxn4b5R3CpnvOHbQ0qYP/rG1FBqDXPiwHaxcwTzYqRO9d/NTwrGORqdp4jPRWqYP6V2XSzkQWUtIdDNyJVaBIudP+q1hqUD6mpKWhW0zhLnWuXgcQhGq8KDTjxQUREpZZ5OWsJC2PyuRDyuR5Y5DUYT/nywAf+zLsWq1paK5FAxbl8UjTdPtrrg1bGnb0yNCIap2pkWOZYkBDFqjjvl9H7YziA/KRjFDJIbC9vmRUClM+K4CyUUXlMUEwSB765NxjunGRKWrNysFscF4q2HctE6rMRdb5yj9edVTOrw7GcXwedy8Ne7F7DykJ2OxkB/6nInsUG+6Jphy3bFgeJqiyXzVrDntFmUH0yGFJvpFDUMYuWfi3D762fxWVkXEoP9UNU9hh98XonFvz2C5/97iTJYwBOYTCT+c6EbSxOCrMoASJJE65DSadvfMrEQo2o9qxjnzVnhGFPr3XIDt0b/mAbH6gdxZ26MVYvDMbXZV9UVVoMPLovHh+c6bD6weYL9VX2sUxLZsiyRWQtq2U2gen95vhd2hTGVHgFWPK5dBUEQiJP5oc3FOtiLDHaExa1yzI+RMk6fekbVdln4sWVlcgijJDIh2A8h/kJWfsWr00JRVO/+Cc07p9vwwLI4q2EWTKSF+yPQl4/zHZ67zh6tG6S1+uwf06B5UIkCK/7E39QNYF2G6yYvIj4XGhae1EsTgiDzE2BvleucXrymKAYAXwEPhakh9GNdFjer1Wmh+ODRpRgY1+KO18/hXIv8csGkN5rw7uk2rH6pCDW943jx9nlWC5eZDCu1CHTj8gYTVIVBiL8QIf7CWbribA8n2+2v6scWO5Y6ytsVeOrD8/AT8PDKnfNR/sJ6/PuRPJz96Vr895l83LMkFp+WduGBt0usTgfcwenmYXQqVLiXRS78hc5RLI5zbiJierg/6vutm6EXpobAh8/FwRr32khR8XlZF4wmEvfmWZdOHKt33cU5XCKCiSTN41qWBzZPUN0zhoyIAJsP89bgcgjwuBzaA2Z2tAQjKj16KKZ5Eh++W50XmHBWQpu93LwgErsuunbhzmA0UTqvKLUGXOoeY5ykDE24voseK/NFF0OnmCDM9l9sdMXZUQFuv3dd6h6Fr5DHaifEGncvicG+S/2Y8NBUs3dUjSgpdZ1ztoVdtPOwUusy6YQFmZ8QQxPM93Ael4PN2eE4VjfosumBVxXFALAqNQQXO0fox+IsblZ5CUH4dMcyqPVG3PtWMeb96jBue+0MNv3tJH6ztxbzY6Q48OxKuzZvTzcNO13P6Ag8DmfWghBV3HO8zBcdcs91iocmtAgNsO1D1TgwgcfeL0ek1Aef7FiK2xdHXx5jcTgEFsUG4re3ZONf9y1EVc8Ybn3t7CyNtbv5pKQTQX4CVu+tE41DWJPu3PfSuowwVol1Ij4Xa9JDcKhmwOqCgysxmkh8XtaJlSnBrJZ+mgaUSHHCjYqOB5fH4aPiTpc9vjPYU9nrFMcJKlalBtPaZOZMebhXUdgnmcMrvENX3DgwgdQw1wd20GHp0E66SFLSpVBdkaHNoLxdAaOJ2Z/4ROMQ1qS5/h7GIQjGa8uyhCD0j2vQpWDOLGCTlOdMNHoj/nO+G/ezaGywgSAIfGdNEv5V1OyUx7OFCY2ecWJwplmOQF/+ZdklFYpJ6oVOZ1OYGoLTzdYnAjfMi4Bab3TZ9MDrimIA2FGYiLcd1OHkREtQ9MPVePPBxXikIB4cgoCfgIe3H8rFB4/m0fo7WqNpcMIpp0dnkRbuj4YZncHMiAA0DU5cVSx7MtmuQz5psx6pd1SN7e+WQsDj4INH8yBj6GxsmxeJz55YBpXOgNteOzPr9+EuBsY1OFI3gDsXR1tdyjAYTTCaTDYtb7BBLORhUmfdDB0wp08NTWhxsctzqVHHGwbRO6bBfSxuQJNaA3wFXJe+lwNEfEh8+LNkSd5C08AEEoL9nJYoNZP50VJU0niGpof7g8chKLt2WVEBjI4D7kJvNIHL4Xg8yfPmBZFW47Pt5WTTEApTqTt751rl4HOZ9cRtw0q3OHNkRgTM2m2ZztKpwr2YhYRiI8vkM2fwxolWPLEqyal6fZlYiGWJMuy75N7J3MlG+iYeSZI40zyM/KRgxr/r0boBtyytJoX4oWXQeuNuaYIMwWIh9lW55vPllCsrQRCbCYJoIAiimSCInzr6eMFiIWKCfB02i5f48rExKxzPb8nAzqfzsed7BVifGWb3BdNkIkGA8PgFdzo50bNlERkR/tAbyVnJQlJfgUfiKA/Z6IdpNJF44sNyKDUGvP9IHivt28LYQHz1zAqI+Fw8+l6Z1TGMK7giA7Be4BW3KqxaJtlLVmQAKzeAtemhEHA5HnWh+KSkEyH+QqxnkUx3vGEIq90wFr8vLxaflHpnt/iriz24dZFj9n1M8LgcGE3Uy5ciPhcpNMt2AVMBIJ6mrF2BJfGedwZKDBGjXT7J6nBqK72jatpOcXGrAgtipLQ7L0YTCS7hnnvY8iTZ5fE8FSmhYgT5CVjtNaSE+VtNwHMGxa1yxMp8aOUGjrAmLRQ1vWOs3bGcQW3fGLIiqbvArcOT6B/XIN+KnrhteBKJbohKJwgCBAGrk0suh8DWnHAcqx90yTTG4aKYIAgugFcBbAGQCeBegiAyHX3c2xdFY0+la04C9lLdO4bsKOemuThKpEQ0a6HQ8iGYqSteEh/okdhMxaTepvHLVxd7UN0zjt/dmo1Mmg80FTFBvnhn+xIoJnXY8UG5W5fvjCYSn5WaZQDxLLow51qHne6gYGFteiirGGd/ER8rkmU4WNPvkpu3NXpH1ShqGMTduTGskulqeukv8M7ER8BFbJCvxyYOdPSMqiETC50+XZhJoK+AVp+fExVAu2znDa2CklYFlnqBMxBgXgoqZRFQYQsavRE8DvVnZUKjR3UPs564omvUJvtRR5D6MifbEQSBvPggVst2ABDkx3dpU0drMOJQTT9uWeC6Q+ejBQl4l8lMwIlY8hToDkAWmVQBgz+xWmeED999pgLZLD3Pb8iJgEZvYnWfsxVndIrzADSTJNlKkqQOwGcAbnb0QTkcAkviA1FsJQrSnZxspE8Q8hRUb/iEYDFEfM6s0VVmBLsOojNpGrBNbqLRG/HK4QbMj5bgxnm26yZzoiX4690LUNk9ih9+Uek2vaxFBnD/UutdYq3BCC5BuGwE7ivgQc0iTx4ANmeHo0uhZhxzuorPyrpAAriHxYKdJabUXVOaOxZHY+f5Lrc8F1u+KOvCXSzS/hwlPykYxTTdu5xoKe2ynb+I79Yu2ExIkoTeaLLqYOIuzEldztU9flM3QGtbWGbREzMkk51rGbY7Tt4eRDwuY3NiaWIQukfUlO+nmWzJjsCBatfJD94/246H8+Ndeo0JFgsRGjB7Ed4VlLUrsCSBfpH7TPMwoqQ+jNLGU01DKEhhXsJzJvlJMpxptl7z5cYHIdRf6BI5ijOuHlEApt89uqe+5jCbssJxuGbAI10sKiZ1RodDFlzBTA9iLodAWpj/rGU7Hpfj9qWqw7UD2JhlfTRu4b2z7egb0+CnWzLs1nRtzg7HTzenY19VH/5xrMmux7CVj6dkAHTWN9M51TiMQhcva7LNk1+fEQYOYbb4cicavRGflHRgTVoo7Sh4Oqeaht16IOVzOciOkuCiFUN5dyFXaiHgcezyS7UVs4MJ9U3bsmxHFRufFRWAGg863LQ40d7QGXA5BEL8hegfc174RFUP/bSyuFUBAZdDG+cLAGq9Eb4C993DcuODUN5O/xmydPVLWDS/YoJ8rS7l2YslStoVCX8zuWdJLL4o73J5XcM0NTGaSJxrkaMgOZjxEFDZPYr50e6ZLADmgzUbSYRZQhGBooZBp8u23HakJgjiCYIgygmCKB8aYnd6JggCBSkyr4ij7ZBPIs6Fvo6OkBjiNysZKDPSvOQw84PnI3CfETpJkpjQGBDA8kY+MqnDq0XNWJse6nA344nCRNy2MAp/P9rkUqNvAGgdUuK4DTKAC50jjDcuZ8A2T14mFqIwNQRfXehxa/LS7ooeDCt1eLwggdX3V3SNYKGbxr4WbpwXib1uXoyh47OyLty9xHpH3RlwGBLJGJftIj0b91xUP4Q1HkixY+KORdFXkhIdZGBcgzB/EW0Rc7ZlGAtjpRDRjLvlHrATXRgrRQXDIm96uD8kPnzWfulhAc49ZADm+9R7Z9uxPT/eqY9LB5dD4Mb5kfjKhbZ9JEnCYKKfmlT3jGFcY2DUExuMJnAIwqkLh2zwF/EwprI+cbpxfgS0BpPTGzrOKIp7AEy/WkdPfe0qSJJ8kyTJXJIkc0NC2Hd81qSZ9ZGe7hZ/w2CA7WmyImdnzWdEBGBUpUf/+NUXkIWxUlzsdGyBkS21feM2aYL/VdSMSa0BP9mc7vBzEwSB39+ag7Qwf3z/8wrGdCVH+es3TRDyuKwuqiqdAUIe1+UXGhGfC63BxOpzc8fiaPSOaRiDG5wJSZJ4+1QbMiMCWB1+LI4C7r44czgElifKGEMI3IFSa4DWYHJrQlu4xIcyXfTKst3sTrLEh48JjeeW7UZU7rGOsgWJLx9cDuGUqOI9lb24kcaKTzGpQ03vOKM+9JQH7ERFfC40evrUTLNMkr2ueGtOBPZecu6u0b6qPmzMDHO5Vn86i+MC0TSodJncqKRNwdh4OcPCn/hi16jLmzdUFKQEX359TCyKDURiiB++KHOuzM0ZRXEZgBSCIBIIghAAuAfA1054XADm4obt4pArkSu1DiXbuJKEYD+0DV9d9F2Oe56hXVoYE+iwqwdbjtYNYj3LoIWeUTU+PNeBOxZHIy3cPru8mfgIuHj9gcUwGkk88/EFaFkk5thKbe849lT24pEV8azeH+Z0Ifd0s7IjJaw05OszwhAg4rlNQ3u8cQhNg0o8vjKBlX6vrE2BvHjnhpywZV1GKI7We1bC9Vlpp9u6xBZWJNGn2zEt23mKkUmdQ13Q7hEVvrrYjRd2VeGNEy1O3Wp/YGkcPi1x7LNFkiSGlTraa8yZ5mGQJBj1n56yE+VxCcY4+aUJQWiXqzAwbv3gEBYgwpBSO8ub317G1HpU9Ywhn+Ew4Soec+HSnTWnnjPNw0gP92e8ZxW3yN2qP7eQSZGzQAVBELg7NwblHSNOzSdwuCgmSdIA4LsADgGoA/AFSZI1jj7udAqSg3GmWe6xi7Anxk62wOUQs3436TRFsY+AefHBWZAkCZWOvX7t4+IOGEwmfG9tilNfR0KwH16+az4udY/h13tqnfrYAPCXIw3wF/HwZGESq++v7Rt3i4MCABSmBuNEo3WpkojPxU0LInGwpt8tBvlvn2pFWIAQ21guUpa1j2BJgmdstgiCwKascByq8YxtnVJrwJha7xKLKCbMB21qz9CcKAkUkzrK5Six0H0hC9Mpahi0KwinpncMq14qQsGLRfjB55X474Ue/PFAPVb+uQhvnGhxitRM4suHgMfBIIuij47K7jEsiJHQ/vnppmH4i3iYR6P/9KSdKFWY1HSWJpoPvCUsnTpuWRCFXRXOkR68ebKF9bXb2bhq6W5wQoMgPz5t4qVGb0R5+4jVFDutwUQrxXElBEGAbal326Jo8DgEvnSSRAlwkqaYJMn9JEmmkiSZRJLk753xmNMhCAKFqcE46aEx5tF693X37IUAriqMxUIe4mW+lJ1CLof55O4MLnaNYgFLDajOYMIX5V1Ymx7KypPYVjZlheOpVUn4pKQTHxZ3OO1xz3eM4Ju6QTy1KgkSX+u6aY3eCBHPteET02G7tAAAdyyOgUZvwn4Xa2hresdwplmOh/MTWLsE6IxGt442Z7IsUYay9hG3aq4tfFrSycr32tlY3qNUjYhshmW77CgJajyQbGfPkl3rkBIPvVMKvcGEX9+Uhf3/sxJVv9qEr57JR3aUBH88UI91r5xAp9xx6dV9S2PxcYn93tfH6gawNp1avkeSJE43DyM/SUZbCFX1jCEnmr6odiW5cYGMy3aZEQEQC3mslu0Ac5HdNDDh8OfxXIscqWH+HpXcuGLp7qsLPbhtEb1LzYWOEWgNJhSk0HeBB8Y1Hp2MR0hFrBxJQvyFWJseiv9c6HZaTeMd3jUsWJUagpMsul6uoN1N5tWOEBXog+6Rq99E2VGzgz0A81id6obmTMzjG3adm4M1/RhW6nD/sjiXvZ4fb0rD2vRQ/OrrGpx1wuImSZJ46VA9gsUCPMxyQeNkI30SlauIkIgotaEzmR8tQUqoGDvPd7v09bxzqg2+Ai6rBDvAHGkbJfX8gustC6JcuhhDxYRGj3GNHpFu7hJbiJP5op2iIMyICACXZtkum2K/wdXoDCbwbEyx6xlV44G3SwAAHz2+FNvz45EZaf57LYwNxAeP5uHzJ5ZBrTfioXdLMEzj28wWiQ8fPgIuK4nATDR6I0AQtIfIdrkKPaNqFDC4s5xpMRfNnkAmFmJERe8vzONykBsfyLpTDJgdhhwJHdLojThSO+CyuHS2OHvpzmgiMaLSM+4fnG4eBo9DII/Bz/tEw5Db9efTWZ4oQzHLHZe7l8RgWKnD0TrnSGy9uygufQtQmgthgiCQH2ZCx4G/ufUlqHVGj4wQbIVq2S4nSoKeUTVGZhie58YH4nyH66ymjCazXyjb39tHxR2ICfLBKhdabnE5BP5+zwIkBvvhmU8uoENuPU6SiRONQyhuVeA7a5JZ2/S5294GANZlhOFonfV4VIIgcMfiaJR3jKB1yDXJUV0KFfZc6sVduTGsOuuAWX/M9nDlSnKiJWgcmHBrIMxnpV24j4XvtavITw6mTCQT8blICRVTL9v58jHuZq/ikjY5ljL4sc5kWKnFg2+XYEJjwPuP5tE2PJYmyvDO9iXoH9fgsffKHNYZ37c0Fh/ZMan6urIXm7PoE0FPN5nvkSsZdLGTWoNb7PzoIEnqqYOF5YkyNA8qWUtMFscF4ULniN0d1rdPteIxljsNrmZxXCCanbR0d6x+0Gok8+nmYcyPkULMcN/qUEyyCqFyFQnBfmhneY9elRqCsACh01xevLcoLn0L2P8j4P1t5sJYOYS1pY8hruSX5j9j+jnltI6ycoj5+61wqmkIK91oXm0vKWFiNA5cXcxYPEVndnSkvgKMsrA8sZfiVvY3qcaBCZS2KXBfXpzL3QX8RXy8vT0XAPDY++UYZeheMNE3psaPvqxEYogf66LFki7kbgeFSKkP+lhaGN26MAocAvjPBdd0i/96pBEcgsCTqxJZ/0z/mNpjndKZ3L80Fh+ec578holxjR4TGj0iJJ77u0dJfdA3Sv3emRct8Zplu7I25pCCmfx8VzV6RtV495Ell6UgdCyOC8S/7l2Eqp4xPPPxBYdGtAEiPuJlfjZNquRKLboUKkYXn1NN5hCGOBn1RGVUpYPEx3MFMQDEynzRqaCXoVj0redsCOsqTA1htTMxk/MdIwjyE7pdp8/EYwUJeONEi8OPYy3mfGhCi0vdY1jN0AU2TLn9eBJbdMU8Lgd3LI7G8YZBp9j1eW9RnHkLEJIODNUDry0DXlsGYqge4+Ik1AWtpf4ZikIa728zf83Owvhil/u7e/Yg5HFnXbCzaIpiwLwR7Kquly1BCx8Xd0DA5bglqQswm7O/fv9idMpVuOfNYgxN2DYW1RqMeOqjC1DrjHjzwcWsta7FrXIsc1GsszX8hDxMsFh+Cg0QYVVqCL4s73b6e6OubxxfVfTg4RXxrAs9jd4IAdd7pjRxMj/oTSZWWjdH+aSkE/d6sEtsga6RZlm266W4Cfm5cdnO7MdKsvIHB4DKrlEcqO7H06uTsISlo8n6zDD84dYcnGgcwl+PNDrycnHboigUNQyyjit+81QrdhTSHyINRhPOtcqxMoU+hOGkm4NvqFgSH4QyJl1xZAACRDycZZFmZqEwJRjHG4Zs0hb3jKpxuLYf97JI0XQnMrEQ82OkrKZ6dHTIJxEb5MvY/bZIUJn8vC90jmJRrOdrnuhAH3QxHKSmc1duDEwknLJw571FsTgE2L4X8A0GVMPm/3yD4fPEAexqpLmgUBTSGKo3fy3zFptfwrhGDz+B6z1lXYXEh484mS+lfnhlSghOuWBxcVSlg7+IR7vwMZ1JrQH/vdCDrTnhkLnRg3V5kgzvPrwEHXIV7n7jHCvNLWC+Af9iVw0qu0bxyl0LkBzK3jqutE2BPBu6Wc7ErMdn92+9Y2UiBie0+NzJ3o9/PlgPfyEPz6xKZv0zJW0KLEv0zO+MjkfyE/DeGdfYKFnoGVXDRJIe7RJbkPkJKPW0lg5rFUVq4qK4QLd5oTcOKJEaxv5z+PLhBgT68vEYy9AYC/fkxeKOxdF442SrQ5ppgiDwvXUp+OexJqtd9hONQ1gYI6UPPyp9C7XNLZjQGMxWbDRT0Yb+caQ7yebSXuJlvpdT46jgcggsT5Kx8qe1QBAEHlgWi38da2b1/SqdAf93vAU/WJ/qFbKJmWzKCkdZ+wjrA9NMPjzXgVsXMocJFzUMItRfyOiA5MkGznSWJ8lYTw7iZH5YlRqC9862Q61zrKHjvUUxDXwOB3EyPzQPTsz+Q5pCGtv3mv/MRvZd6sMNLG2jvAGZn2BW55Nu2W5hDHPSkL18XdnLennh68peTGgNeMCFC3Z0FKQE48PH8jA0ocWd/3cOTQMU76dpkCSJf59px+flXfjummRszqbX+M3EZCJhIknwWHaznE1WZADrG/nyJBny4oPw+vEWp3WLi1vlKGoYwjNrkllriQHzlvSiOM9YsdHhI+BicVygSwM93jvThkfybSvaXMXC2EBUUBS4lmU7qgN3TpQEl9zkhX68YZC15vxcixynmobxnTXJdulrX7ghA4G+AvzkP5dgcFBGcdP8SEY3CpXOgJONQ9icHUH9DVNT0ajddyGYGMOKcJJyKupJK7bpsHn+/KRgdI+oWXcHASA51B8LY6VWAxxMJhIvH2rE/6xL8eodoadXJ+G1omabZUn7q/pQmBrCuN9iMJpwcmpHg+nfQ2vwjj2q2CBfm5xfvrMmGfJJHT4vs9/lBfDmotgifbAUtpZC9/1tuD1dgC9dvCUPAO3ySSR4UGxuK1lR1Mt23SOzl+04HAJcwrnWbCRJontEzdpW7T/nu5ESKsZiDxU+ufFB+HRqw3zL30/hF7urKbtiTQMTeOCdEvxmby3WpofiBxtSbXqei10jWBjjueKOIAjWNnwEQeDZ9SnoH9c4ZXGBJEn86UA9wgNErF06LBhtGIu7k01Z4fimbsBpAQLTOd00jMVxgfAReP6mBJiL31oKj9kry3azi2JLmqI7GNfoIWXhIW9xiwkPENl9CJf6CvCbm7NQ3TOOtx0MXVgYGwguh8D7Z9tnhQqNa/R46VADdqxk0N5PTUVlqlZ8I/opAt8rpJyK2poq6kqkvvxZ96HpWNwxqJY7mShMDYHBRNJqtZVaA148WI87Fkd7bQCXBYkPH2vSQ/F1JfvUPrlSi+qeMRRacYu40DmKcY0BaxhCPQbGNQj1F7F+blfCZAtJRV5CEPLig/DGyVaHrs3ed8exULvryof8mWLzf1PSCGHDHsQG+c7uFjMU0lct37GgeXDCZt9LT2M2Sb/6d2JZtqum6BQuS5Sh2IbFBmtU9Yxdfj5rdClUKO8YwS0LozzaxciOkuDQ9wtxT14MPi7pxOqXjuMXu6vx8qEG/ONoE372VRW2/P0UqrrH8OubsvDmg4tZSUOmc7JxmDFpyh0siQ9i9AqdTn6SDEviA/FaUYvDKYB7L/WhomsU319vW4embXiSdnHI05jHtnF476xzZRQ6gwlH6wewicFpwN0IeBzaw1TO1BSK6qbF4xIuOTRMZ1ipRZAfuyLnaN0gLnSO4lkb34cz2ZIdjo2ZYfjrkUbacBO23JsXixXJwXjpYAO+ruxF76gaf/+mCf8+3Y4dKxMRLmEoTsQhmLx3F+RkAKTkGO1U9HTzMFZ4IK2NisVxzK5HyaFiBIuFOGtH3Px9S2NR3KbAF2VdlzvNJhOJnee78WpRMx4tSPCaw4E1ViQHQ67UoaiBncXYGydb8dRq6wEkRQ2D4HEIrGC4F3naim0m1hY0Z/LMmiT0jWmwywGLO+8tivN2AFtfvvIht0gjtr4M5O3AHYujZ3eLGQpp1O6y6en3XerHDTk0oysvRSzkzUpgyo6kX7bLSwhCqQ3ekNY4XDOAjVnUBvMz2TOVX+9pn0jAbAD+u1tycPgHhViRLMMX5V149Xgz/nKkEZ+WduLO3GgU/Wg1tufH2yyBIEkSOhvs6VzF0sQg1gcggiDw7LpUc7fYAW3xwLgGP99djZwoCe5YbNsi5YmGQazyAis2OpJDxfAX8VHe7rzPz8clHbh/aZzHR90zIQiCcpkpJ5p+2S4nSkJ5EHcmRfWDWMPiPUKSJF450oh4ma/N78OZEASB396SDQGPgxd2VTnsvpEcKsYL2zIRKRHhQHU/Hl4Rj2fXp7ByXDnXIoe1Z5/Q6D3uPGEhK5L5PUEQBPKTZDjbYl967Q/Wp2BJQhBONQ3jpUP1+MP+OiSHivGTzekIC/CO7idbHi1IQJdChSO1zIt3B6v7sTxJRq87n0ZR/SBy4wMZv9fTVmwzWZ5IHzdPxarUEGRHBeD1Ey12h7uwM1j1FHk7rv5/ccjlrwl53Mvd4ssLT5bvz7zlyml5+15zQTzzsRjQG03QGoys/We9GYkvH7FB1Mt2PC4HJpKE0UTa3P1dBPNZAAAgAElEQVSciVpnBIdDsHZj+LqiF4vjAl2SYGcvSSFivPFg7uX/N5rMvxu2yWtU1PVNICPC8x0KIY8LndEEkiRZFV0rkmXIjQvEq0UtuDM3xuai3mQi8cMvKqHVm/C3exbYfJgYUmq9ZoxHxz1LYvD7feYbL5sRPhPNgxPQGkxIDvW+6VRqmBjNg0qkzVjWmr5sN9PeKjcuCF+Ud2FRrOtkQ23Dk7gz17qLwMWuUdT1jeOPt+U4RY4TFiDCjzam4Zdf1+BY/SDWZbBrBDCRGx+EXJZuGAAA5RCyjtyPYGIcpG8wCODKVHSqkTSm1nvUm3gmAh4HBiNzobIiWYavK3vRPKhEig0LlIC5qE4I9ptTkkcmHloej09LO7G/qg9bZzToNHoj3j/bDhGfi+0sZGl9Y2rU90/g+S3ptN+j9wIrtpnEBPnOCiVjgiAIfGd1Mp7++AL2V/XhRjuabt71G7ARym5x3o6rl+qmFdJsOdEwhNUMuhtvRkxhv5VDs2wHmEdaFzodX7gzf3DZjX3r+8dR3z+Bmxd4vkvMBJdDnyLFlqIGdt0sd5AaJkbTILtgDoIg8NwGc7f49/vqbH6ud0634XTzMH5xY6ZtMqTSt6Aa6YOPpQh30GfclRAEge+tTcE/j9m+GDOdMbUeHxV3MmtIPciCGCkuUlwjMhmW7SS+fKeEEdCh0RtZfza/LO+CD5+LbfOcN/m7b2ksEkP88Pv9dU7dy2CLsfq/iNC1o18YD4JmKnqmeRgFXiKdsMDnchhlNRa/YnskFNci9+bFwjSlh//XsSaUtMpxqKYfLx9qwNacCFYFMQD0HPoHZBi7YsVGcV290DHiFVZsMyEI9rpiwLzzkRTih38da7ZrIXZOF8WWbrE15wBbKe8YYTTA9maodMXZURJ0KdSUYRUrkoMd3qQ3mUjU9o0jPZxdR3R3RS+4HGLW6feaYipERmlJkvKC4m5NWiiK6tlHYeYnB+PJwkR8WNxhk4yipncMfz5Uj42ZYbhniQ1+oFMb9aZ/b8PKSDjFZ9zVSHz52JoTgU9K7dt4NplI/PVII76/PsXhaY2riJL6UHozMy3bAeabmcnOEaY12NpGqXVG7Kk0d9qc2TXlczl4fksGWocm8amd//aOcE52O36ufxj1mz6llBcCQG3vODK9YEo1HWtOODFBvogO9LF52e5aZtu8SPx4UzoeWZEAjcEEsZCHF7Zlsp+ylr6F3No/YKfPH5Dip6a9rpotMD1vxTaTeJmfTfp9DofAcxvS0DAwgffOttv8fHO6KAbM3eJPS7uclqzUPjyJCInI63R9bMmKDJjVubm8bEcRy2oJ/XDk93e4tp/1cpDJROLril6sTAlmzGef00wVd7p3tyJdrPGa4k7qK7C5e/fjTWkoSA7GC7uqUcHCZqtxYAJPfHAegb4C/On2ebZ9jqY26sXjzZj39WaHfcbdxeK4QPjwudhphyPO/51swT15MQ7LL1wJ079hThR9sl1SiBitw66JDD/fMYJcFq41B6r7oNQacKcLwoHWZ4RieaIMf/umyW1hJRYO1/ZjJ2cLluWkXfnitKkoSZIgQXqdx/6iuEBcsOJhnZ9k1pHaqwm9VvET8rAqNcTmxUld2k1oJqORQHaBeG057XVVo/cOK7aZLEuSobjVtt2NrTnhWJseilcON6J7hP2iHnANFMVCHheFqcH4po59B4yJT8s6cbct3S0vQyYWYmRGRzg7ytwtoOvoLIoNtPlNZ8FkIlHaNsI6mOJC5wh6RtVeL51wiKniTqBoxI1nbveq4i5YLMTgBPsoTB6Xg3/euxChAUI89eF5xgTAE41DuP21s9AZTXhn+xIE+dlY6IlDQG7fAxUvEIQTfMbdyW2LouEv4rEO9iBJEh+ea0disB/rCYsnoZJlAeZlO/mkjjJKPDc+kDHFzF7IqT0INjr1L8q7ECfzZR07bwsEQeBnN2RgRKXDq0XsAiScgclE4nDNAApTg2mLmPr+CaR54fsqyE/AaMsGmCUU4xqDQyEp17nC6T7gbu3PoBMG0eY39I9pvHYZMUrqg74x21JECYLAb27OAgD8YneNTU2/OV8UA8DqtFAUt8odDhuo7R1HYrCfV56WHEHqK0BMkA+l9g8A1mWE4kjtgF3d4sO17B0nALN0QsTnYEOm99hOOZ2pUeYkLxActXcVd+syQnHMxgNkoJ8Abzy4GKNqHTb97SReP96CSe0VlxOVzoD3zrTh0ffKEBXog93fWYGcaHbWfDNpH570WhmBNTZlhSM13B//PNrEKBuQK7X4zd5a5ERL6cMZvIx50VJcokivy2aIko8O9EWPDUsybGG7vNopV6G4VYE7F0e7bPKXHSXBbQuj8e/T7TaFTjhCVc8Y+sc12MhwDT3dNIyVXqYntkCCZLzXFKQEgyCA4w222aheh5q9lX0IEPEZl0xPNLIPwfEEJGmbrhgwX39+uDEVx+oHsa+qj/XPXRNFMQA8nB9vl35kOl9d7Mbti5w/ZnM3Ai4H+uI3rvJmXh5mQlL7J5TfTxAEViTLcJrG/JwOc5eYvQ7JYDRhX1Uf1mWEQXwNOHswMTChhjfWdnEyv9m+j1P658tQ6J+zIiXY+VQ+5kVL8OLBehS8eAw/+LwCm/92Etm/PIRf7anFqtQQ7Hw6n5WdFCXKIUi/vA1CncJhn3FPkZ8UjHUZYXjxYD3ePtUK+VQYDEmSGBjXYN+lPrx5qhXPbUjFghjvW2qhY160BJXds8feTMt2ruJE4xAru76d57tAEOYuviv58aY0cDjAiwfrXfo8Fg7X9oPLIbA2nX4ZfFStQ6Ctkxo3ESX1pbTxsxAsFmJetBTHbNh/uA41Gr0R5bWN+Ez4O/P0jea62qlQIU7mva4dSaF+aBmy3Rf84fx45ERJ8Kuva1n/zDVTFMcE+YIAbIoFnE55uwLzoqUei+J1JuuUX4N/8P9dedMrh/D84I/xnP4tTJx8jfJn1qaHoqh+yKbT2JG6AWzIZN8lPtcqh2JShxvnUHS2XSiHIPzoZvjoR7yyuBPxuVfy4af0z9PfK3T65+woCd57JA9fPZOP+TFSnGoaQmiACN9dk4x/P7wEbz2U69hhp3YXAidbAVnyFZ9xWbJdPuOeJDMyAM9vzcCN8yOx83w3Xj7UgL9+04QjtQMIl4jw/JYMr7LKYoOfkHflPTMNy7IdVRcZAMIChOhnKIDsYUKjt+rLapwKbliZEmL/IY0l4RIRnihMwt5LfU5x8rHG4ZoB5MUH0Ra9Sq0BvgLvbTosjgvEBYYQDwBYmxaKyu7Ry4fK69jH8YYhrDKcQZi2nTa/wRut2GayPDEY5+wIGuNxOfjjbTmU0i86vPs3YSPb8+Px7pk2mwX6JElif9XcC+ugQ5p7J8bESeY3/WvLgNeWIXCyFY2mKJT5FlL+DEEQWJMewjpFR2cwobhVjuVJ7LdV91f1wU/A9eoxjVOo3QXpZKtTQmRcQUFKME41TRXnU/rn6e8Va/rnhbGBeO+RPJS/sAEfPJqH5zamYU16qMOyB0flT95GWIAIT65Kwo82peG5Dal4YFmcxyLNnQXVoTmbYdkuNz4IZU4MOOkeUSGCZbBF75gGdzoY1sGWJwsTzSFAe2udtvRNReuQEk2DSkbJ2hkvSrGjIiXUujXk2vRQkOR1CYWj7LnUi/2ibTBu/jNtEJq3WrFNJ1wiwoCdh+vsKAnO/HQt6++/popiEZ+Le/Ni8a9jti097L3Uh1VpIV63qWsvEZEx+Dj91SsdStUwSN9gbDf9Amf66f/JC5KDcbrJepoQSZL42zeNeHRFAuvXpDeacLC6H+szw645zfZMFFnbcTr1J7QXIU+zIFp6ZRRueW3T3iue0j+fFa6ERpoCyJuvFOjyZq9YULwOEB1Ibc2WE0W/bJcW5o/6/tmuN/ZysLofW7Kt7yMcqDZ7XdsyyXIEPyEPP9qYigudo9hf1e+y5zlQbX5spr9XVfcY5kXZp+l3BxwOYfUekxUZgBB/IY6xbNJcZzYqnQHH6gaxJScc3GVP0uY3FLd6pxXbTGz1K56OLU5X11RRDABp4f5YHBeIT0rYeUeebRnGsFLrVXnfjkK1VEIAyIqQoJxhbEUQBDZmhWHvJWZR+sclnShMDbEpja64VY4Rlf7a9iae4lBNP6I2PutwiIyrsBz+vM3y6IKcB8Fj+72iQL/ObHKipKiikEnMm1qqvEShOeZwCBCgjom2h2GlzuoNzmgicahmAGvSQ9x6AL9jcQzSw/3xp4N1s3Y6nOFTbjKR+LS0E8sTZYgOpL72kiQJE+l9VmwzuUrCRQGHQ2BNWghONg55JBzlWuBo3SDUeiO2WZErag3eacU2k2QWEwZncM0VxYB5PCzgcXC0jjk3vLpnDOXtI3jEho7nnEA5hLtqnr5SVEwVGX9SPo++nk7Gi9GyRBlGVDp8WU4d1nCqaQgCLsfmk+W+S2bpxLV0+KCjXT7p9VGjC2MCUdE1ckVDzLCE4Q4sHQDOHPUH/zaQEiZG48Dsm1JGRAD4XAIXaXysF8cF4rwVDSkb2oYnkRBs/SB+oXMEw0qt2509uBwCL9yQidVju2ftdDjDp/xk0xC6R9S4f1ks7feY45G9Lyp8JjlREspD1HTWpodiQmNwynvn28ieyl6EBQixhCE+vHdUzUqO5A0sT5Sh2A5dsa1ck0UxYA71aBuexMclHZRaxfbhSey51Ivvrkn2wKtzMbW7EKxugzYw9SpNa7C6DZuIYsot8uk8tDweoQEi/OVI4+WYxJFJHfZX9aG0TYG7bPRx1htNOFjz7ZBO9I9pEO6lfo/TWTEllUHtrisaYg/qn6t6xrBIpveKAv061PC5HBhNs7t2Ij4XmREBqKQpipcnyZySUHaohl1I0IGqfgi4HI/EqxekBGMsYSuaEW2TTp8NHxV3IlgsYLRiO9k0jMIU7288LIiVWg0DWpEcDD6XsCmF8zpmxjV6HG8cwtacCMZdj2P1g4wuJt5EaIAIg+OuX7z03hVVJ/D4ykQ0Dyrx96NNCPMXYkVyMIpb5egf10xpwNK8fsxkF3k7MDShQbl4FbZYxs7b90JVsRMf7o1FWLt1DdGq1BDEBPrgt3tr4S/iQ+rLR15CEDazTK6bzrkWOUZV+mtmkZGJfVV9uHGe9/89fQRcaA3GK5KOzFuuSBS27zUXxG6Ue5xqGsbjwtNXiofte81/8P62KwW6l8hPvu2QJDlLojU/Ror/nO+G0UTOugmL+FxoDSbKn7OFUZXeavIfSZI4VNOPlSnBHnP4+N6N+bjv7y+gyPen8FNNHQYclAH1jqpxrH4AT61KgoBH38tSTGohmwNJoQEiPpTTvM6p8BfxsSQ+CMfqB/H81gw3vbJrg8M1A9AZTLhxPrN0ondUjag50ikGAM5UdLwr67ZruigGzDqUn2xOR4d8Euc7RrAhMxzhEu/v5DlK0Jrvov5oE7ZYviAOgW/B00gpPcGoK55OYogYv7452+HXsu9SH8RCHgrtlE6caBzC2ZZhCHnmLrPRZMLdubGIlbHXNLuLoQktQudApxi4kimfMLPY9ID+WaUzQLjmKYDH9XiBfh16Yqd8rmd6mi6IkeKDcx1oHlQiLdx/1s+lhIqnRvuz/4wNDf0TSAu3Lguo7hlHz6gaz65Pset5nEFKmD9+G3sR2j4j/Cz3btIEXPwQWPmcXY/5WVkXSAD35tFLJya1BvjMsUmctYPS2vRQ/G5fHboUKpt2WL7tfFTcgYRgPyxk8EJX6ebe+yU13B8NA+zCe+zlmi+KLcTJ/LzanNrZcDkETJQWSYHYd6nP5actC3qjCYdq+7E+I9Rm6YTJROKtU60ICzB7u1rQGUz4vxMtiA/2w01WTsLupHVIiUQv1xJPZ016KHZX9ODxlYkefR3dI6orXrJeUKBfh5550RJc6h6bdS2dP3XzregaoSyK16aH4vOyLruL4m/qBvDQ8jir33egug9cDoENGe5xnaDk1F+wqf8NgAAmOAHwF/IAtQI4+mvzn9tYGOuNJnxW2mme3jEUhmdb5Fie5L1WbDOJk/mhQ65CPMM1c81UUXy8YRAPLo9334tjYFSlw0fFHWgcUGJMrce4Rg8+l4N16aHYkh3h8WbN+Y4RVHSN4jc3ZzEeOM40y7EiZe68XwDzztOeyl6XFsXXrKb4OmZmWpjkxgVhXGNwyxYnYL5Qj9rhOjGm0uM3e2tRmBqCWxZGXfVnAh4H/7MuBRIfPl48WA+VjnkM5y4O1vRjEwu7KG8hxF8I+aTO0y8Dx+oHsS7dg0XMdViTFGLu+M4kQeaHABEPFV3UIR5SXwFG1ewN9KdDkiQmNAarcgiSJHGwuh/LEumDLdyNzmiC3uiY88bRukEMTmhx/1LmQ0FF1whjZ9DbYLOAmRjsh8RgP5fa3LFFManDS4fqUfBiEV4+3IjK7lGMqHQQC3mY1BrwxwP1KHypCDf84xQO13ju9b57ug0BIp7VdN7KrlHMj5477xfAbK0mV7r2nvWt6RR/G4mS+qB7RH1VdyE33hweUN6hoOzoOJuvK3rhL+KximWdzr+KmvCDDamQ+NDfCFelhiAj3B8vHWrACzdkOhwe4QiWGzfT6/VGAn35GJn0bCTswLjmWyFpuhagm0BxOATmxzAvT4X6CzEwrkGYjfKiyu4x5LDw3W0aVKJ1eBKPFHjYTWiqE0ye/SdkagWgA0ifIBD537O5S0ySJN4724YIiYhxcdBieTeXdmTiZb7474Vuxu8hCAI3LYjE3482oW9MjQiJZ/SvB6v78cMvKqDSG7E1JwLfW5uM9PCru5VdChUOVvfj8/IuPPHheWzJDsevbsqy+f3uCN0jKhyo7sOOwkT4MaSLmkwkSMzW/88FXK0rvt4pvobJiTYnTU0nNsgXwWIhyttdb3Oj0RtxuKYfm7PCL+uB2VDUMIgl8UGsCszQABHuzYvFa0W2BbY4m+qecVY3bm9jbXoojnlwu3tco4dYOLcOEt92CJhvSjNZECNFQ/847eRmfUYYvrFik0nFgao+xgQ3Cwer+0EQwCY3BXYwsvBBEMSV26tKbwIWPmjzw+y51IfiVgWeLEwEj0t/uzZ3iedWWiLbpctbF0aBJIHdFb0ufkXUfHCuHU9/fB4pYf44/P1CvHrfolkFMQDEBPliR2EiDjy7Ej/elIaj9YNY/5cT2HmeufB3Ju+fbQdBENhuRWpS0zuO7Mi5d78CzBaQtX3OCwSayfWi+BomNcwsSp8OQRDIjQtEeYfzolfpON4wiAmtATctYK/7VeuMON00jI02uFykhvkjO1pC663sDr6pG5gz1jbTSQoRo3nIPVIaKk42Dn0rvKuvJRJDxGgdnpz19fnRUphI8wGRipggX3QqVDY9V+uQEtGBPuAzFIQWDtf2Y2GM1POLrjO8vyd5UvgZRqF8c7NN1oJjKj1+s6cW86IlVvW0p5vkKJhj+lDAnAQ4oWGW1cTJ/LAoVoqvLvS4NEJ7JiRJ4sWD9fjF7hqsSw/DpzuWsdLE87kcfGdNMg59vxCZEQH40ZeV+O3eWpeHJSm1BnxW2oUt2eFXdjRoONk0hJVz9Lq7NFGGkjbX1S/Xi+JrGP75dyDQTDO7nkpVyo0PRJdCjcFx+7LE2fJ1ZS+CxQIstyHo490zbXjMjvHnmrRQaAwmnGtxvbn3TNQ6I0iSnJMezARBQMjjUHp5u4O6vnFkRLhexnMd55ETLUFVz2yZxPRlOzqSgsWzpldM7DzfjTtzrfui942pUd0zjg0MHr5uY4b3N+97JejgxkI83oyx81+wfpg/HayHYlKLP9yaY3XMrZkjqWQzWWBFcmPh1kXRaBiYcGmHcDokSeIXu2vw+vEW3Lc0Fv/3wCL4CGz7/SYE++Hjx5fi4fx4vHO6DY++V4ZxKwcAR9hZ3oUJrYHV/XNSa4CYQV7hzQT5CTCqcp2u2KGimCCIOwmCqCEIwkQQRK6zXtR1nEDpW8D+H+GumqdBKgevSlXaot4DADjnwnSYCY0eR+sGcUNOBOPYbzqNAxMQC3lWT7l0PLA0FscbByFXut7gezp7Kntt6oZ7G2vSPCOh0BtN4BKEQ96113E/CTI/tA3N7hSH+AsRJfVBJc2yHQDctigKX13sYfU8PaNqBPkJWBV7R+vM798NmV4wrcnbAWx9+bIvsVASDvKhPfit6VHsqF/EKra4vF2BT0s78eiKBGRbZFmlb1FGR7cOKb0+QZOO+dFSXKKIDp/JtpwI8LkEdrF87zjKO6fb8GFxB54oTMTvb8lmfQ+bCY/Lwa9uysIfbs3BmeZh3PbaWXSP2DYtYYNKZ8Bbp9qwKFaKhbHMMhp7dP3eBkE4Lzp+Jo52iqsB3AbgpBNey3WcSeYtl1PsTK9enaoUsfxeyPwELi2EjtQOQGsw2VQsflHWhfuX0vtwWoMgCHx3TTJeLWpx25iNJEk0DymRHDp3u50Wmy13U9KqQF6CbXHh1/E8HA4Buk+XtaQyHpeDedESVtG9n5d14W6W6Znf1A0gXuaLpBAviTjO23FVUEd8XDwyb/4hStsUePjfpRhT0XcMJzR6/O9XVYiS+uAHG1LNX5xqclBFR/cd+SfWzUHpFmAOEWIzpQr0E2B1Wih2V/S6XIZwuKYfv99fhy3Z4fjp5nSnHNrvWxqLDx7Lw8C4Bre/fhb1/c7teP/lcCN6RtX46RbrISdHauem1G86mREBqO11zdTAoaKYJMk6kiQbnPViruNExCHA9r0wiGTgquVXonO37wXHPxRr0kNxvOH/t3fn8VFV5//AP2dmsu/7RlYSCWFfZEdQXABR0KqttYpasS5V+2vt6q/tr+3XbvrV2qq1aFXcWzdUZBFwY1/CThIgBBICZCeBrJPMnN8fkzFDMpO5s2WW+3m/XnkRksmdQ7gz97nnPOd56r9p4+xuH+8/g4zYMEy0c9dqZirOH+X0HblZVGgQFoxJxbu7h2Zzw56qc4r/jb5KCIHoMN2gF2pP2FbRgOnDGRT7I40QVt87JmTG4nRzB+ou2E7Num5sOlYdGHzTVENrF0J0GkVd6dq6erC1vBFXjkzx6VWHb00ahiduGoudJ5qw5PktOG4ll3/XySYseGYTyuta8T83jO6rINA7yWGtdfSeqDl+0cXOFlsbN/u7cUIG6i50YUu56y3DbTlY3YJH3tmHsRkxeOqW8W6tcDBjeCLevW86AODmF7Zhp5vyYvefasbLW07gtqlZmJIb3/cNGysLp5s7/L4RyrS8eOw44ZmVbuYUBzhbuWjzCpPR0tGtaMbGUU1temw+1oDrxqUrvkh9sv+M3ZaUSl2aE4/61i6csLIZyN3Wl9ThypH+fdcNAAtGp2HNobND9nxdPQZoNRq/LAlEpk6hx62kUJjzigdLodBoBKbnJQwa3Ly8+QS+o3CWeNOxeugNRlzpC1Un7Lh5cibeWjYNLR3duOG5LXjqsyN4r7gaOyoa8cS6Mnz7X9ugEQLv3jcdl4+weF/pneRAeKJpgqN3kqPppvcRFusDedQuyE+JUrTZ9/LCZESH6jyWQlF3oRP3vLYL8RHBeHHpZIdziJUoTI3G+/fPQFJUCL737x349IBr77ndBiN+/v4BJEWF4OcLCvu+McjKwqymD138V3hfbHgwmj00iWM3KBZCbBBCHLLysdiRJxJC3CuE2C2E2F1fr3wHLjmp90Ug2hvQrovrezPtfZHMKkhEkFZ4JIVi9cGz6DFKLNV9ZvVOtb/mdj3CgrVu3Shy72V5eGXLCUX5e86qu9CJhIhgl2e3fUFuYgRONro/182WL4/UD1p3VZVszOz4IlPKTfOAMY+O0WOpbj32Vg1+s31Vkak8W/80Jykl/rHxGGYVJCqe/VxfUoeYsCBMznZuxebQ6RZsPd6A3SebsP9UM7p6PLvp9NKceHz04EzkJUXiH1+U49F39+Pby7fjuS+O46ZJw7D6kdmYlB1v/0AANpU3+P1SuJImHgAQGqTFtWPTseZQDZrc3HSox2DEQ2/tRUtHN168YzKSozyXczssLhzv3TcDYzJi8OBbe/DMhmNOp/st/7oCZTUX8IfFoxFtuapiY2WhOXI4smff5qZ/iXdpNNZXq1w+rr0HSCmvlFKOtvLxkSNPJKVcLqWcLKWcnJTEi6HHWeyAXjH+bcgHtvW9SEpWIio0CFNzE7DRA0HxB3uq8ZPYr5C25ddW71T7X+jfK662233HUUFaDe6ckYOXNp1w63EtfbDnNG6cmGH/gX4iNToEZ1s6huS59p1qxng/6r7lcYPM7PhiYJwVH474wysGjDnsrcX4ne4V0/cGIYTAkvEZ+J9PS7GjohFSShiMEn9ddwRzRiRhhsJ2xQajxOdltbh8RJLDN6dGo8Tyr4/j0OkWaIRAR7cBdRe68KfVZQ5VyHBGZnw4Vj44E2V/mI8vHp2L1+6egvfvn46/3jTOelWAfmXezJMcMzbfibywoXnNekp6TCjONCv7N9w9MwedPQYs/7rCrWP467oj2HGiCX+6cQyK0j3XQtgsPiIYby2bihsnZuDpDUfx0Nt7Ha4AtPNEE57ZeAwLx6QOLGFqY2Xh9RHPYlim8/t2fMmYjBgc8kBesX/W5CD7piwz/Vm0BAllnTjVFYmspatMwXLv964oTMbvV5WgsrEN2Qnu2b1cVnMee6qaceNVtwBlX/XdqQKmF2ZSoekutlePwYj61i6PdDTLS4pEfEQTdp9swuQcZTMvSnV2G3C+o9uvc/n6Wzg2Dcc+eRpp19/Tt0motf6ic8YdLnR2IzJE59P5n0OuaAmw6yW7rxdfIYTAkYR5mNf6yYAx14fl4sWmsbhNbxh0CXpcZizGZMRgfWktHv+0FB3dBtw1Mxf5yco3y+2pOodz7d0Op06063vwv58dxQ0TMvqqO/SaV5iMN3dWYevxBtw1M1dRjWRnhei0yE2MsF89wrLM29JVAICeV65FUuMRt78+h5oj73h0pHQAACAASURBVAMFKVG4flw6Vmw9iXtm5yLRDe+/qw+exfKvK3D7tGzcMMG9kzODCdFp8b83j8MlKVH4y9oylNe14vEbxmCSghWPj/adxk/fPYBh8WH43fWjFT2fEfC7jquDmZIbj//sqnL75IqrJdluEEJUA5gO4FMhxDr3DIvconcH9NhhMThwutkU6Fi8ec7rzYV1ZwrF2zuqEKzT4NppY63eqZrLFJltKK3D1R7MBbxlciY+PXjWboF4R72xvVLxznh/kVz6OmYe/bPHZys/O1zr0f9zv2RjZqf/68WX6EMT0H37xwPGXHbNW6g1RNtNoQBMS6DXjErFY9eOxK8XFTkUEAPAhpJaBGkFLnOgEcH5zm78cXUp7pszfEBAbB7T7dOyMW9kCh7/tBT6Hs+lYCnWr8wbIpPw4dgXcG7OH/06IDaLDQ/GOYUpEQ/PK0CXm2aLy+su4Kfv7seErFj8elGRy8dzlBAC980ZjpeXXoqWjm58659b8csPDtiswyulxHNflOORd/ZhfFYsPujNTx7AysqCpr0B3yl5wKEGMr4sJiwI5zusd890havVJz6UUg6TUoZIKVOklNe4a2DkPvlJkSivG7iRITshAsOTItwWFHfoDfhg72ksHJ2KuIhgRT/j6eoNQgj8JPYrvPLZrr4vupirWXehE+16g9tm131G0RK0RA63usPdnbOVx+paFXWGIt82IiXKagWF8Zmx0AhguwN10IUQTu0pWF9ai2l5CRfnU9rxyuaTeGTeJdaDCQvDkyLx/Vm5ePKzI4qqI3hcvzJvx9vDEXf5g14ckPtMzIrFHgU3UYDp/2XJ+Ay8tu3koFVO7Glq0+PuV3cjLFiL52+biGCd9/aGXF6YjPU/noNls3MRvOdlvPrXR/DzFRvx2raTOFpRgYoP/4DPXvkDrn92C55YdwSLx6fj9e9PQWy4jeuseWUhIR94YDvwwHY0hmYhuOmo6XsBItgDjaf8f4eQIyw3hex8Eagt6wuOfHhTi6t0Wo3N2o7zRqZge0UjWrtcv+P65MAZXOjswXenZtvMgftmFhJAeV0r8pMjPbuMvvNFRG78Jb5f/hDW7jzoltnPFVtP4s6ZOe4dpy+ITILu7k/RHhTnsdnKhtYuJEYqu2FSFQWvF18zLr4baR/ePGDMUe/cgJlpEts92IoVML1/VNS34cqRylcdjte3IjpMZzcgNsuMD8fi8el47otyZ4fpEaea2jEszrkmR75oVHqMzfbg1jw0rwDdBol/feXcbLG+x4j73ihGzflOLL9jMtJivP+7jAzR4bGkLfid7hX8CG/hoRMP4LWP1kL76gLk7X8SV1c+icX6T/GbRUX427fHI0Sn/CaypUMPTQCmq03NjXd7y2f1BMWWG1k2PWX6fPls05+bnvLpTS3uIACrO1yvKExGt0Fi8zHXL7xv76xCfnIkLs2JG9DqFA9sv2ijHwCsPXQWC0Z7uJxQ7y7ciPPlmLP+OvQ8O9Wl2c/DZ1qQGRfu0MyUP4kI9uw2gzUHz2LBmDSPPodfUvB68TVp1WsR03rc6phvi9yDfVXNHm0fvu5wDQDg6lHKg+I3t1fhtqnZDj3PqPQYTMyOw+vbKx36uQHcWF1k9cGzWBhAr6NgncahSkG5iRFYMj4Db2yvRE2LY7PFUko89uFB7DzRhCduGutbdeaLlgDhpgmIYajB+tBfYLim5ptv3zMlCXfPyrU/kWSuPtFYDjw/DSEvzkJcR6XP7lFw1sTsOOxxc1lZ9QTFliVKtv4DEFrAoDf9ufUfHlkm9iVZCRGotFJya1J2HKJDdVh7qMbKTylXevY89lY149YpWaYXrJUcOCxdZfralGWQUqK1y6CoOL9LLHI1w7rPQdfZCGNYglOzn1JKvFdcjZsmDd1mjCHVO1sZ3n0OncEDy/i5SkqJqqZ2ZDjZxjug2Xm9+CIx9V5szP2p1TEHTf8B9AYj9lbZ7m7nqnWHazA+M1bxLN8XZXWYVZDg1DL5zPxEhOg02Ops4whXq4tYBNRSSnS21Nit8OFvIkN1Du39eHhePoQAfvBGMTr0ym6+pJT4x+fleLe4Gg/PK8Di8T5WPcj8GuoNzS7qHZmQD0y43bHj9L6Hh+qbfH6PgjOCBlkFd5Z6gmLLk6SjCZAGAML0Z0dgnjCWxmTE4ICVMkNBWg2WTMjA6oM1LuVnvdW7we5bliXK+uXAWW70M+USe6ckV3u3AZ1O1CJdue80rhyZEhB1ia2ymK3816g3Ybx/m1tnKzcda8DsgsB8fbnFIK8XX3V42LfRFWpR2aV3zJNz4h3OK3bE6eYOHKhuwXyFK036HiM2HWu4uCGGg26eNAxfHq1HQ2uX4z88SEc6uxMx/QLqsuMVuOvYQwG3snlpTjx2nVS+FJ6dEIFnvjMBB6qb8fA7e+0GRwajxG8/Poyn1puqjvxoXoGrQ/aMiAQgtH9ZOAHc8mbAxieuSHGgpJ8SAXp1p/7yk61vtgOAO2fkQG8w4s3tVU4du6WjGyv3nsaiMWm2E//7+epIPea6cIFSzEquZmRPM86/MB/6llrFh/mirA4deiNm5iurn+qXLGYrJxZdgk1nhVtnKzcdq8fsggD+/alQYWoUys5eGPD1mLAgFKVHeywoXte7snVN//qsNvx39yl8d2qmS/sXhBB4eF4Bnv283PGNd65UF+kXUGf/Zx6iLxwPuJVNU0MYx+pDXzMqFf/vulFYX1KL339y2GYTjHZ9D37w+m68tq0SP7gsD/978zi3tnB2m9Z64JX5QGf/FRYJvPMd5St2Fte9jqA4GP1gj4Kz5hQk4euj7vs3qScotgyOwuJNaROQpj/D4gP2hDHTaoTNN4y8pEhTbc4dlU7lAL7w1XG06nvw/dm5ih7fbTDCKDE0u31t5Gomd57Auv++gEYFsz47KhpxvL4V350aGEXPB9U7WzkrP9HUhtdNs5VHai6gIDmKtYkDzNhhsVZXoABgWm4C9p7yTF7x2sM1GJESZb++L/rSdvKTXa94Ehmiw82Th+HVrSddPpZiJSuBm1Z8E1CHd58DgiICbmXT2aXwpTNycO9leVixrRK/+vAQDla3fHOta+3qwfvF1bjpn9vweVkdfr94FH65cKRvBsSA6f+60WJTZ2gsvgnTzp0A9r6u/Dj1ZZBJhXhxzFvQ+MEeBWdlJYSj+hxnih1nGRzNeMiUNqENNv0546GAPWEshQxSvuTuWbloaNXjk/1nHDrm2ZYOvLz5BJaMz8Co9IE1P60Z0hnDQXI1L7/jMTz7RTlKz9re9XywugU7TjTh+7OUBfyBQgiB7IRwVFgpueWMj/efxvXj091yLPIdKdEhqLWx0WlaXgL0PUbsO+XevOKG1i7sPtmEaxSmTri77OOo9BiEBmlRXOnArndnq4uYUyf+exsgLTaidbcpD5D8SGSIDuedqCn/i/mFuG1qFt7dfQrXPbsZc574Este243J/7MeP3l3P853dmP57ZNxx/Qc9w/anaYsA4puMH2ekA/8cDdw/7ZvNt8hROGNXe91b9ecFRhXWOAXexRcoRFwW8tn9QTFlsHR7B+bPr93k+nP2T8O6BPGbHRGDA7amNWZMTwBhalReHnLSYf6sP9t/TFICfz4qksU/8yuk+dwqZs7zA3KRq5mZIgOv762CJuPNeCJdWVYc/Asznd2o6vHgDUHz+JPa0pRXNmEh67IV+UM5w0TMrBy72mXj1N/oQuRIUFO1aEl3zbY6+LS3HgIAeyocG/JpA0ltTBKYL7C1IkNpXXfNCpyl1unZGL1wRqbTRYGcLa6SNESU3DUWA50NEHC4ve9782AW9m8NDceu5wosaXRCDx+wxjs/r9X4q/fGoucxAgcPt2CmyYNw3v3Tcemn13ucNdDr7nlVVMsctda07UqpRB4YJvj8cmUZfiiGphlTvnzgz0KzhqfFYv91e65+VZXm2fLE8L8eUqh6c8APmHMJmTF4T+7qqwGpEII3D0zFz97/wC2VzRh+vAEu8c7VnsB7xafwp0zcpEZH65oDG1dPQgL0vrM8pVGI7DssjwYjRKHz5zHm9ur0NFtwLzCZMwfnarKYNgsPFiH6LAgnGxoQ46CZWpb3i0+hVsmB1b3P+oTGqRBh5WWzjFhQShKM+UVPwL3bWpae7gGWfHhGJlmf9aspaMb4UFat7dqNucX/23DUfxmUZH99wnztaVoSd8N+tJV9ls0RyYB428DNv7O9LyQpnS/8HhToOznLZ77G5MRg+e+KMc8B2pPW4oND8Ytl2biFn/vNtr//9SJ+KSlvRuRITpofeRa60nT8hLwypaTmJTt+mSbemaKCTFhQbjQabtJx/Xj0xEfEYx/bz6h6Hh/XXcE4cE6/PCKfMVj+KykxqG6okNFoxEYMywG988djh9fdQnGZcaqOiA2u316Nl7bVunQ6oGllo5udOgNSIxU1iyB/E9RejRKbKQgzRiegOLKc25rs36+sxtbyhsU37B+vP+Mx9J2YsKCcP24dLyxQ+EGZWeri0y4HQi2uCkVGlMlggBc2QzSatBj8IHugQHg4wNncN1YdaSshQfrFJfls4dBcaDrVzA+TN8EuWO51YeGBmlxx/RsbCitxfvF1YMe9quj9VhfUov75uQhXmFLZwA4WtuKwtT+5WbIV4XotLjskkRsLHWuFfgLXx3HXTPVlY+tNmMyYnHQxtLlNaNSoTcY3dZKfmNpLboNEtcouLGWUuJUU7tH27FPyIqDwWDEATct3Q5gzkXWt6EjyKJ2+HtLA6ryhKXoMB1aOtxzE6Vm1U3tyEpQtoIbCGLDg9DUpjCdaRAMigOZlYLxt5U9CLHmpzbrWz4wNx8zhifg5+8fsFmofmt5A37w+m4UJEfibgc2oLV0dCMqVF0ZO4Fg7ohkbKtodLiKwNdH6zE6PcahmybyP0lRIWhotX4xmpgVh+SoEKw56FpzILP3iquRGR+GCZn2N84dqG7B2GHKNv+64o7pOfhw72m3zYZfpDcXuTEs11Q3PICrCJhNyU3ATg+3CA90xZVNGJfpnT4A3nJFYTI2lCovs2oLg+JAZqVgfEzrcZyPGm5zliFYp8E/vzcJeUkR+MEbxThSc3EN0i3lDbh7xS5kxYfj7XunIdyBtsAbSmpxlZO5Yj7Dja1a/cmdM3KwwoEyVK1dPfj6aD0WjvFwG2/yaRqNwILRqfjyaB3a9bZTt5Q41dSOLeWNuHlSpqI9CZ+V1ODqIs+ffxqNwENXFODvG485nWZk05Rl6LjqL3h/zL8QEZ8W8FUEAGB0erTNDeGkzNpDNYpreAeKvKRInGxoc/k4DIoDmZWC8TI8EW+MeG7Q+pYxYUF45a4pCAvS4q5XduKZDcfwwlfH8dwX5bj71V3ISYjA28umOZwnWl7fioIU12uFek3/mfdNT5kKrZs7SwVwgJwZHw6NEDiosLj+8q+O4945eczLVonY8CCcs7F0uWBMGjq7jfjyiGuVEt4rroYQwLcUtFnv6jFAK8TQ1EIHEB8RjCtHpuDd3YOnnTnj1e6rsGjG2L4vBPimcJ1WA4PRPeW11Kj07HkUpESpYoNdf+HBWpdvvhkUq4zSl0lGbBhevvNS9Bglnt5wFH9eU4Yn1h1BfnIk3rxnKhIcDIgDInXCcub9mXGmHeGN5aaSSdmzTcFygLVetXTP7Fx8caQOO+x0KVt98CxykyKQHBU6RCMjb5uQFYu9p85Z/d6lOfFIjAzG6oNnnT6+0SjxXnE1ZuUnIiM2zO7jNx9rwOxLhraxxdS8BLR0dKOsxnbdc0cdr29FsE6DdAX/5kCSFhOG025s3evXHFydXLnvNJaMzxiCgfmeOZck4ysXb74ZFAcyGwXjv1PyANqa7F+gRmfEYOdjV+L4Hxei9PfzsffXV+GTH85yOCAGgPUltbjaX+pE2mI5895tsUzT3gSsuLavDmmAboARQuChK/Jx8HQLPi8bmLvV2W3AU58dgUYI3DDB/mweBY5R6TE4dNp6MKjVCFw9KhWfl9U53d1uW0UjTjd3KC7tV1x5DpPc2LBDqbtn5eKdnaecakDRn5QSr2+rxB3Ts90wMv9yRWGy2zZn+jUr+4IGm3ypbGxDekzYkK2Q+JrRGa6n3qjzN6cWNgrGx7dXoGbbO4oPo9UIhAVrERcR7HR94eP1rW5ps+p7BNDR1HfjEWCtV/sTQuCe2XmoPd+FJ9cdwXvF1dha3oAdFY340+pSfGdKFuYr7DRGgSM0SAt9j+0l74Wj09CuN+Cro87N4vx39ylEh+pwlYIba32PETqN8EotdK1G4EdXFuDJdUeU3QAMMgv40b4zuHZsmttrLPuD9Ngw1LRwptjavqDBJl/eK67GzZPVOyEhhIBOqxn0vcge9b3a1MRGi2P91X/B2vDrhmwYAZE6AQyceQ+LB6DOmpq3TsnCQ/PyMSUnHhqNQEtHN35z3SjVLfNSH40ADEbrr4epefGICw/CGidSKFrau7HmUA2WTMhQ1BVxy/EGzMgfojbyVsSGB+PBy/Px5zVl6B6s9ewgs4Dtm/+Jo7UXhrbzp48J1bmeH+r3rOwLGjD50ntjVXe+E5EhOoTrzwVsCp8S0/Lisd1Oit9gGBQHOisF44Nn3Oe2QtdKBETqBHDxzPvST01dpcyCI0xvWOYLnAqE6LTISgjHtLwEXD0qVZUbO6jPJalRA6rVmAVpNbi6KBUbS+vQ1ePYe8/HB85A32NUnDqx+2ST14PJlOhQ3DkjB0+uOwKjjRsFW7OAMqkQL9SPwbLZeUM7aB8zIz8RW8udD25UweLG6t2v9+DWorCA39tiz5SceOw66XxJPwbFKqXVCJeWGBxRXhcgqROWM++Vm/o22c37LfDw/oCvH0o0mIlZcdhTZX2zHQAsGJOKC109WHtIec3iHoMRK7aexMi0aIxKt9/0p9tghEYIn7hBy0mMwJIJGfjj6lK0dlmZ8bRRHejvmU/hmqljEKfy+t7jM2Ox75SHmqL4Cxv7gr6ZfLG4sbrnwG2IfmV2wO9tsUen1cAope2bUTsYFKvU5Jw4FFfavoC5y7k2PWLDgzz+PEPGPPNuDpDvWgvM/rEq6ocSDcaUB9pp8/uzC5JQmBqFp9cfHTytwMI7u06hvK4VP7qyQFF5v63HGzF9eILiMXvayLRo3D93OJ5cdwQlZ+xXpWjXG3Bb0CaMirYobxfApR4Ho9UISEj31372Jzb2BX0z+RKZBP33PkabLhYhevXsbbFnfGYc9jp5Q8WgWKUudXGJQanVh85i4eg0jz+PV1hJTWFATGSdViPws/kjcLKxHe/sOmX38ec7u/H0+qOYmhuvOP1q54lGTM31naAYABIiQ/CbRUXYerwBr2w5gYbWLtM3LGYBe0IT0KaLRUTPOSTu+JOp/rmCagOBrigtBiVn3Vfizu/Y2BdkOfny392nFOXaq8nsgkSnN/UyKFap0CCtw7l9zjjV1KGq/utEahYbHoQmG008AODyEcmYkhOPZzYcQ5u1lAILz39xHE3tevx6UZGiWeJugxECvpE60Z9GY6raMndEMj7adwZ/XVuGr1YuB+rL0BCWi/WXr4Thvm2mdCzAlJqloNpAoJt9SSK+Ptrg7WF41yCTL9WnqrCweBm0HY3W0ytUKjRIC6NRKl6RssSgWMXiI0L6Zi084ExzB9Ji2MCBSC0mZMVh7yB5xUII/HxBIRpau/Dy5hM2H3eqqR0vbz6BGycMw+iMGEXPvaOiCdPyfGuWuL/cxAh8f1Yufja/EJfe8nMYFzyBxAfXY8HUsYhOTDelY8377eDVBlQkOjQIrV2u13wORN0GI/atewXx7RW20ytUbO6IJKe6aDIoVrHZBYnYUu65u/A1h2qwYAxr1hKpxeiMaBy2kzs7KTsOVxel4F9fV9icVf7z2jJoNMBPrxmh+Ll3nGjE1Dz/KWEWHqyDZuq9A2cBJ9zuvUH5oISIENSdt52rrkZSSjy9/ihGLXnUbnqFWk3Kdm7fFINiFStIjsSx2laPHb+htYutfolUJESnVbRk+bP5I9Cu78GP/rPvona+LR3d+Ml/9+PTA2dx72XDkapwpclolDAYpf83urBXbUCFFo1NwycHnG8RHohe3nIS80YmIzcxwhT8lqzsOz8ik0ypNirMQbckhLCbzmWNn7+DkCvMeXqe2N1bXtdqesESkaoIIdBjJzDOT47CbxYVYeeJRlzx5Jd46rMj2FBSi/l/+xor953Gw1fk46Er8hU/5/7qZozLjHV16N5nrjYQmWKqhW65HL76p94enVckR4ei7kKn49epQboF+rNP9p9BekwoJmX3roo42ApaTa4bl45VB8449DMMilXuktQoHKm1XnDfFesO17DdL5EKjUq3n0IBAHfOzMXnP5mLa0al4u+fl+Oe13YjPFiLD+6fgR9fPcKhWd9NxxpwWUEA5NxOWQYU3QC01gLvLTV97aYVpiC55EPVBjnjhzlYszhAA8WP9p3G+c5uLBhjUdHJwVbQapIRG3bRSpQSAdB7l1wxKz8RH+ypRmGq/cL4Skkp0drVg+jQAKpPTESKTMtLwNs7qxTN3KbHhuHvt07A0hnZ2FvVjO9Ny3a4vJSUEh3dBoQFB0hZqoVPAPWlfUEOYEqhUHGQM29kCv7x+TFMyIpT9gNFS4BdLwXM77Cz24BnPy/HpJw4LB6fcfE3zXnEz08z/RsBVW/O7G9EShTKapSX9WNQrHLxEcE41+5Yzo09+041Y9wwZTvGiSiwxIQF4UKnYxUDJmXH9y0HO6i8rhWXpEQ69bM+iUHOAME6DYQQ6Ow2KLtp8vDvsKalE8fqLqCpTY9zbXr0GCWyEyKQmxiOzPhwhOjcd4NW2diGlzadwA/m5GFYHMubOmr+6FQs/7pC8eMZFBNCdVp06N0307K+pBb/56pL3HIsIvI/ITqt8gDGRRvL6nDrpVkefx7yrmtGpWDd4ZqBM6VDpLPbgHWHa1By5jxSY0IxJiMGo9JjEB8RDI0AqpracfjMeaw5WIOuHiOCdRpMzIrDpOw4p66t5XWteH9PNRIjQ/B/F420HWj335wJ9G3OVPGNlFl4sM6hUo0uBcVCiCcAXAdAD+A4gLuklCpvVu5/5o5IxsayWiwam+7ysc616REZqvP/XeBE5LTJveWQZuYnevy5znd0IyaQWskzyLFqVHoMPt53RllQ7Obf4dpDZ7GnqhnXj0vH9ePSrTaTiQ0PxthhfSlDnd0G7K1qxr83V6Cz24jQIA3GZ8YhPzkSKdEhA45hNEqU17di98lzqGxqQ3Z8BB6ZV2D/xtKyFfTSVaavrVjUV6tY5aXZAAxdUAxgPYBfSil7hBB/AfBLAD938ZjkSTtfNOVUmd8UWusx+vSHWNU02y1B8Qd7T+PGCcNcPg4R+a+J2XFY/nWFx4Pi6nPtSI8N8+hzDDkGOTalx4ah+ly7/TQCN/0Om9v1eP7L45iel4BfLRzp0FhDg7SYPjwB04ebArJ2fQ8OVLdgY1ktaltMdZeFEDDX1JBSIj85EnNGJCEj1oGVD/O/xfK6vnSV6s8VZ7kUFEspP7P463YAN7k2HPIo847cXS9d9EYh6stweeEvcKY5x6ULjNEoUXehU3FtUSIKTKFBWuh7HG+x6qgNJbUX78QPBAxybLp+XDre3FGJH15RMPgD3fA7LK48h7WHzuLBy/MRGx7swqhNzMv4Hum62P/fZNEKmhzjzjXuuwGscePxyN0GKd0y4oo78PF+x+r59ff1sXrMCYSySETksugwHVraPduit/ZCF1KiA/AmfMqygZ3uGOQgLsIUnNZdUNDhzoXf4WeHa7C36hx+tXCkWwJi8h92g2IhxAYhxCErH4stHvMYgB4Abw5ynHuFELuFELvr69XZmcfrzDtyzV2SzPlWS1chLjkDze3dMBqdb+Sx7XjjN0tFRKRu0/ISsK2i0WPHD8jUCbLr9uk5eH1bpceO//bOKrR29eCe2XlWc4cpsNkNiqWUV0opR1v5+AgAhBB3AlgE4DY5SMsZKeVyKeVkKeXkpCTOJvqiqXnx2HGiyamfPdVkukDxTYSIANPGqJIzLR47/tpDNVjABkGqExMWhISIYFTUt7r1uFJK/PPL40iPDcONE7kvRq1cSp8QQswH8DMA10sp290zJPKY/jtyzTPGvV1/LitIwqZjzs3iv1dcjSUTvFMqh4h8j1Yj4MLCk12NbXokRoZ47gnIZ906NQvv7DrltuMZjRJPrz+KqXnxmHMJJ+3UzNWc4mcBRAFYL4TYJ4R4wQ1jIk+x3JH7wHbThznHuGQltBqB8GAtzjtYeP94fStiwoIQExZAZZGIyGUpMaE42+JYm1Uljte3Ijcxwu3HJf8QotNiZFoU9ladc/lYPQYj/rK2DAvGpGGi0o55FLBcCoqllPlSykwp5fjej/vcNTDygCnLgIVP9tVpNOcYL3zymw0Ii8am46N9yjfcSSnx5vYqfG9atqdGTUR+alZ+IjYfa3D7cdceqsE1o5g6oWaLx2Xgw72n0W1wvspJh96Ax1eX4tuXZmJkWrQbR0f+ih0W1MbOjtycxAjUtHQo290LYGNpHS67JBHBOp5KRHSx3MQIVDS0ufWYUkpc6OzhypTKaTQCd8/MxRPrjji1QfxsSwf+uLoU988ZjrykAGoTTi5hJEMD3D83Hy98WYFB9k0CMHXs2VbRiLkjkodoZETkb2LDgnCuTe+245WcPY+idM7qkWkSZ/H4dDyz8Zjd65WlvVXn8OqWk3js2pFIDsSSfuQ0BsU0QGSIDlePSsGHe08P+rg3tlfijulMmyAi2xaMTsOaQzVuO97G0jpcOZI34mQyKj0G04cn4N+bT9h9bGe3AS9tqsDeqmb8YkGh/RbKpDoMismqaXkJqGpqx5lm65tk3i+uRlSoDtkJ3OxCRLZlJYSjqsk9xYkMRol2vQHhwS41Y6UAMy0vAXlJEfjT6lIUVzYNmDU2GiU+2FONumKiWwAACstJREFUp9cfxVVFKbh7Vi7Lh5JVfGchm+6bMxy/X1WC707Jwqj0aAgh0NltwDMbj2Hm8ETMKkj09hCJyA+k9VahSItxrdnGxtJaXD6CJbNooCsKU3BZQRI2lNbij6tLERkSBENvcGwwGnHlyBTWHya7hCN5OO4yefJkuXv37iF/XnJcW1cPvjxSj8NnWqDTatDSrsf9c/ORGsM8LCJSpu5CJz7Zfxbfn5Xr0nEe/7QEv1o4krN8ZFdXjwEhOqZHkIkQolhKOdne4zhTTIOKCNHh2rFpuHZsGrp6DNAKAZ2WWTdEpFxyVCgaWrtcOkbp2fMYkRrNgJgUYUBMzmB0Q4qF6LQMiInIKTkJ4S615l114AwWjU1z44iIiC7GCIeIiDxu/qg0rD3sXBWKpjY9woN1rBZARB7FoJiIiDwuJjwILR3d6OoxOPyz7xdX46ZJ3CRFRJ7FoJiIiIbELZMz8c7OUw79TLfBiMY2PVLYZIGIPIxBMRERDYnhSZFobNOjuV15h7t3d1czl5iIhgSDYiIiGjJLp2djxdZKRY+tbGxDc4ceozNiPDwqIiIGxURENIQSIkMQEaJFVePgXe4MRomXN5/APbPyhmhkRKR2DIqJyGTni0Brfd/fW+tNXyNys+9Ny8abOwafLV6x9SRunZqFYB0vU0Q0NPhuozYMfMianS8Cqx8FViwynROt9abPVz/K84PcLjRIi4nZcXhlywkYjAO7qpbVnAcAFKZGD/XQiEjFGBSrCQMfsqVoCZBUCNSXAc9PM33Ul5m+VrTE26OjAHTNqFRMzU3A7z45jJMNbQCA2vOdeO6Lcqw9VIOlM3K8O0Ai8m0emOQTUg68S/e0yZMny927dw/586qeOQiuLwPCE01fa28wBT5LVwGRSfaPsfNFU5BkfmxrPVCyEpiyzHPjpqHRWm8KhtsbTH8PTwQe2K7svCBykr7HiJc2V6CxVY+0mFB8a+IwxEUEe3tY5E94XVIf8ySfOX4B+uKbhU8O+L8XQhRLKSfbO6zOI4Ml3xSZZDp5+gc+jgTEqx8Fdr008CQE+AZERA4L1mnwwNx8bw+D/BWvS+pUtMT0f25e3QT6JvlcWN1k+gTZZ16isFxif2Ycl9gDiXkVob3BdKMUnmj63JxqQ0Tki5j6pU7mST7ztcp87VI6yWcDg2I1cSbwscxDBoCbVgBCC3S3ue0kJB9QsrLvQvLAdtOH+UJTstLboyMiss5DwRGpE9Mn1MQy8Om/zGQr/6r/EoU0AtIwtOMmzzP/31vm5S1dxbw8IiLyPf0n+YC+ST4Xboi40U5tnNmQ0H8DFgCExQNC4/hGPSIiIndxxwZy8j/caEdu0T/4jUxyfCZQaIE71wARCfZnmomIiDzFmRVQ8n8eWt3kTDENrv9deE8HoG+7+A2IbzxEROQtLMlGdiidKeZGOxpc/w1YD++/eAOWMzPNRKQu7KRJnjRl2cVpErwukZOYPkGD4wYsInIF68gSkZ9g+gQREXkON0IRkZcxfYKIiLyPdWSJyE8wKCYiIiIi1WNQTEREnqO0kyY34xGRl7kUFAsh/iCEOCCE2CeE+EwIke6ugRERUQBQ0kLcsp18a31fIL36UQbGRDRkXNpoJ4SIllKe7/38YQBFUsr77P0cN9oREamIvTqy3IxHRB40JBvtzAFxrwgAQ1/KgoiI/Ntgm/EAzhYT0ZBwuU6xEOJxAHcAaAFwucsjIiKiwOFIneKejr7PpRFoawTeW8qaxkQ0JOymTwghNgBItfKtx6SUH1k87pcAQqWUv7VxnHsB3AsAWVlZkyorK50eNBER+QklqRGWjxFaQBpMjzN/zjQKInKB0vQJtzXvEEJkAVgtpRxt77HMKSYiUpHWeuD5aaZgGDAFxw9s7wtyzbPJSYXATSuAVxcAHU2m7wVFAI/sZ0BMRE5TGhS7lD4hhCiQUh7r/etiAGWuHI+IiFTIsp08AAiL7S5BYUM/HiJSJVfrFP9ZCHFICHEAwNUAHnHDmIiIKFD0r1McFHFxnWJzPWJzYKykpjERkQe4NFMspfyWuwZCREQByLJO8dhvAxt/B2iDTV/b+zpw4D99G+mAvsf235RnWcKNiMgDXK4+QUREZFP/1AhzEBwcAWx7rm/TnWUdY8vPl65iQExEQ8JtG+0cwY12REQqZW/THRGRmw1J8w4iIiIiokDAoJiIiIZG/0133EhHRD6EQTEREQ0Ny013D2w3fSQV9m2kIyLyIm60IyKioWG56Y4b6YjIxzAoJiKiodM/+I1MYkBMRD6B6RNEREREpHoMiomIiIhI9RgUExERke/a+eLF1UnMrcGJ3Iw5xUREROSbdr4IrH4U2PXSwNbfAPPRya0YFBMREZFvKlpiCojry0ydEIGLW4MTuRHTJ4iIiMg3RSaZZojNjV7MjV+WrmJrcHI7BsVEREREpHoMiomIiMg3sTU4DSEGxUREROSb2BqchhA32hEREZFvYmtwGkIMiomIiMh3sTU4DRGmTxARERGR6jEoJiIiIiLVY1BMRERERKrHoJiIiIiIVI9BMRERERGpHoNiIiIiIlI9BsVEREREpHoMiomIiIhI9RgUExEREZHqMSgmIiIiItUTUsqhf1Ih6gFUDvkT+6ZEAA3eHgT5HJ4XZA3PC7KG5wVZw/OiT7aUMsneg7wSFFMfIcRuKeVkb4+DfAvPC7KG5wVZw/OCrOF54TimTxARERGR6jEoJiIiIiLVY1Dsfcu9PQDySTwvyBqeF2QNzwuyhueFg5hTTERERESqx5liIiIiIlI9BsU+RAjxEyGEFEIkenss5H1CiCeEEGVCiANCiA+FELHeHhN5hxBivhDiiBCiXAjxC2+Ph7xPCJEphPhCCFEihDgshHjE22Mi3yGE0Aoh9gohVnl7LP6EQbGPEEJkArgaQJW3x0I+Yz2A0VLKsQCOAvill8dDXiCE0AJ4DsACAEUAbhVCFHl3VOQDegD8REpZBGAagAd5XpCFRwCUensQ/oZBse94GsDPADDJmwAAUsrPpJQ9vX/dDmCYN8dDXjMFQLmUskJKqQfwDoDFXh4TeZmU8qyUck/v5xdgCoAyvDsq8gVCiGEArgXwkrfH4m8YFPsAIcRiAKellPu9PRbyWXcDWOPtQZBXZAA4ZfH3ajD4IQtCiBwAEwDs8O5IyEf8DaZJNqO3B+JvdN4egFoIITYASLXyrccA/Aqm1AlSmcHOCynlR72PeQympdI3h3JsROT7hBCRAN4H8CMp5Xlvj4e8SwixCECdlLJYCDHX2+PxNwyKh4iU8kprXxdCjAGQC2C/EAIwLZHvEUJMkVLWDOEQyQtsnRdmQog7ASwCME+yfqJanQaQafH3Yb1fI5UTQgTBFBC/KaX8wNvjIZ8wE8D1QoiFAEIBRAsh3pBSfs/L4/ILrFPsY4QQJwFMllI2eHss5F1CiPkAngIwR0pZ7+3xkHcIIXQwbbScB1MwvAvAd6WUh706MPIqYZpFWQGgSUr5I2+Ph3xP70zxo1LKRd4ei79gTjGR73oWQBSA9UKIfUKIF7w9IBp6vZstfwhgHUybqf7LgJhgmhG8HcAVve8P+3pnB4nISZwpJiIiIiLV40wxEREREakeg2IiIiIiUj0GxURERESkegyKiYiIiEj1GBQTERERkeoxKCYiIiIi1WNQTERERESqx6CYiIiIiFTv/wOATZrrFSM5+wAAAABJRU5ErkJggg==\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "# let's do some plotting!\n", "xx = np.linspace(-5, 5, 200)[:, None]\n", "\n", "mu, var = model.predict_f(xx)\n", "\n", "fig, ax = plt.subplots(1, 1, figsize=(12, 6))\n", "ax.plot(xx, mu, 'C0')\n", "ax.plot(xx, mu + 2*np.sqrt(var), 'C0', lw=0.5)\n", "ax.plot(xx, mu - 2*np.sqrt(var), 'C0', lw=0.5)\n", "\n", "ax.plot(X, Y, 'C1x', mew=2)\n", "_ = ax.set_xlim(-5, 5)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Optimising the noise variances\n", "Here we'll optimise over both the noise variance and the variational parameters, applying natural gradients interleaved with the Adam optimiser. See [Natural gradients](natural_gradients.ipynb) for more details and explanation. " ] }, { "cell_type": "code", "execution_count": 17, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:58.218804Z", "start_time": "2019-04-05T16:22:57.828288Z" } }, "outputs": [], "source": [ "# Stop Adam from optimising the variational parameters\n", "model.q_mu.trainable = False\n", "model.q_sqrt.trainable = False\n", "\n", "# Create Adam tensor\n", "adam_tensor = gpflow.train.AdamOptimizer(learning_rate=0.1).make_optimize_tensor(model)" ] }, { "cell_type": "code", "execution_count": 18, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:59.302441Z", "start_time": "2019-04-05T16:22:58.220913Z" } }, "outputs": [], "source": [ "for i in notebook_range(200): # notebook_range() behaves like Python's built-in range here (only needed for continuous integration tests)\n", " session.run(natgrad_tensor)\n", " session.run(adam_tensor)\n", "\n", "# update the cache of the parameters in the current session\n", "model.anchor(session)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Plotting the fitted model\n", "\n", "Now that the noise variances have been estimated, we can see the final model fit.\n", "The predictive variance is higher on the left side of the plot, where we know that the data have different variance.\n", "We'll plot the known underlying function in green to see how effectively we've recovered the ground truth.\n", "We can also print the model to examine the estimated noise variances:" ] }, { "cell_type": "code", "execution_count": 19, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:59.570865Z", "start_time": "2019-04-05T16:22:59.303838Z" } }, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "# let's do some plotting!\n", "xx = np.linspace(-5, 5, 200)[:, None]\n", "\n", "mu, var = model.predict_f(xx)\n", "\n", "fig, ax = plt.subplots(1, 1, figsize=(12, 6))\n", "ax.plot(xx, mu, 'C0')\n", "ax.plot(xx, mu + 2*np.sqrt(var), 'C0', lw=0.5)\n", "ax.plot(xx, mu - 2*np.sqrt(var), 'C0', lw=0.5)\n", "\n", "ax.plot(X, Y, 'C1x', mew=2)\n", "ax.set_xlim(-5, 5)\n", "_ = ax.plot(xx, 2.5 * np.sin(6 * xx) + np.cos(3 * xx), 'C2--')" ] }, { "cell_type": "code", "execution_count": 20, "metadata": { "ExecuteTime": { "end_time": "2019-04-05T16:22:59.648913Z", "start_time": "2019-04-05T16:22:59.571983Z" } }, "outputs": [ { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
classpriortransformtrainableshapefixed_shapevalue
VGP/kern/lengthscalesParameterNone+veTrue()True0.3091906360603466
VGP/kern/varianceParameterNone+veTrue()True5.11779936937855
VGP/likelihood/likelihood_list/0/varianceParameterNone+veTrue()True0.01046763426205227
VGP/likelihood/likelihood_list/1/varianceParameterNone+veTrue()True0.5016463405422175
VGP/q_muParameterNone(none)False(100, 1)True[[1.141104975484783], [1.0567668719464667], [1...
VGP/q_sqrtParameterNoneLoTri->vecFalse(1, 100, 100)True[[[0.3053892368983257, 0.0, 0.0, 0.0, 0.0, 0.0...
\n", "
" ], "text/plain": [ "" ] }, "execution_count": 20, "metadata": {}, "output_type": "execute_result" } ], "source": [ "model" ] } ], "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": "3.6.8" } }, "nbformat": 4, "nbformat_minor": 2 }