diff --git a/docs/examples/classification.ipynb b/docs/examples/classification.ipynb deleted file mode 100644 index 90e264f1c..000000000 --- a/docs/examples/classification.ipynb +++ /dev/null @@ -1,779 +0,0 @@ -{ - "cells": [ - { - "cell_type": "markdown", - "id": "f78d3318", - "metadata": {}, - "source": [ - "# Classification\n", - "\n", - "In this notebook we demonstrate how to perform inference for Gaussian process models\n", - "with non-Gaussian likelihoods via maximum a posteriori (MAP) and Markov chain Monte\n", - "Carlo (MCMC). We focus on a classification task here and use\n", - "[BlackJax](https://github.com/blackjax-devs/blackjax/) for sampling." - ] - }, - { - "cell_type": "code", - "execution_count": 1, - "id": "1610bd93", - "metadata": {}, - "outputs": [], - "source": [ - "# Enable Float64 for more stable matrix inversions.\n", - "from jax.config import config\n", - "\n", - "config.update(\"jax_enable_x64\", True)\n", - "\n", - "from time import time\n", - "import blackjax\n", - "import jax\n", - "import jax.numpy as jnp\n", - "import jax.random as jr\n", - "import jax.scipy as jsp\n", - "import jax.tree_util as jtu\n", - "from jaxtyping import (\n", - " Array,\n", - " Float,\n", - " install_import_hook,\n", - ")\n", - "import matplotlib.pyplot as plt\n", - "import optax as ox\n", - "import tensorflow_probability.substrates.jax as tfp\n", - "from tqdm import trange\n", - "\n", - "with install_import_hook(\"gpjax\", \"beartype.beartype\"):\n", - " import gpjax as gpx\n", - "\n", - "tfd = tfp.distributions\n", - "identity_matrix = jnp.eye\n", - "key = jr.PRNGKey(123)\n", - "plt.style.use(\n", - " \"https://raw.githubusercontent.com/JaxGaussianProcesses/GPJax/main/docs/examples/gpjax.mplstyle\"\n", - ")\n", - "cols = plt.rcParams[\"axes.prop_cycle\"].by_key()[\"color\"]" - ] - }, - { - "cell_type": "markdown", - "id": "a96a36dc", - "metadata": {}, - "source": [ - "## Dataset\n", - "\n", - "With the necessary modules imported, we simulate a dataset\n", - "$\\mathcal{D} = (\\boldsymbol{x}, \\boldsymbol{y}) = \\{(x_i, y_i)\\}_{i=1}^{100}$ with inputs\n", - "$\\boldsymbol{x}$ sampled uniformly on $(-1., 1)$ and corresponding binary outputs\n", - "\n", - "$$\\boldsymbol{y} = 0.5 * \\text{sign}(\\cos(2 * + \\boldsymbol{\\epsilon})) + 0.5, \\quad \\boldsymbol{\\epsilon} \\sim \\mathcal{N} \\left(\\textbf{0}, \\textbf{I} * (0.05)^{2} \\right).$$\n", - "\n", - "We store our data $\\mathcal{D}$ as a GPJax `Dataset` and create test inputs for\n", - "later." - ] - }, - { - "cell_type": "code", - "execution_count": 2, - "id": "9abfffa3", - "metadata": {}, - "outputs": [ - { - "data": { - "text/plain": [ - "" - ] - }, - "execution_count": 2, - "metadata": {}, - "output_type": "execute_result" - }, - { - "data": { - "image/png": "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", - "text/plain": [ - "
" - ] - }, - "metadata": {}, - "output_type": "display_data" - } - ], - "source": [ - "key, subkey = jr.split(key)\n", - "x = jr.uniform(key, shape=(100, 1), minval=-1.0, maxval=1.0)\n", - "y = 0.5 * jnp.sign(jnp.cos(3 * x + jr.normal(subkey, shape=x.shape) * 0.05)) + 0.5\n", - "\n", - "D = gpx.Dataset(X=x, y=y)\n", - "\n", - "xtest = jnp.linspace(-1.0, 1.0, 500).reshape(-1, 1)\n", - "\n", - "fig, ax = plt.subplots()\n", - "ax.scatter(x, y)" - ] - }, - { - "cell_type": "markdown", - "id": "9db6c8d8", - "metadata": {}, - "source": [ - "## MAP inference\n", - "\n", - "We begin by defining a Gaussian process prior with a radial basis function (RBF)\n", - "kernel, chosen for the purpose of exposition. Since our observations are binary, we\n", - "choose a Bernoulli likelihood with a probit link function." - ] - }, - { - "cell_type": "code", - "execution_count": 3, - "id": "1ac7588f", - "metadata": {}, - "outputs": [], - "source": [ - "kernel = gpx.RBF()\n", - "meanf = gpx.Constant()\n", - "prior = gpx.Prior(mean_function=meanf, kernel=kernel)\n", - "likelihood = gpx.Bernoulli(num_datapoints=D.n)" - ] - }, - { - "cell_type": "markdown", - "id": "8240564e", - "metadata": {}, - "source": [ - "We construct the posterior through the product of our prior and likelihood." - ] - }, - { - "cell_type": "code", - "execution_count": 4, - "id": "335b6ead", - "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n" - ] - } - ], - "source": [ - "posterior = prior * likelihood\n", - "print(type(posterior))" - ] - }, - { - "cell_type": "markdown", - "id": "d5553eee", - "metadata": {}, - "source": [ - "Whilst the latent function is Gaussian, the posterior distribution is non-Gaussian\n", - "since our generative model first samples the latent GP and propagates these samples\n", - "through the likelihood function's inverse link function. This step prevents us from\n", - "being able to analytically integrate the latent function's values out of our\n", - "posterior, and we must instead adopt alternative inference techniques. We begin with\n", - "maximum a posteriori (MAP) estimation, a fast inference procedure to obtain point\n", - "estimates for the latent function and the kernel's hyperparameters by maximising the\n", - "marginal log-likelihood." - ] - }, - { - "cell_type": "markdown", - "id": "91485478", - "metadata": {}, - "source": [ - "We can obtain a MAP estimate by optimising the log-posterior density with\n", - "Optax's optimisers." - ] - }, - { - "cell_type": "code", - "execution_count": 5, - "id": "0192a42a", - "metadata": {}, - "outputs": [ - { - "data": { - "application/vnd.jupyter.widget-view+json": { - "model_id": "0e78f1d26f0a48c79e5db78bfa48948d", - "version_major": 2, - "version_minor": 0 - }, - "text/plain": [ - " 0%| | 0/1000 [00:00" - ] - }, - "execution_count": 6, - "metadata": {}, - "output_type": "execute_result" - }, - { - "data": { - "image/png": "iVBORw0KGgoAAAANSUhEUgAAAqEAAAE5CAYAAACgf/ntAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAABJ0AAASdAHeZh94AACwP0lEQVR4nOydd3wcZ53/3zOzfVfSatXlvnK3E9uy05yeyKEEEhNswsERqq0j1ASwyfGDO+6AIB8dLsQO4YAr4FiENFqkQHqV1yXutuQmN1la7apsn5nfH7s72pVW7l7J9vN+vfzy7swzzzzz7Gjns9/nWyRd13UEAoFAIBAIBII8Io/0AAQCgUAgEAgElx5ChAoEAoFAIBAI8o4QoQKBQCAQCASCvCNEqEAgEAgEAoEg7wgRKhAIBAKBQCDIO0KECgQCgUAgEAjyjhChAoFAIBAIBIK8I0SoQCAQCAQCgSDvCBEqEAgEAoFAIMg7QoQKBAKBQCAQCPKOEKFAJBJh69atRCKRkR6KQCAQCAQCwSXBBStC29raaG5uJhAIZL0/E1pbW5k9ezatra3ncIS5CYfDhMPh836eCwUxH9mI+RhAzEU2Yj6yEfMxgJiLbMR8ZDOa58OUrxMFAgEee+wxVq9ezfr160/pmJUrVwJJgenxeGhoaMDtdgPg8/lYunSp0dbr9dLU1HTOxy0QCAQCgUAgOPfkRYQ2NzfT1tZGIBAwLJcno76+Pkt01tfXM3/+/Cxr5erVq/F4PHi9Xmpra8/DyAUCgUAgEAgE54O8LMfX1dWxfPlyvF7vKbUPBAKGcE2zcuXKIUvudXV1LFmyRAhQgUAgEAgEgguMvC3Hny5+v5+2tjZDYHo8HoAsYRoIBPD5fPj9fhYsWGBYTU9ER0cHx48fz9q2Z88eIBmgdL79JkTwUzZiPrIR8zGAmItsxHxkI+ZjADEX2Yj5yCbf82G320+57agUoW63m+7u7qxtaQtoXV2dsW3t2rXU19fj9XpZtmwZ9fX1Wftz8dBDD/HNb37z3A9aIBAIBAKBQHDKSLqu6/k6WWNjIytXrjyjKPT58+dTV1dHQ0NDzv3Nzc0sXbqUvXv3ntAiOpwldPHixWzZsoVZs2ad9thOh7Sl9XR+KVzMiPnIRszHAGIushHzkY2YjwFOdS66u7sJBoNompaPYY0YqqoCoCjKCI9kdHAu50OWZYqKiiguLj7rvmCUWkIHs3LlShYsWDCsAAVYsGABgUCAlpaWE1pDy8vLKS8vPx/DFAgEAoFgVNLd3c3Ro0eRZRmTyYQkSSM9pPOGLF+w2SfPC+dqPnRdJxqNcvToUYBzIkRHvQhtbGykpKRkiAAtLi5m3bp1huBMWz9PNfpeIBAIBIJLhWAwiCzL1NTUYDKN+kf/WZG29AoxmuRczkcikaC1tZVgMHhOROio/oSam5vx+/2sWLEiaxsk84JmRtunA5ZEpLxAIBAIBNlomobJZLroBajg/JK+h86VS0feRajf7x+yzefzUV9fP2TbunXr8Hq9NDc309zczKpVq4wo+bq6uiwR2tDQcFppoAQCgUAguJS4mJfgBfnjXN5HeflJ5PP5aG5uZu3atQQCAVauXElJSYlh4Ryc/zMQCHDrrbcSCARYs2ZNVl/pOKqGhgZWrVoFQFdXFzU1NVkWU4HgQqb1Jw/Ru20HM77zr1hTP7wEAoHgRITbD3Hk6T8x6Z8+JQSn4IIgr9Hxo5WtW7cye/ZsER0/Aoj5yCYcDuN/+VU2ffCjACx87k8UTJuCcgnOj7g3shHzkY2YjwHSc7H9018g0R9i6j9/Bad3IuaiIqNN2mXtTFYL93f1MaHEdcb7843wCc3mXM/H2dxLgxGfkEAwytjzjW8h22xM/cYDhA+2Ez50ZKSHJBAIRjmhtn0cfebPxLu7CbXuJXzw0Dnp9+cv7GLJmpd4cfexnPtf3H2MJWte4ucv7Don5xvMqlWrWLVqFWvWrDFeZ9Lc3Mz8+fOHrJqOJgYHTC9atGhUjzefCBEqEIwioh3H6d+1m+Ir5lM8vxYkiYO/+V8iHcdPfrBAILhkOfy/vwNdp6zuFo48/Ud2ffd7aLHYWfW5v6uPX77aSlzV+FKjb4gQfXH3Mb7U6COuavzy1Vb2d/Wd1fkyCQQC1NTUUFdXx4oVK1i+fDkrVqygtraW+fPnG+3q6upOWqRmJGlraxsiOFeuXDmqx5xPhAgVCEYRwTdbAHDWTEJx2Am2bGDvf67m8O//MMIjEwgEoxn/Cy9jKiqiqHYuoX0H6PzbC8TOMmXhhBIX319Si1mRhwjRTAFqVmS+v6T2nC7JL1u2jCVLlgzJeFNXV8eCBQuGBDOPVnLlNx8cWH0pI3I1CASjiHgggOJy4Zw+DUmWKb3petoeepijT/6RScs/iSQqgAgEgkHE/H76d+zEc81VmFxOiubMpnfLVoK+Tdjeueis+r5hSgXfX1JrCM4vNfpYPGcsT2xqzxKgN0ypOEdXk6SxsZGmpqac+xYtWsSyZctYvXq1sS2zbUtLCytWrGDy5MlGgLPX68Xv97N+/XrjuObmZpqamqipqWH9+vU0NDTQ0tJCfX099fX1uN1uVq9ezfr1642Kj16v1zjX/Pnz8Xq9XHvttYwdO5a2tjZaW1uN/hsbG2lpaTF8KNPWz2XLlmVVgEwHb7vdbgKBAHV1ddTW1uLz+Yy2NTU1AKxbt844/4mu7UJBiFCBYBRR+f7F2CdMMP4wze4iiubOIbB+A/2tbbimThnR8QkEgtFHtP0w1jHVOKdMRrHbKZw1E4DAhk1UnKUIhaFCdJ3vAMB5E6Bp0eYZJjNIWqwFAgGjUI3H42H58uUA3HLLLbzjHe+gtbXVEGlLliwBMJbG29raqK+vN8qINzc3s3LlSlavXs2SJUtoamqiqanJGMOSJUvw+/1ZYvfuu+9mxYoVhoh1u90sXbqUNWvWsHz5cpYsWcJbb72VlQ0ofVxXV5cxjmXLlrF+/Xpj//z581m3bh21tbXcfffdrF271hCsTU1NNDY2smTJkmGv7UJCLMcLBKMINRRCi8UwORzGtvK6W0DTOPLkH0dwZAKBYLTimjWDmf/5I0pvuRFJkrBWJkVh5NC5CU6CpBBdPGds1rbFc8aecwEKA1HXaTE6mPT2tAAFDEth+ni/34/P56Ouro5ly5Yxf/58Vq5cyQc+8AEgaaV0u900NjbS2NhIW1sbLS0tRh9pN4C0wANYvnw5zc3NQwRwa2srbrcbn8+Hx+MxhO1wZI67sbGRBQsWZO1fsGBBlkUzc3/62oBhr+1CQohQgWCUEDlylLbv/oC+bduRrRZju2fh1UiKQsdfm9ASiREcoUAgGI2ooRBaOITJlfTJtJaXARA92nHOvjNe3H2MJza1Z217YlP7sFHzZ0vaGpmLpqYmw+p5MrxeL93d3TQ0NBg5yCGZX3zBggUsWbKEJUuWsHz58ixrZKaozeQDH/gAa9as4bHHHjPGUF9fz8qVK3G73cMeB8ll98GkLaInIlO0nsq1XUgIESoQjBJ6d+zkyG8fI7z/IFJGaT2Ty0n5u27DNX0aiWDPCI5QIBCMNiKHj7D9iyvp2bgZxW4DQLHbGXfPhym9+Xr0WPyszzE4CGlp7ficwUrnkkceecSolphJY2MjPp9viO9jpvWxra0Nj8dDbW0tDz74IG1tbdTV1bF69Wpjeb2+vj5n32kGp1VKs3LlSh588MGsY1paWmhoaMDr9dLV1UUgEDD6KikpMYRm2tKa2Xd9fX2WBTbd7u677855/rQVFhj22i4khE+oQDBKCO3dD4C1vHRItZOaz99LuP0Q8WAQS8mF90UjEAjOD50vvkLnX5uwVlciW63G9qrF70VXE2jxGApnntA/VxT8DVMquH5KeVaw0rn2DXW73bS2trJy5cohFsTBy90lJSVcccUVNDY24vf7aWlp4a233jL2pZfeASOq3uv1snr1aurr642UT3V1dUaQUFrEDk6l5PV6s9JC1dXVsXbtWtasWYPH42HRokU0NDQYS+bLly9n2bJlrFq1yui/qanJcBeora2loaGBlStXUlNTQ2trKw0NDUZg0tq1awEMwdzS0kJLSwt1dXXDXtuFhKiYhKiYNJKI+Rhg+9f/jb0PrWHqg/9G2ZULhuwPtx/C5C6k5JqrL4koeXFvZCPmIxsxH0k23ftFDq1tZNr3vkPpvLnG9kQoRKKvj6J5c7GVlZ5RlZv9XX0sWfPSsFHwgwVq4/LrR0XlJFExKRtRMUkgEJyU/rZ9oChYK8pz7j/6zJ9pufujhNvPXbCBQCC4cNF1na6XXsFaXYWtqiprX/v/rsV3z6eIHT3zpfIJJS4+sbBm2Cj4dNS8WZH5xMKaUSFABRcWQoQKBKOE0N69WEpLkG3WnPtt1VVokQhHnhJR8gKBAEJte4kcPoJz+lTkQRZhxWZDjyfO+kfrp2+cSuPy64ddar9hSgWNy6/n0zdOPavzCC5NhAgVCEYJZYtuxX3tNUhmc879pTdeD4rC0af/JKLkBQIBXS+9CoBzsjcrowZgrKiEDx0+6/OczMIpLKCCM0WIUIFglDCp/pOUv/sdyMOIUHNRIe55cwn6NtK/J3f+PIFAcOngWXgV4z9+D85ZM4cEM1rKSgGIHD57ESoQnC+ECBUIRglaIoGuqicMOip/x62g6xz6XeOwbQQCwaWBtbKCitvfiTUlODOxeIoBiB47nu9hCQSnjBChAsEowP/GW/ju+RS9b289oQgtue5aTAUuet7eghqJ5HGEAoFgNBHt7CJy6Ahqfz9KRoW1NKbCQgDiwR50Vc338ASCU0KIUIFgFBDet5+et7eix0+cWFq2mJnz8E+Y8KmPEuvy52l0AoFgtLH3oTW8fONthA4cRLZYhuw3FxYw/d++zpgPvE+IUMGoRYhQgWAUEDnWAYC5uPikbW2VlWixOJFjHeLhIhBcohz7819RXC6c3kk590uKQsGMaZiL3egJ8T0hGJ2IikkCwSggaohQ96m17+xk57cbmPmtf6H8trqTHyAAknkVY8c7iXZ2kujpRbZYcNfOBaD7zRb6du4CSUJxOjG5nJgKCrCPG4t9TPXIDlxwXtE1jWjHcWyVyTREaihMPBjEWl42KgtD9GzZRv+uPZTdehOmggKGk5jxYC9qOETRnDl5HZ9AcKoIESoQjAIiqYTS5lOs/WspKSHUto9dD36P0ltuQjaJP+U0aihM6MBBosc6KL3xOrR4nCNPPsOe7/2YcHs7WnjAl7ZwzmXMeehH6AmVfWt+yZE/PDWkv8o738OMb/4/JJOJXd/9PtFjHRRMn0rBjGm4pk3FOWUyJudQn7yLETUcJt4dQA2HMblc2WmANA2z243icg6J1B5N6KpK77YddL/ZQvebLQQ3vU14/wG0WIwb3nwRSZLofOFltn75ARS7HdeMaRRfdQXVi99LUe1cpFFQhefQY78HwH3FAhSHHTUWy9lu93f/g3h3kIp33ZbP4Z01bW1tNDQ0sGbNGpYvX05NTQ2QLNc5f/58li9ffk7O09zcTH19PQ0NDSxZsgSARYsWsXTp0lM6RyAQMEpmnu6xgiTiySUQjAKixzpQXE4Uu+2U2jsnTqD8tlvp+Gsz+3/5GyYt/8R5HuHoQYvHiRw5irWiHMVqJXzoMDv/7UFC+/YTOnCQWEcyGlgymbj66Ub0RILQ3v0kentxTZ6MpdSDqbAQ2W7DUlJC347dIEu4r5iPwzsJXdfQIhG0UAQ1EsY2pprAhk3IZhM9b2+h5+2tHG96bmBAksTM7/47Y+9eguKwc/xvz+OaMhn7+HGjQrCcLvFgkFiXH6d3Erqmsbvh+xz+/ZNEj3ei9vUZ7aqX3sWUFfcBsLH+cwR9GwGQrVYsJR6slRVMunc5le95F7LZTM/bW7GNqTaitvOFGg4TWL8B19QpmIvd9O3ew8s3vcPYb60oxzVzOtbSEgK+TShmE4m+XkpvuZHo0WP07dxN0LeRfT9/hKJ5c7j66UaUESwVqsViHHrscawVFRTOueyEgt9UWEj4YPsFtxzv9XoNEdrQ0JAl9NJ13s+F0Kurq6O2tjZr28qVK0+pHGVbWxuNjY2sWLHitI8VDCBEqEAwCpj0T5+ie70PTmPpb+I/LcP/6hvs+nYDJddfS+GMaedxhOcGXddB04wlztC+/cQ6u0j0h1D7+0n09ZHoD1F85XzMKV+3bV/7VyKHjxA5dJjwocNJ1wVdp/bXj+CaMpnQgYMcbvwDpsJCrOVlOGsmYfF4sJSVEj58BJPdjnv+XDxXX4FkMiX/KcoQgegYP274casqWiLBjH/7BolQmEh7O6H9BwgfbCdy+AioKt3rN6DHYrT8w0cBUJzOpMV09iwKL5vFmA+8f9RZTMOHDtP5t+fp3b6T3m076Nu5m2hHB67p05jz8E9Q+0NEO46jxeM4Jo7HXFiI4nQim01YKyvo2bwFXdcpnDUTS2kpaihEoreXRE8v/Xva6N22HVtlBZLVyhvveT9aNIqpqAindyLOGi/OqZMZ/5EPYS0vO2fXFNp/gOCmzQQ3vo3/1dcJbtyEHk8wecV9lN5wHYlQmKol78NWXUXhjGlYSkuRrVZkizl5b0gSjokTKLnuWvR4AjUSpmfLdvyvvIq1opyAbxPWinJ6t++k9MbrMBcWnLOxnwqRo8ewlJZQeNkszEWFJ2xrKixEi8ZIhEI592/753+hZ8vW8zHMYSmcPYuZ3/nmSdtlCs9MFixYwLp1686ZtdEzaPWpru7U3JsaGhoMC+3pHisYQIhQgWAU4LnmKmSrhcRpiFBzYQGTv/JFdnzj3/Hd8ykWPvsUllMIbDqX6KpK5PARQvsPENp/EElRGLP0fWiJBPsf+S+O/OEpEr19JPr7UUMh1FCY6vffyfR/+Rq6prHp3i/Q/UbLkH69X/gMnve8Cz0W49C6x4l3+TEVFWEtLcFxxXzMxcWEDx1GSySQZJm5jz6EubAI2WJGtliQzOZzuiQsKQqKooDVisnlxFZeaviS6qqKGo2iRaIk+voY/7GPEGo/RPTwYXp37iKwfgPIMsVXX4ml2E3vtu3sf/Q3FF42k8JZMymYNQPHxAnnxWqq63rSmre7lf7de+jf00rf7j3MeejH6JpG599e4O0vJi05ssWCbUwVnoVXYx8/jlDbXmSLhYrb30XV4vcimcxIJsUQ8JnjdU6aaPzA0FU1KdrjCbR4jHhPD2ooTMXt7yTacZzY8U762/YR3LAJgKLauTj7JxDed4BNn70Pp3cS9vFjsZSUJC2q5eWM/YelQPJHS+/2najhMKGuLhI9vWj+bmxjqxl/zz+iRaO8fd8Kul54GQDF4aBw9iyck2swFRURD/ag2O1M/NTHhi0KkUaSJCSLGdlipmThVZQsvAo1HCEeCNC3azfbHvgGisOB9zP1TLp3Wd6so9aKcmb9x3eIdXSgWHOX+E2TFqnxLj+4hwrWni1b8b/y+nkZ5/mipaWFu+++GxhYTq+vr8ftdrN69WrWr19v7Hvuueeoqalh/fr1hkXV5/OxevVqw6La0tLCokWLAPD5fCxbtoy6ujoaGhqAAdeA+fPnEwgEqK2tJRAI0NLSQltbsmhIWnxmHtvY2GhYRpuamoCkFdfr9fLII4/Q0tJCU1PTkPENJnNMadHb1NREQ0MDPp+PtrY2mpqajHOkrz2z7wcffBC3282aNWvweDy0tbXR2trK6tWrhz3HunXrsvo8XwgRKhCMMLquo0Yi6AkVyXLih8pgSq69hnEf+QciR48ROXoMU0HBefMPVaNR46F37Nlmdn7zQfr3tKJnlBB1TvbimDAeXVPp3b6D0MF2FLsdk8uFpbQExWZFcTgI+DYCOkXz5+GYODFpibJakCwWFJsVa2Ul4dY2JFlm5rf/FcXhQLHbkBQTktmEbDIjmU2jwvdQUhRMDgc4HFg8xcYythaNkghHiB45SvhgO307d6PYrHS9/CodzzbT8ZdnjT4Uu52yRbdw+U++j2Qx0/XCS4T27kdxuTAVODG5XCQUBVt1FfYpkwEIbt5CrKvLsDwmevuI+bupuvM9OCd7SfT08vyCa1H7s61gssVCR/PfsHg86JLMpM/9E84JE7CNHYNityFbklbB054HSQJFMazcyiBN5v1MfVKcxuJosRixQIDwgXbi3QGCwR7CBw6g2G0EfBvxvzogjMwlHormXY4kybSvbaTtx/855NwFs5KCXovGcF8xH+fkGhwTxuP0TkJxJt1cTiY6TwXFbkOxV2IudjP+4//I4cYn2PWdVex/9FdM/soXGfeRD51X/+y+PW3IJoV4t9+oiHQi0iI0NowILZw965yP8WSc7jnXrFmD2+0mEAjQ2trK3XffbSyB19XVsWTJEkOEpa2abW1tfPrTn6a1tRVIirKVK1fS0NDA0qVLje2QFFtpamtrufvuu+nq6gKSPp+LFi1i/fr1hohbt24dq1ev5q233qKkpCRrOT7z2CVLluD3+7OEXHrsbW1t1NfXDxlfWhRmkh7T2rVrDWG8fv16Vq5caYy9qakJn89HbW1tzr6/+tWv8vDDD9PQ0GBcy9KlSw2f21znaGpqorGx0fCVPV8IESoQjDDhg+08P/9aKt/7bqo/+dHTPn78xz5CtLOL0N59qP0hAi0+xn/0w2dlmUn09tGzbTs9GzcT3LiZ4KbN9Lfu5dq//xk0ndDe/cQ6OymcMztpsfJ4sJaVYq2uIh4MIikK1e9fzJil7wdZRpIlkGSQQEr9jyRR9d73JLfJMkhSlnUtGo0CYD2JtWc0Iskyit2OYrdj9RRTOGtGSpjGKF90C56rr0z6sO4/QLj9EJFDh9FiMbrXb0A2Kez9+S/oevHlIf1WffiDWL74GQA21X+Ovl27h7TRNY2ym29AiydwTZ+Kxe3GUl6OtbIC+9gx2CorDLFpLS+jaPaM8z4faSRFQbErKHYb5qJCnBPGJ8es69irq3DX1qLF48QDQRI9PcR7etCiMfp27EZHx1JawvhPfhTZYkE3m1GcDuxlZdjKStETCRSblfJbbz7nlvDBKDYbY+9eSuXt7+bQY7/n6FPPsPXL/8yxP/2VBf/7Xznzdp4t/jfe4o3FH6Dy3e+k6v2Lkz98ToK5qAiAeCCQc/+pLIuPNEuWLDmpn2XarzMtmH7/+9/jdrtpbExWlvP7/bS0tPDYY48N6Wuw9dHtdhtCMt0+3eZkLgCZx6bbr1y5kkBq/tP9NDY25hzfiViwYEHWeTJdAdxuN21tbdTW1ubsO20dTgtTn8+Hx+PJEuODz+H1evH7z38uaiFCBYIRJtblT/lJms44HYy1tIR4MMi+hx/hyB+eZvd3v0fJTTdQftutFMyYjrNmEubCAUuIrqqooTCxri7C7YcItx/Gc82V2MZUE+vs4m+zF4CuG+0tpaUUzZuD/40WrJ5i7OPHMefnPzF86c6FleliJylMbSkBVoR97BhKrluYFKfxOHo8jhZPoIYjVN11J57rFqKGQmihEGokSizUj7WynODmLUgSlNx4He6rr8BktyPbbSgOB6aCAmwV5ajhCLLZzPRvfC1pOTabR3WQVHLp22KIN8swqcqcEydQsvBqYHT8SDG5nEz4xD1U3vke2v/3d7gmewlu2oytspLOl16h7KYbsFVXnfV5jv/9RTZ84p+QZIWCy2efsg9txe3vpODy2RTNnkXHWY9i9DLYN7Orq4v58+dnWfGWL1/OqlWrhvU1zUVgGPE+mLQVMhcf+MAHDGtuWsR2dXWxYMGCIeM7Eac67lx9f+pTnwIw3Bbq6+upqanJEsync45ziRChAsEIE+/uBkBxOc7KemMuKmLcPR9GcRXgf/kVjj39J449/Scg+TCa1fAtJFnG9/F6ut94a8jxkz73aSreUYcWi1O+6BaUwkLs48ZSMKUGS1lZ0npmtY5qMXMhIsly0s0hQ0zlqgU+WHQ5JkzIzwAFJ8Va4qHm8/eS6A8ROXKM7rd87Pj6v4Ek4Z4/j9Kbrqfk+mspmDEdS8mppWED8L/+ZtK3+slnUOx2vJ+tx3PlglNe8pfNZmRZQUvEwXRhrSicqgDM1Xb58uW84x3vyNqWXloevOSd9uvM1deSJUt48MEHs1IxpfspKSkxRFxLS4vhKzqYlStXMn/+fGOZG5JiMO2HOnh8Z8twfafHmbaKdnV1EQgEhj1vIBA4rc/gTBEiVCAYYdLlN00u11n3ZS4sZMLH/pFxH7qb8MGDBDduJnz4KPbxYwls2IgkK9jGVFN89ZVJC6bdgaXYjaXEg2PCeOLBHmSLhUmf/TSKzToqE3ULBKMVk9OByenAXFSI9wufoeulV+h5ewuBFh97vvdjTIUFXP/q31FsVo43/52jT/8p+cNOkkCWUSMRCmfNZGL9J9FiMXZ9ZxX+V16naN4cxty9hMLZs045jRtAIhQi4NuAFo/B/Lnn78LPMW1tbYZYbGhooL6+Pqel0efz0dzcjMfjoba21ggQ8nq9/PznP6e+vt4IQKqrq8Pr9bJu3TpWrlzJFVdcYSw3r1692hCRTU1N+P1+w7r53HPPGUIy3Q8khe6yZctYtWoVdXV1+Hy+Icemx1JXV5cVOe/1elm9evWQ8eXC5/Oxdu1aIOnfmfl/XV2dcb500FSuvm+55RY8Hg/r1q0zgpMWLVpEQ0ODcfzgc7S0tNDS0pIzjdW5RNL1jDW3S5StW7cye/ZstmzZwqxZ59dROxwOA2AfwTxzowkxH7Bv9aNs++d/YfLKLxGZt4DKQtuQJcYjwRBVRQ6OBJNBJunXVUWOrP2D26a3pSOWdVUb6FSWONYXBUWh2u0c0k/6fUdvhDljPSc8T2abNJva/ca2Te1+ygtsQ/pOX8vg9pC0/B3tiTChrGjYOcjFcOPJNZ8n41TOleZE/eWa18zjcs1FeYEta38sFgdgQllR1vHpNoPnePA5cn0Gg8eXa5y5tp/ouga3G+4+HTwXg8c93Hyl8diSP5CsVquxf/Dnnj5f+t463es7EScaZ7pPXVVRe3qwH9hL7/ad9IUiFN1+OxWeAo488TSHfvvYkH6Lr76SyV/+InoiTmDD25hcDgqmT6NTMlNd7Mp5rv3Hg0O+N44EQ8SPHePgpz9D4e23U/UvK5Ak6ZLIY6lpye85WazaAOd+PtLW43NxLwlLqEAwwsRSy/FvHI+w7pktfPaGyVxVU2Hsb/Qd4MlN7dwxZwxPbGwHdK6eVMqb+/zcVzcdgB827+DOOWPxlrn4YfMOrpzo4c19fu6cM5YlteOTaXUGWTUbfQf4w8aDSBLcXzeDtuN9PLmpnfvqplM7zoPvoJ/vNW1D12Ght5RrJ5cb5wGMMT2+4SAAd80bx9La5BLxT/++g1fbOlnoLaWyyG60+cptM42+v9+0HdD50qKZvLKnw2j/uZuT17SxPcDPXtzDnXPGGdc1pdzFzmM9SJLE/XUzqB2XLTR9B/1879lt6JDV1+A26etYUjt+2M9l8PzkOtcPmrejajqyBF9aNHNIm1znS/Z7AB2QJYmrJpbw+t5O45rSc5GK3eLqSaW8sa8LTdORJFg8dxxPbGxH05PvZUmiqshOe3cIWYL3zR2Pt8zFD5q3J9sAU8sL2XGsJ+szyHV/PbXpUNY4c20/0XVl3j+ZfQ++T9P9DL4P2o73DZnzdBtNT85zmvdeVs37Lh9Do+8AT2w6iKol7Snpzz197snlLnYc7WF6ZSF7OvqM86fHP/hv5VRIf4aZ92HmtjvnjOXJTe3GZ/alRTNh0kx++OxW5Of3ccesSu5Y8j46r7iWX76wk4keJ/u7+vj49VOYPKEcSZKQ7Q7KbrkB2WRKjXXzoHsoOU9XTSzhzX1dWd8b6XvTEgnxSeCt7QeZ1dHL5IoT5xUVCPJN3iyhgUCAxx57LCuP18lYtWqVEaHV2trKAw88kOU4e7L9p4qwhI4cYj4geryTbY8/ww9ea6erwINJlrgv9WA7Egyx4vGNJDQt57FK6qmsanrW6zQmWWbVXXNzWp++8vgGo60sJYNDVE3HJMvcMWcMT25qz+pLlkDTyXmeNHfNG0exw8Kjr7QO2QegSHDn3HFZfUtAZk+fXFiDx2Xlh83bSQxzXelrzxSHSbGyjcxmn7y2hrrpA4EhaeGR0LRh5ybX/OQ6V1qADjeeXOf78qLp/EfT9pxzBwNzfDZIUnJOh+tHkeD+lGDOdX+lx/m9ph1DtqfnK9d1pdubZJn76qZTVWgz+s78DNP9HOmJZM2hIgFIqPrAnKfFXK75UmT44k1T+dHzu4fsX1I7nic2tuf8uzHJMh+9ehK/fn3vsNd3InLdGx+72suvXm874ecKA5+JIkt87Bovv35t6BgyRTyc+j2U/t4AjHmVNI3P/O67tI6diu3b3+Sm6dVMmZwdxHMxIiyh2YxmS2hePqHm5mYee+yx03J0XbVqFZB0DF6+fDl33303S5cuPeX9AsGFgrmokPGzp/Oxm2ZgkiUSms4Pm3fgO+inqsjBHXPGZLXPtAapmm5Y4tLv0xhiIMdDtarIwf11MwxxoOnJNDmKLJHQNB7fcHDIQy4tQO+vm2FYQwfz+IaD/NeruQWoLIGqY/StyBKylC1AAX71ehs/MATowHWZZJm75o3LEjQ/aN6O76DfEIWDNcCvX9uL72DS72vww3y4uck1P7nOlSngFTkp4NOf23DnmzPWk9XvYE4kQAfHrA0Xwqbrw/eT/gxOdH/dMWcMc8Z6cm7PJUDT13Vf3XRMskxC0/hh8w6O9ES4r266MTfpz/y+uukc6Ynww+YdxjZFllD15EVl3svpeyXXdN0+qxpV08llR3liYztXTMztjnHFRE9OAXqi+yGTXPfGr15v4845Y4eMU5IGflhkCtA754w1BGj6vs6cuzO5hxKpe/T7TduMe1OXZeKKGWs8xmVj3MPedwLBSJFXn9B0BYHBualyUVxczPr167OUdua2k+0/HYQldOQQ8wE9W7bRu2MHlmIPGwJxfvbiHhIp0XXT1HKe39WR9cBUZAld108oWHJZVHIxWFANtkqml4QzH6A3T60YMqaTWfBkCS4bU8ym9m5j25yxxWw5HMgSczDQz2DrbKabwGARmHlc+iH/1KZDxsM7cx5PdW5yzU+uc92fsj5lioWTnS+XJfVESIAsQ6ZLryJLaJo+RMQPx13zxlFTVjDsONOc6vZTsdrdNLWcv+86lmU1zLx/0n1kzl+u6xp8X8LQz+JMOZ37IZNc94auD/37geQPPFnXQIfLxxSx/XAQVdMwyRKfvWkqc6rdbGr389Dzu4jrOrLJzPXTKnl+T+dZ30Mff/zHuCrLmfJ/jyKbTcIn9BJkNFtCR6UI9fl8zJ8/n+7u7qzl9ZqaGurr66mrqzvh/swKBoPp6Ojg+PHjWdv27NnD4sWLaWlpYebMmWd0badKJBIBwGY79QjHixkxH/DmovcS7+xkxk++h2a38fbhHh5+dT+JLKumxLtnVfGnrUeyttsi/VR2Haa86wgF/UHs0RCOaIhyRcMi6egJFV1NpP5XAZKJ42U5mTRelkho0JfQ0JHQpeQ/TZLRZBlNktFlGU1W0CUZNbVNk2V0WaHC7eBQbww1vS3VRpeSxyTbycZ+TUr3lTyHnkpQr0sSKgPbBv4ly0Peflk1NWUFRkWeVn+YdZsPE5dkNCl5Hk2WkRSFj15Tw+xxxWw51s/q1/cTI9mPLsvJh/4Nk5k71n3Kn8/G9gA/fXFPDncA+Ox1NcypcKLFYry9v5PfvLIHKR5H0RIoagKrpnLnjDImFlrQozG0eAwtGkOLxTnS1UNL6zHkRLKtrOtIuo6ka8n/Sb5H15P70JHRqXBZOdYbRZMkdCQ0SYLUfCb/z54/VVbwVhQxbawH2WzmSH+CF/Z3E5dMqIqCKivoJjPzvWW8frCHmCSjKgqanJzr66dX8bc9XcSQ0BQFTVKQTCY+feMU5lQXgq4nA950DV3TeLs9wKOvtKKpKrKmIesqZl1LirCEiqwlt5vRWDyrEm+xDT2hsrejh6ath9DVgTaypqLo2sBrTUXWdTQkNFlK3WcyyArTq91sPdZHXJJJmMzETWYSioWY2UzCZCGumEmYzMTMFhKK+bTvBy0aJdHXj9rbSzzYQ6Knh337j/H6ln1YI2Hs0RC2aAh7NIw5HsOkJjCpyc/WpMaHtVyf8JyShCYrmCxmrJ5iLKUl2CdOoGhBLYVzL2fTkd6c92YakyzxTy1/oKjQQen3v4PJamXSpElnMJILCyFCsznX87F37140TWPs2NwrYqdjVBqVgUnptAnDVTI42f4T8dBDD/HNb47+KhGCS4dEIIDiciLJycChy6oLuX5yGX/fNZBe+vrJZdw1Zwy90QR/39VB5fF2rnz7JcYd3Tvswy12GmM44+RQ7TDlTI89DbQXYHBtoLuGaas+AZtSrz+VsV0HdEVBazSzIV0DXVGSRQJMSqoueuq1nHyvJ1TMiTif6AkRCceS4lJTUVQVk6ai/o+KL+McH8419uegLcd2gDNNfHLaMmInHMl4e32uNi/AHbm2Pw0fybFZ/S1Z157Jx05xWFoz7Ml4f+spHpeTDXDdaTRXZRnMFqQ/2XnbYkmmSrJYkqJaS2WTSKho4TCJvn70WO6/qCvPZswnQdZ1ZDUB4QTRQ2Gihw7Tu+ltOp58Btv4cUy+77PcMOi7IpPrJ5cx945/RotE6BeJcASjkFEpQk/kN3oyv9KT+Zzee++9Q3xH05ZQm82Wt2XhS3n5OReX8nwkgkGcZTVYHXZURWFje4CX9nRmtXlpTyfFTisv7+rgWl8z83a8OaSfXkcBYauDqM3JxHHleIochqiSTaakFZFkWcfkgzb5//GeMFva/aklwwFLnJyyOqX/VzQVNA1F05B0DUXXKLLI9IeiSGlrlZ60WEmajqyrA9tGwQNQAiRVRUtZhE8Ha+rfuUaVZVTZREIxpSzJSetmpmVTl0hZqZP7i+wWguFY0kpKcplX0jE+t0zLqaRpKJqKRdeQEvHzcAX5RUtZd0FHGSZY71RRNA2iEeLRyFmPK2q2ErY6iFjtRKwOomYrqpL8XI3/ZQVdlgApmRlBlrluSjlji50ka9fCoWCYF3YeAzX5Y0fWNEy6xoIqF+54mOixDvpb29ATCSIHDrLjy/9M680fhPLchQte2tPJ/Aons91mwrKcjLrPsIb1t+1FG0Zcnw9kiwWnN3+WWGEJzeZczYckSSiKck6e26NShA4X4T64/upw+09EeXk55eXlZzYwgeAco0YiqKEwJpcLKSVAh/MJfdx3gFvf+CMz2jYDkJAVttXMoW3cNDo8VcQsAy4Np+MT+rPm7ahjzrNPKDqXVxWy5aA/JWo1ZlcWsfNIN5qWFKkySQGMpqdEbnKbrumYgA8tGM/UMic7D3fT2LIPUsu2JkP4JoWwgsbcqkK2HvSDqmJCx1ts48DxXki1rx1TRJnDjJ5IuipoasJ4beRUTSTojWu0BiIkZAVVNqGllq+TosKEblK4Zlo1ksVC854uYpKCbjYzc1wJG4/1E5UUMJv50LVTmDW+DDlVmnLz8T5+9EIriVNcpD0nPqFzx+L1OPjPZ7eiJ2JYdZ2F44toae2ARBxFTaBoCSyaxpyqArYd9KNniCGzrjG11EHb0SBoCUy6zjWTyxlT7ARJRlJk2oNhXtjTmYxwV0xMqSpie0c/iZSbBorC7HElbDrSSxwJSTFx99VeUBR+89ZB4gAmEwmklAuHYrh+qCmXisx7Dl1P3gcZP4IUVcWkxjEn4pgSMcyJ9Os45tR7UyKOWU1es0VNcFmZnSKFpLVTkgZSmikyis2OqbAAk8uFqSD5z+wuYk8Y1mw4Sr/ZjqYoJ/YJzdg2Z2wxWw8HSWgaG2WZ+64Z8HX+afMOEtMnDPnbfyPj7znR18+RJ5/mwK//F1SVd7zQyG/f/Sn6nEVDPvOEpvH0//6JuDmI95srh+zXYjF6Nm9Ftpz/srtaLE7h5ec35kJw4TEqRajHk3xwZpbKSlNTU3PS/QLBhULcnyrZ6XSw8XAwS4CmHzqFdjOPbzjIZbvWGwK0q6iMP96whJ6CYiAVoZ0K4gGMKNsTCdFcgRVSRh+QWsIepHLmjR8YU5pMAZpLkGpIbD7ai2YyA8kH3oZgAt3uGtI2O+0UQHJMP96X4I4iF08eCqBWTDTa3j8oLQ3AFoBpNUMCmtLBL6+dgkj3HfTzn5kphIY511vpeasZl3W+KRnn+/GOfu4bO47astQ4XtqLehpegjo55vQ0BCjA4xvbk5+NyYzJYuVzqXHaffuzPst0vtd1ObbfVTshax7XyzL3XT8wvz9r3kFi6sSsgKMnBs3hXXUzmMpAINKP2pKWOLWwZNigu7TAy+Q9s6vxljj5yQu7SQzaZ5Jlrpjo4bW27BUFgGu8pby1z5/1I+rF0wxO8h3086Pm7ag2l3Fdd84ZyxMbD2aPM5UuK3PblsOBrMC5HzbvMHKxDg5CmjfeY8xT5t/z8Rtuo3lnkLpXnsQaj3LV2y/x3NXvyTl/Yw7tQd37Nl33f5byHJYr2WLGMfH8l4AN7dt/RselM+G43W7D0HSiuI/zxaJFi1i6dOlJa7wLTo9Raauura3F7XYPqena1tZmlJA60X6B4EIh3tODZFKIKhZ+9PfdKQEqGQ+bI8EQT206hD3SzzWbngeg11HIE7f8Az0FxUZ6m0xLZRojVU6OajNHgqFhBejgVEhpZGkgVdGTm9pzXs9d88bx8YW5fwhqejIXZLpvVUs+LAdLsY9d7eX+umS6qrTlL1fqqMy8nLXjkmlrBmeg+eg1kwxhUTsuRxqhHHOTa35ynSszvVVm+qETnW9T+4kjmk+UQWewCBtOgKbTAuUi/RkMvr8yeWrTITa1+3NuPxIMDXtdQ9JfFdqGpmFKpbGqKrQNTd+UMsNn3svpeyXXdP1x62EUWUqWvBzE4rljeWufP+ccvLXPz0evnoQpw6p6svshk1z3xseu9iaT0w/+jFLXIw9KPfXkpnY+es0kYw4f33AwZxT8ie6hHRNm0Tp2KgDT9r5NaW8X99fN4EuLZmb97aZXSHYf7j7lbAyjgUAgQE1NDXV1daxYsYLly5ezYsUKamtrjXKU+WTlypVCX5wH8i5C00FFmfh8Purr67O2PfDAA0Yt03Sb2tpao4bpyfYLBBcCBdOncf0rf2PS+9/LnXPHGdG66YdQVZGDO+eM5cqtr2BJJC1Gz1/1bubNSlqb7q+bkRJsMovnjDNeL/SWYpJl7pwzdtg8oYvnjEOWkuLgS4tmsnjOgCVvae0E7q+bYeSmXOgt5UuLZhrnSbe9a944o8+0Ba1uehULvaXGcZlt7l800+hblpJ5Qr9828ys9nUzqqgd5+GzN0zGJEtZ1zWjstDIyzk4MXztOA9fWjTTELULvaVZierTbdIP9eHmJtf85DpXphCVh6mqNPh8c8Z6Uv0mxaIiSyz0lhrX9KVFA3Mhpfpd6C1Niq3U+7vmJccFA32MLXYY47hr7nhDiKQF6fSMSjn3Z1R2St9fmbkq0+PMtT09X7muK90+LaIy+868T9P9DPxwSN4H9y+ayeK52XOeea+kry/97z2zq7msuojFc7J/MC30lvL+eeONc0+vTF779MpC4/x1M6qM8Z/sb2X4e2PgPqybUZW1LS2e05/ZlxbN5P6bp1Ie7GDavq18uHs7Nc//kXtb/8aYrsPMKrHhiEe5d/dzuJ9cS/vaRoKb3kYNh09wD0m8Nfem5LzoOv+QaB/yI0mWoLwi+VlXO80XVJ7QZcuWsWTJkiHP9Lq6OhYsWDBEM5xv0rXnBeeWvKRo8vl8NDc3s3btWnw+HytWrKCkpMQwqQ+XumnVqlXGcvtwFZNOtP9UEXlCRw4xH9C3aw+927bjmDghZw3oRH+IN5d+GD0axX755VT96zdOu3Y8QKK3FzUcQde1pJ9bgYujvcmgjOFqe4va8aJ2vKgdn5tTqh2v60iSRLndxPoPf9xwv8mk9J/qmXDdVRzz99L++S9k7ZMUheKrrsD7uU/jtzpznqvja1+jb/tOrNVVzP/NLwzLcHq/3vwsex9aw4Q//Bb72DFZQqp3x076duzK23K8a/pUCqZPO6X2kiTR1NSU0/rY2NjIsmXL6O7uxufzsWzZMurq6gx3vMcee4xnn33WCMRpbm6mqamJmpoa1q9fT0NDQ06tEAgEWLNmjVGJcf369axevTrrHA0NDcYYmpqamD9/PuvXr2fRokW89dZb3H333UPG09TURENDAz6fj7a2NpqammhqajLOu2bNGjweD21tbbS2trJ69erTmtuTIfKEjnKECB05LvX56NvdyrE//xVzkZuiObOJRqMAWSL06B//QusPfgLAzO98k+Krrjhpv/FgkK6XX0NXVcpuuQktFmPvf66m8/kXBxpJEqaCAspuuZEJn/oYssVC/959WMvKMBeNfI3pXHNxKSPmI5vRPB+9O3dx6LHfYykupvKO20GS6Gj6G8TjuKZPwz5uDNbKCqxlZShOB7KSFNSJ/hCx7m5Ce/fRs+ltul55ncihw0z/969jr67C5BqaTO3QY79n3+pHAZi75j9x1mRHnx/7y7Ps+Y8fMe6x/8Y5acIFIULb2toMwZhrdbO5uZlFixYZucJXrVrF2rVrjZLgS5Ys4QMf+AAf+MAHaGtrY9GiRYaRq7m5mXXr1uUUeulS4EuWLAGS4jDtA7pq1Sq6uroMESpJklGta9GiRVnL9YPHU19fj9/vZ926dUb7hoYG49rS1+p2u1m6dCmLFi06p76no1mEjsrAJIHgUqHzb8+z85vfYcoDXx62Tcdfk7+YLSUluBec2N0k3tPLwd/8L0ef+TN6PI6ltATPNVciW6x4rrsGU1EhkiyjhsLEg0HiwSBmTzGJnl7UWIxd32ogfLAd29gxFM6eRdHcyyiaOwdrWek5vW6B4GIkcuwY+x/5FZ1/fwGA4quuwD6mGktpCSXXXYPJ5UI2Dx+JbgWcTKR4/jzGLHkfWiJB+GA78UCQcPshYv4AjvHZCcJLrr/WEKH+198cIkLt48ZSfNUVyOYL53GfFjdtbW05RWhaBGVaMxcsWJB1fNr1r7GxEbfbTWNjI5B0CWxpacl53rq6Om699VYefPBB6urqeOCBB4x9J8pDnsvNMHM8brc7K2g6HdOSvra0QPb5fHg8nlOqKnmxcOHclQLBRUi8txcAxenMvT8QpHfrdgDKbr0pmTZmGHq372DHv3ybWFcXzhovVXfdQdltt+IYNw7FbsdzzZXJ4yUpVeUmlYw7XcUnGmXsP34Q/2tv0Pv2Vjr+8iwdf3kWgKn/byWea65CMpmIdweEKD3P6LqOFokQDwRJ9Pfjmpx8gEWPddC7dz/oGpLFgsnhQHE4MBUVYKuoGOFRX7rous7Rp/7Ivod/gRaLUVQ7F+/nPo3nmquweIpP+Hd7ImSTCeekiWixGB3PNtP2k58z6z++Q+HM6UYbW1UllopyYsc66Nuxc0gfhbNmMrH+k/QUFJzx9Y0ES5YsoampybBKZtLU1DTEUjicK15XVxcLFizI6mc4K6PX66W7u9uwlt56662GNXMwK1asYOXKlUbw1GC3gdNxDayvr8ftdlNfX09NTc1Ji+5cTAgRKhCMIPFAEACTK7cIDfg2Gq9PtAyvJRLs/FYD8d4evJ//DOM/eQ+2yopkkvphkBQFLKAw4Aox5ctfRNc01FCYvl278b/yGl2vvYFr2lTi3d307tjNrm99F2tlBYWXz6Zo7hxck2uwjx97QguPIDeJvn4Uhx1JlgkdbGfvQ2uItB8i2tmJnvIBlS0W5qz5GbFIhO5XX+fwb/5vSD/2ieOZ+e1vIikKnX97noBvI7aqShw1Xlw1XhyTJqLYL93SuOcbNRzm0LrHMRUWMPlLX6DqzvdgKStFOkfLn7LFQsn117Lnez9m38O/4LIfNmQJW+fUKcSOddC7Y6fhh5qJJMvJ4hQXEI888gjz58+nubk5S+A1Njbi8/lOaC0MBAIEg8nv1vr6ehYtWpS1v7GxMae4ffDBB43S4HV1dVnH5cpDnl6aPxsaGxtpaWkxxG5XVxeBQGDYMV5sCBEqEIwghggdxkoR8G0AQLZZKZg5Y9h+dFVl0mfqMblcVC1+L+bCM7d6SLKMyeXEXTsXd+1cJn2mHrU/RKK3FySZsttupWfT2xx/9jmOP/scANbKCuY89CMks5mezVvo37sPS7Ebk8uF4nRgcrqwlHqweE4tD+PFiK5p9LftpWfzFnq376Rv5y4ihw4z88F/w1peTuT4cYK+jVgryymcNQtzcRHmoiLMxW5sY6tRdB2Ty0VBjRdJktFiURJ9fST6+jA5nJiLitDVBJGjxwhu2ESgJaOopiRRecftTPjUx1CsVmLdgaSFbhRVlIkHgoT27Sd86DDxYJAxS+9CTyToaP47B3/130nLvZYsZiA77DinTmHi/Z/DpOuE9u4jHuzB6Z2Etbwsb2Pu292KfWw10Y7jTP3nFRTOmkHBjGnIFss5P1fJwqupfM+7OPrUH/G/+RYl11xt7HNOm0L3S68Q93cT6+zKWqmI+f20/uQhnF/4NJZi95B+tVj8jHN4ng5a7PQqdrndblpbW1m5ciU+X3aB2EwB6vP5jEw5zc3NAKxfv94IFqqtrWX16tXU19cbqZ2GS7VUUlJiLN8DRgS+z+ejqakJv99vZOJpa2ujuLgYj8eD2+02gpZyjSf9f11dndFHIBCgtraWuro61q5dawQnpf1Fcy3xX4wIESoQjCCJYBAkCcUxNDJX13WC65MitOjyy3JWNdF1nXggQKKnl+IFtRTMnDGsVfVMkWTZqBRjq66ifNEtJPr76d/div/NFvq27YCUpUXr6+N4898Nn7hMSm9OBkBJksSu736Pvt17UGw2ZJsNxW5HsduovOM9lFx7DZKi0PHMn5BtNpxVVVjLSrFWVZ7Qsjva0HUdPR5HtljQEgl8H1tO9MhRY7+1ohzPdQsxuYtwTptMweWzqHznIhSHA9liQTKbkM1mw+IVDodhxnTs77ot9/lUFS0ep/Dyy5i16tuEDxykZ+s2erdtp3f7LhwTJ6L29hHt6GT7A18n0ddPwYxpFF4+m8LLL6Ng+tTzIp5OROToMfatfpS+HTuJdhwf2KHIlFx7Ter+sGKtrEAymZHShQzCYSRFQQtHiEdjHFr7e7peegUAS3kZRZfNpmj+PDxXX3leguziwR5af/IQXc+/yOSvfJGyultwTJyAtbTknJ8rkykPfJmjT/+JI394Gs9VVxo/IpxTpxht+nbsHOIuE/RtxNbbz+A4ZNliyWsVozO5v05mbaytrR2yZP7WW28lz5ean1zL5bkYLgl+bW1tVjR7c3MzV1xxhRFoFAgEWLlyJatWrWLFihVDxjP4/WArbrqfNJdSPtIL5xtdILgIsVZW4JwyOae4inV2GQ/monlzch7vf+0Ndn2rgZr7P4fnmqvOuQDNhaQomAsLcc+fh3v+PCBZ/k+NRNCiMab969cY/9EPE/N3EwsEiAeDJII9OCaMx1xUBLqGc8pkkGW0cBg1krToxY53Eu/2Ew8G0RIJ2h/9NVo0lnVea2UFk7/8RQpmTEOSZSJHjmKrrho1Fr2Y30/Al7RCBjdsxH3FAsYsfR9IEu4FtchmM8VXXUHxFfOxVVclxXdqOf5skRQFRVFQbDagAGtZqfH56KqKGo6ghpIW7dKbbyS4aTPBjZsJpH7oSCYTMxu+ReGspMVdi0ZzRmOfLloiQWjvfvp27qJv5y56t+1g4qeXYa+uItrlp+ulV7BVV1F68w04J9fgnFKDa3INDu8kFKsV95XzmVT/SSQ55c8MoOuE+/vR1QQWWWHSZ+rxXHtN8po2beb4c3/n+HN/p/LO9zDuIx/CVOBCj8dRzkEWDv+rb7DnBz8m3h3AfcV8PDdeR+Hsmal5P78UTJ2Ce0EtwY2biXYcx1aZ9AN2TK5J1nXVNHp37qLk+muNY9I/cDU1MaS/fNZxv5hIW0PTpKPaM4Wq4NQQIlQgGEGmff2rBFp8OQMX+ncP/Fp2TZs6ZL+uaex/5JcAFF9zFZYRDBZK10QHsJaVUjR75pA2uq6DpqFrGu4FtcbSqp7ahqaiqxp6IoEajzPz4Z8QPXwUtbOLSPshwgcOED54KFlh6lgH/Xv3sfNfv41it+PwTsQ1dQqu6VNxTZ2CfdzYnJV0zhddr7zGgV/+JmtZ0zamGmt5KY6JEzC7i5KWPZcTZQRSCkmKgsnlxORyYi0vY85DP0KNRol1+el+/U2632ghuHET1vIyIkc76Nmyhdbv/yT5I2nSRJw1Xuzjx2GrrjRS7OiaZohCLRIh0dNLvLcXxW7HPqYaXdfZ/vV/I9DiQ48PLMWai4uJ9/RSeNks7N6JXP9yczJVkd2ObLOesiA3JWu6YrPbsVVWUL7olmRAVzhCX2sbnX97HvvECYBOf2sbW764Ate0KZTechOlN16X/EF0GkSOHWPfw4/S9eLLKA47k7/0Bcbe8yHsY6rzeq+N+cBdJHp6CbcfMkSokrIWRw8fIdKeXelKtlqTKxWqmrcxXuysWLGCVatWGblHIWkNPRc+opcaQoQKBCOInkgKL9ky9MHbt2eP8dpZMzQfW+Ct9YQPtFP1/jspumx2Xh+EZ4IkSaAopxwpXFaS9B9N55DV4nG0SBQ1GkWLRlGcTqrev5i+Hbvob2szsgiYCguYs/pnKDYbob37iRw+jHNyDY6J488qeErXdWIdx+ndvoPebTvo272HaV9/AC0WI9rZSay7G8/111J85XxKb7oRV80kTAUFKI7RmQNXsVqxV1dhv+tOqu+6Ey0WI9HXj9rfjxaLUnL9tfS3tuF/7Q38r74OgH38OGZ++1+RTCYO/t9ajj39JyPbQpqSG69n3D0fApI/TgpmzaBg2lQKZs/EPX8ejgnjkxH9TscZR40PhyRJKA47RZfNouiyWeiaRqKnh55tOyiqnUugxUfP21tp++nPKZpzGaU330jFu98x7N+OrmnoqprMCuHvpuvlV3EvqGXKA1+heP48TAVnbyk+XcZ//B5Kb7yeni1bk2NLzaG1uoro4SOE2w9ntZckCcVmEyL0HDMS9esvRoQIFQhGkD3f+xFIUHH7u4bs69uVFKG2MdU5l9mPPPk0SBJjPnT3WQUiXSjIZjOy2Ww8+O1jqim7+QbUaBS1v5++HbsIbNhMzO/H5HSghiMc+8tfjeApSVGwVlRgq66k5IZrqXj3O5Ekid5tO9BiMcOqp2sqiZ4+XNOmYKusIBEKsXXF14i0HyLR22eMR3E4CB85iss7ker3L2bCJz+GubAAk9N5zsVVPpAtFiweC3iKGTNuLNWL34saChPr7qZ32w5Ce/ejaQnMJR70hIrTO5Hia64ETU+KykIXJlcBBbNnUjB9GrLVguc/f4hityHb7SNjAZZlzG43JQuv5uqn1hFuP8SRJ5/h2B//QnDDJmJdXRRdPhvF4cD/2pv0t7YimU3oqkb4YDv9e9qovutOPNddg6XEQ+0vH8Z9xXys5WUj5gIiSRKmoiIUhwM1HDZcJmzVVfQAkcNH0DUta3yK3YaeECJUcG7IlYHhTBEiVCAYIXRdZ/+jv6Lw8suovOM9Q/b3706KUNeUyUP2RY930v3m+qRf5mWzz/tYRzOK1YpiteJZeDWehcmIYTUSQQ2FsXjuo+yWm+jZspW+XXuIHDpMcONmrBXlFMxK1pnf84OfENq7b0i/4z/+ETzXLUSSJOKBILbqauwTx1M4eybFC+bjmj4Vc2EBitN5UaankhTFCEhzjB9nbNd1HT2RwD1/LmgpC6iUbC+dhqU730iKgmPCeGo+fy+T7l1OeP9B+nbtxlJWihoK0f3Gm4bFF5I+svbx45AtZlyTa7CUlmApLRkVwXHBjZvZ9eD3GPuhuylOFbCwjakGkr68sS5/VnBS1fsXEy8vRRPWUME5QNM0TOfo72Dk/5oEgksULRxBT6go9qGBKfFAkFhnMmGxc3LNkGNjXV3YxlRT/s5FmHOkXbnUUWw2FJsNi6cYd+1cIB08FU0G5/T1IykyejzBhGUfJ3r0WNJnVddBkjC7iyiacxn2sWORLWaubXoG2WpNRmvbbaPe9eF8IkkSktkMF7Dwlk0mnDWTcNZMMvLizn3kP4kcPpK0issy1vIyTE4nJqdz9LlU6Bo9m94mOHOGIUKt1VXG7nD7oSwRWnn7O+lCJx6P09/fj3OY4hgCwcno7+8nkUjgyJHR5UwQIlQgGCHiqWTKit02RISGD7Ybrx2ThtZ1dk6uYca/fZ2Cy2aNmsjw0U46eGqw64Jr6lBLs+DSIZ0X1+RyGoE+o53iq64AWaZv127D19OWIUIjhw5DZkYNWcYVChHUdQ4cOIDFYrmof0ilU1FdzNd4Opyr+dB1nVgshizLlJWdm3y84uklEIwQ8WAPQNISOmgJM5wR4WofO2bIsWp/P4rLlTP5tEAguLgxFxbimlxDeP8B1HAESOZHlVJLpJFD2cFJ+x5+lG333keZqwCHw3HRizNN09AusApR55NzNR+SJOFwOKiqqsJyjnIKC0uoQDBCJHqSIlTOkbswLULTuTEz6X7jLQ6te5wJyz6OufDcJ+IWCASjH9eMaRx98hli3d0o5WWpwLtyIocOEzl6LKutGg4R6+rCruuUTpw4MgPOI+FwGBjIrHGpM5rnQ1hCBYIRwlxURMXt78Q+buyQfWkRmqtKUHeLj+CGTViKi0dtEIhAIDi/FKYCEkNte41taT/QtD95mnTC+kRff55GJxCcGsISKhCMEK5pU5iy4n76W9uG7EsnnLanIl4zCW56G1NhIQWzhq8lLxAILm4q3/0OdFXFXFxsbLOUJMuGxroGidCUBUzt60MgGE0IS6hAMIJo8fiQwCJdVQmnfLoG+4Mm+voIte2lYOb0EUmULRAIRgeuaVOovusOzIUFRuCJpTQtQv3JilYpFGfKEtrbm/+BCgQnQIhQgWCEOPTY79n21a8T7fJnbY91dhllDm3jskVo77YdoOsUzrkMk0izIhBc0shWK2o0ih6LAQMiVE8kjMBHAFN6Ob5fLMcLRhdChAoEI0Tvth0EWnxDtkczggrsVVVZ+/pS9eTd8+YIf1CB4BJn82fuY9vKrydzmzIgQiF7Sd6z8Gq8X/wM9vFD070JBCOJ8AkVCEaIeCAAYJTdSxM7ftx4ba0oz9pXcv1CdE2l8PJLu0qSQCAAS1kpid5e4r29mAoKDJ9QSAUnpQpd2KoqKbxsNibnuUkwLhCcK4QlVCAYIYw8oYPqwseOD1gwLBlVTyBp6ah4Rx2WjGAEgUBwaeKYmLRsxg4nV0+spYNEaApd09CiURLhUH4HKBCcBCFCBYIRIh7sSZaCtFqztqctoaaiQhSbzdiuxeNEj3UgW21GtKtAILh0SYvQ6NGjAJhLPJBKRJ8pQnu2bGPzvV/k6BNP53+QAsEJECJUIBghEj09KE7HkOj42PFOAKzl2UvxfTt2sfFT93LsL02jr5a1QCDIO44J4wCIpn64yiYTZrcbyPYJHUjRJAKTBKML4RMqEIwQ4z72j/Ru2YakDBKhHckHirU8uzZv/759ADhrJomgJIFAYFhC453+gTRNJR7i3d1ZltD0j9ZEfz+6rl/0ZTsFFw5ChAoEI0TFO2/DVl4GGZZQXdeJdaYtodkiNHwwmcDeNXVK/gYpEAhGLbbqKq7+4+P0HGxHTyQAMBe7AYgHg0Y7wxIaCoOmgfgRKxgliOV4gWCkUFV0Vctajlf7+tHCESCHCD1wEMlkwjFxfF6HKRAIRieSLOOs8aLYbOjxlAgtKgQgHsgQoY60CA2hq2r+ByoQDIMQoQLBCKDFYry48GYO/Ob/kOQBq0RWeqbBIrT9ENbKcqMOtEAgEIQPH6Zv6za0VIELc1ERQFayetlqBUlCDYWzKikJBCONEKECwQgQ7+klHgiiJ+JZPqHpoCTITs+kqyqx453YqquHRNMLBIJLl30/f4S9//EjoySnKSVCtUgENRoFQJIkLv/ZD5j0uX9CV4UIFYwehE+oQDACJFL+WrLNluUTGu8OGK8zE09LisK8Xz6MJMvINiFCBQJBEmtlJQDxLj/UeDG7i4x9iUAQJVXwwlZZmQxeEpZQwShiVFtCm5ubCaSqyuSira0tq036vUAw2jES1dvtWZGqcX+38driGZSQXtexlpchm815GaNAIBj92KtSItTvBwZ8QiE7OCl0sJ2+XbvQNeETKhg95FWErlq1isbGRtasWcPKlStPKDABGhoaKC4uRpKkrH81NclSZD6fj0WLFhltFi1ahNfrzcOVCARnR1qEmgb5d8a7kyJUcbmQLRZje9+eVrpbfCI1k0AgyMJaWQEM/IBN+4RCdnDS3p89TOv3fyJ8QgWjirwtx69atQqAJUuWAEkBuXTpUpqamoY9xuv1Dtm/bt066uvrjferV6/G4/Hg9Xqpra09DyMXCM49id6UJXRQ0vm0CB1sBT3252c5+sTTVLzrtvwMUCAQXBBYU8vtaaunKWM5PjM4SbHbiRw5KnxCBaOKvInQBx98kPXr1xvva2traWlpoa2tbVjr5aJFi6irqzPet7W1UVNTkyU26+rqhPVTcMHhWXgNs77/3SGJ6tM+oYNFaPToMVAUbKmlN4FAIACwVVZg9hQDErquZ1tCg9lpmtRIRKRoEowq8iJCfT4fgUAAj8eTtd3j8dDY2MiKFStyHpe2mqZZvXo1DQ0NWdsCgQA+nw+/38+CBQtwp0qWDUdHRwfHM9LgAOzZsweASCRCOBw+lUs6YyKRyHnt/0Ljkp0PpwPr1MnEjh4jmopgBYilltQUtztre+ToMczFbuKSdN7v0dHCJXtvDIOYj2zEfCSRKsq57Im19L29lUgohGQ2JYMdNY1Il3/ge8RqAVUl1N2NXlgwsoM+z4h7I5t8z4fdfuplpfPiE+pPOUwPFohut5uujPq2J2LNmjUsWrRoyPa1a9fidrtZsGABy5YtO2lg0kMPPcTs2bOz/i1evPiUxiAQnCti3YGsICRIVktKpJbjzYMsobHjnVhKS5FEUJJAIBiEZDIhKQp6QkWSZUwpkZnIUTUp0S/qxwtGD3mxhJ4oAOlkwUlpVq9enbWcD0lLaaa1tL6+nqVLl7J3795hLaL33nsvS5cuzdq2Z88eFi9ejM1mOy0Ffzbk6zwXCpfafLR+49858MvfcPlDP8KayvuZ6OtHjyUTTtvLSge2h0Ko/f3YqipwFhYaD5NLhUvt3jgZYj6yEfMBh158md63t+L+wBJMVisWt5tEIIjW1298j1g9xSguJ2ZNv2Tm7FK5zlNlNM5HXkTocILwVAVoY2PjKfl9LliwgEAgQEtLS5YvaSbl5eWUl5ef0nkFgvOFER3vchnbYqkVAwBzhuuK2h/C4Z2Ic+IEJIuwhAoEgmyOPf4UfVu2Me79i4GBhPWZPqETPvkxyupuHVKJTSAYSfKyHJ/2Bc0lOtPplk7E6tWrc4rQ4uLirOX3tNg9VXErEIwUiWAQyWRKltNLMVyOUGtZKdP/5WuM/dDdyCZRX0IgEGRjKS9DDYVQQ0l/cXN6Ob6n12gjSRISughMEowq8iJCa2trcbvdtLW1ZW1va2sb1mKZSUtLCyUZ1WPSeL3eLHGa7l+kahKMduI9vSgOR1bez0xL6ODoeC2REDXjBQJBTtIlftN5QU0FyRWWRG+f0SZ86DDH//4CofZD+R+gQDAMeUtW/8ADD7B27Vrjvc/no7a21hCMPp8vK/9nJsNZNgenZ2poaGD58uUiZZNg1JMI9qA4s0VoZslOc7HbeN35wssc+f0TaLEoAoFAMJi0+46RK7QgZQnt602W6gT6du6m/X9+R9/WHSMzSIEgB3lb21uxYgWrVq1izZo1ALS2tvLcc88Z+09UcnO4RPQNDQ1GEvyuri5qamqGTfckEIwm4j09SUtoZt34tP+WLBsPEQD/a69zvOlvTP2auLcFAsFQLCXJlZMBEZq0hOrxBFokimK3GYUx1JCIjheMHvLqYHYigTg40j2T1tbWM+pTIBitXP3M4wRa1mclq48HUsFKBa4scRrr8iNbrVlJqAUCgSCNo6YG97VXGz7mmQGPib6+pAhNRUbHM5boBYKRRkQ5CAQjgGKzYnI6k0mlUxhWjEFiM9blx1zsFjlCBQJBToqvvRpLeSmxtn0ARp5QgERvL9ayUkOEqiJPqGAUkTefUIFAkESNRul8/iUiR45mWTwTqaACsztbhMa7uzEXu5HN4jejQCDIjaQMfD9kWUJTls/0crxIVi8YTQgRKhDkmeixDjZ/5oscf+75nD6hpsJCY5sWj5Po6cXiKUYyCUuoQCAYSqK3lz3f+i6dz7+IrmmGT2h6HySFqXPKZCyeEnRNG6mhCgRZCNOKQJBnEqlE9bLNmr0cn8oCYSoaEKF6IkH5bXW4pk8VllCBQJATyWyh69nnKL76ypQIzViO70taQs3uImZ88/9hKnCha1rWD2CBYKQQTzWBIM+kLZ6Kw4EkSUAyD2h62SzTJ1Sx2xn3kX/AXOIRPqECgSAnis2K7LCj9vejq+ogS2hGIJIso+uphPWi8IVgFCDuQoEgz6RLdqZ9tAASPT3Ga3OGJRRAU1Vkm9UQrAKBQDAYs7uIRF8fuqomC1tIEui6sRwP0P67dZjdRbhr547cQAWCDIQ9XiDIM+nleFNGBaR0pRPItoQe//sL7HrwP4i0H87fAAUCwQWH2V2ctHqqKpIsY3I5gWxLaNcLL9H92hvoqvAJFYwOhAgVCPKNLGMu8aBk+G0ZierJ9gkN7TtA/87dwh9UIBCcELOnOGUJTQrMgapJAyJUsdtRw2ERmCQYNYgnm0CQZ8Z+cAnOyV6ixzqMbfHuARGamaIp3pWsJ2+trMzfAAUCwQVHxfvuwOmdgBaPA7nrx8t2e9L1R1NHZIwCwWCECBUI8oyuaegJNXfJTrItobFAAGQZi6c4n0MUCAQXGBV33UHRlMlGMvp0rtBMn1DFbid67Bi6po/IGAWCwYjleIEgzxx95s8ceeKprG2ZgUmZ6VXigQCmAheyxZK38QkEggsPSVGQFCUZ+U7GcnyGJdTksKOFI0YbgWCkESJUIMgz7f+3lgO//A0oirEt0ZO0ViguJ1Lm9mAP5sJCJJFORSAQnICOJ59hw6c+Td/uPUDGcnyGT6j7qisoW3SLsWQvEIw04skmEOSZeLAHxelEzhCW8ZQlNNMKCjD2Q3cnE0ubFAQCgWA4tHic6JGjAyngnMnoeLU/hK7rSJJE5bvfSbTzOCLZm2C0IESoQJBnEsEgJpcTScmoG5+yhA4WocVXLUAym5BFyU6BQHAC0uV+1b60T2hShOqJBFo0imKzgSyhq7qIjheMGsRyvECQZ+LBHhSHPTswKWUJVQoHRKiu62gJFUk2CUuoQCA4IemARjUUApKuPWnU/uS2zr+9wK5vPUj/3v35H6BAkAMhQgWCPBMPBpNLZfJQn9BMS2iobR++ez7J0aeeET6hAoHghKQrrSVSItTkHBChab/QeDBI/542Yn5//gcoEORAPNkEgjyiaxolC6/BXOzOsoSmo+NNhdkJ7PV4Atlmy2orEAgEgzGW40NhdF0fJEKTS/SKPVkqOKuevEAwgggRKhDkEUmWuexHqwhs3GREwWvxOGooDAxEtMJAKU9raWn+ByoQCC4ozKWlzP7+d1GjUdA0FNfAd0k6d6jisGe9FwhGGmFeEQjyjK5p6KpmBCZlJpM25SjlaSkrye8ABQLBBYdis1J+263Yx41BV7UTWkKFCBWMFoQIFQjySN/uVrZ+9ev0bN4CqSX2tD8oDFqOT1lCLSVChAoEglNDi8bQNTUrMCnRn1x+N5bj+4QIFYwOhAgVCPJI+OBBOv78LLHOLiQpma0vnlEtSRlULQmEJVQgEJwar77zTnZ9Z1XSEpoZHd+XDFayjx/LuHs+TMHsmSM1RIEgC+ETKhDkkbR1U3E6jG3DWULHfnApRfPmYvF48jdAgUBwwWIqLCDu7wZNRXbYjTKeaUuoxeOhbNEtWErFD1vB6EBYQgWCPJIWoSbHgAhNVziBbJ9Qs9tNwfSpKFZr/gYoEAguWMyFhaihULLKmiQZS/JqxvK7JMugiYT1gtGBEKECQR4xSupl+msNYwkNtx8iHgyKHKECgeCUMBUVkugPoSXU5PtUhHwiFYikhiNsrP8sbT/5T3RVHbFxCgRpxNNNIMgjiWAAyLZ4pnOESmYTss1mbN/xL99CttuoeOeivI5RIBBcmJiLikDTUMPJlG9pt5+0JVS2WlD7+on39ApLqGBUIESoQJBHPNcuJHzkGGZ3kbEtnq6WVFhoBCtB0nrhKC1BkkXJToFAcHLS3ytqKhm9YQlNiVBJlpFttqRIVYUIFYw8QoQKBHmk9KbrMTkdWRVL0pZQ86C68Wp/PyaXU9SNFwgEp0TN/Z/Hc91C48dsOldoIiMvqGKzoYUjwhIqGBUIn1CBII/oqpr88ldy1I1Pld0DkpYKXcfkchmVlQQCgeBE2MrLsFWUG+9zBSbJ9qQlVNeET6hg5BEiVCDIIxs+8U/s/FZDVi34eA5LaPqhoRS4RGCSQCA4JWLd3fTt3mNUWxuwhA6svCh2e1KEisAkwShgVIvQtrY2mpubCaSSdqffCwQXKr3bdhA73pll3cxpCQ0lk0ubCgqyBKtAIBAMR+fzL/H2575E79YdAEbCei0SRUskAPB+7tNM/vIXQNNHbJwCQZq8mlhWrVqF1+vF7/fT2trKAw88gNvtHra9z+dj6dKlxnuv10tTU9NZ9SkQjCTxnh7sY8cYdeN1XTdqx2emZ3JMnMDcR3+OfdzYERmnQCC48EiLTjUSAUBJBSZBsl68XFSEa8pkEj09whIqGBXkTYSuWrUKgCVLlgADAnOwqBzM6tWr8Xg8eL1eamtrz0mfAsFIoGsaiZ5eFKfDsISq/SHjYWDOsIQCSEjGcppAIBCcjHQ0vJYSoZnfH4m+fsxFRaiRCJEjRyiYNWNExigQZJI3Efrggw+yfv16431tbS0tLS20tbXh9XqHPa6urm7Y/Wfap0AwEiT6+kDXUewOSC2xJzLqxg9OVN/dsh5rdVXexykQCC5MBovQzKIYaipC/uBv/pejT/2Rhc/9CfvYMfkfpECQQV5EqM/nIxAI4BlUA9vj8dDY2MiKFSuGPTYQCODz+fD7/SxYsMBYaj/TPjs6Ojh+/HjWtj179gAQiUQIp5L8ni8iqS8HQZJLaT7CR48lX1gtxBIJZCDU1TXQwG4nGosCEHj1dQ6ufhTHrBnYL5uZ/8GOAi6le+NUEPORjZiPAdJzoZuTj/RYf4hIJIJusRhtwt0BzNEo2JJlgPuPH8d6np93I4W4N7LJ93zY7fZTbpuXiAe/3w8wxFfT7XbTlfkQzsHatWtxu90sWLCAZcuWGYFJZ9rnQw89xOzZs7P+LV68+PQuSCA4A8yeYqZ8+18pumK+EWyUWbJTyayilE42XVSEQCAQnAqKy4lssyXzhGqaUTEJMjJuOJPWUjXju0cgGCnyYglNR7ef7r4lS5YY/p4A9fX1LF26lL17955xn/fee29WsBMkLaGLFy/GZrOdloI/G/J1nguFS2I+7Hbk226lZ/MWbOnrDQ/8QnWUlqBYrKntSQuFs7zs0pibE3CpX/9gxHxkI+ZjAHtxMTe++SLBjZuxmMzgKTb2SbEYVqsVa1Hyx64UjV30c3exX9/pMhrnIy8idLho9ROJxVwsWLCAQCBAS0vLGfdZXl5OeXn5CdsIBOcDLR5Hi8WztmX6hJoLC0nXMEn0JS2hloyHiEAgEJwMSVGQZBldUzE5B6Lj098pRu7QjO8egWCkyMtyfNpvM5dArKmpGfa44uLirLygaeGZ6Qt6un0KBCPFobWNvHTtLfRs2WZsy1yONxVkPDB6+0CSspboBQKB4GQc+0sTgfW+ZBCkww6pEp4JYzk+KULjvWI5XjDy5EWE1tbW4na7aWtry9re1tZGXV3dsMd5vd6sKPf08bW1tWfcp0AwUsSDSctDpp9WulqSMqg8p9ldhGPSRBSzqJYkEAhOndbv/5gjT/4RXdWQZBnFkfy+SUfHu2vncvnDP6Fs0S0jOUyBAMhjxaQHHniAtWvXGu99Pp8hJtPv6+vrs44ZnJ6poaGB5cuXG9tO1qdAMJqIB1Kl9DISSKctoZklOwEmLv8Es1Z9G2RRN14gEJw6itOJFomga0nnnnQC+7QlVDabMTkcEE+M2BgFgjR5M7OsWLGCVatWsWbNGgBaW1t57rnnjP25SnI2NDQYCem7urqoqanJSr10sj4FgtFEvDsAZOcDTftlmYqyE9XrqopkMmVZRwUCgeBkmFxOYl1dkBKh6eX3tCVUi8XpeXsL9nHjcC+oTUbSCwQjRF7X+k6UD3RwJPypHHMq+wWC0UI80A1kW0LTS/SDqyW1/3YdjokTKL7qivwNUCAQXPCYXC7UcA5LqCFCo+z5jx9RdtutjLn7/Ugm4fIjGDnythwvEFzqxDr9yHY7is1mbEv7hJoz8oHqmsaRPzxF92tvCEuoQCA4LZQCF1okgpYqB5z+0WvkCU2l6VH7+kX9eMGII34CCQR5YvYPGuh65bUsYWksx2cs0avhcDKytcAlRKhAIDgtCmdOJ3zgIHo0WX0tHQiZTtEkKQqyzUYiFAJVG7YfgSAfCBEqEOQJc3ERtqpKJCW5AKFGImiR5IPCnOETmrZYmAoKhL+WQCA4LaasuJ/Sm24kHswOhEz0h4w2isOB2h8SllDBiCOW4wWCPBFo2UDs+HFIWTezcoRmLMenAwjMg4KVBAKB4KTIMshSzsAkw0/U6UANhdA1IUIFI4sQoQJBHlBDYVo+eA/tv1tnLLHHs6olZSzHGyJU1I0XCASnh/+V19j70BoiR48BAxWS0HXUULIcsLWqElNhIbpYjheMMGI5XiDIAzG/H0haJdIiNBEcEKGZKZoUl4viq6/EOdmLQCAQnA69O3Zx7Jk/UzBzBjAQHQ/JH7gml5NpX/8qal+fWI4XjDjCEioQ5IFYV1KEmjJEaDyYaQkdsHo6Jk1k4rKPU3zlgvwOUiAQXPCkfUC1cNLqqWSI0HTCekmW0VVNiFDBiCNEqECQBwxLqMtpBBslMpbjM6PjdU0DSRKR8QKB4LQx8oJGIsn3zgwR2p+MkO/ZspVDjX8wgpcEgpFCiFCBIA8YltDMRPXDiNDOpudo++nPiXZ25W+AAoHgosCwhKZEaKYlVO1LRsgHN2zm6BNPEz3Wkf8BCgQZCBEqEOQBLRJBtliyApDSPqGKy4WcUbUk1LqX4IZNSLL48xQIBKdH2hKaFqGZP3zTltB05o14IJDfwQkEgxCBSQJBHhj3j/+Aa9pUIhmWh4GSnQVZbdN5Qs1ud97GJxAILg6sVZVUvX8xtrFj0XU9ezk+nYM4JUzj3WI5XjCyCFOLQJAHdE1DT6hZFk+jWtKgfKCJ/n6QZUwFLgQCgeB0sI+pZvrXv0rRnMtA04yKSTCQ/i393RIPCBEqGFmECBUI8sCxvzzL8ef+nhVsNGAJzRahan8/JqcT2SwWKgQCwRkgy0iShK5pyGYzss0KZFhCC5KrL2I5XjDSCBEqEOSBvQ89wr5HfmmU7ISBwCRTDhGqOB1IsoiOFwgEp4eu62xc/lkON/4ho0JSqnRnqn68payUotq5WEo8IzZOgQCET6hAkBdiHccxu4uyLKHp5fjB5TnL3rEIxaSIFE0CgeC0kSSJ4KbNuKZMAU0HSC7Jd3WhpurH28dUM/n+z2MuLkbXdSNtnECQb4QIFQjyQLSzE/vYMUhK8k9OjUTQIlFgqAgtvm4hdncRKGKhQiAQnD4mpxMtGhmwhLqyLaGQTFiPpqInEkhm84iMUyAQTzmB4DyjRqMkgj2YCgqQTKmSnT29xv7By/G6piHJsrCECgSCM0JxOlHDEdDTIjQZIZ8OTALYu/pRDv7P70TVJMGIIkSoQHCeiaWSzpsLC5FS0fGZieozLaGJvn62f+HLHPyf3wkRKhAIzgiTy4kWiaCrSRGqpNI0pQOTAIIbNxPc9DZ6QohQwcghRKhAcJ7R43HcC2qxVlcZwjKRUTc+M0WT2tdHItiTXCITfloCgeAMMDmdqJEI6Emf0FyWUJPLSaKvD10TIlQwcgifUIHgPOOYOIHLf/ZDerZuM4RlPEOEZqZoMvL4DfITFQgEglNlwvJPJK2cKYGZLt2Z6Os3ApFMLhfRY8eEJVQwoggRKhDkAS0RJ9OumciqG5+9HA9Dg5UEAoHgVKl4121YS0tQozEAo2qSnkigxWIoViumwgL629qET6hgRBHL8QLBeebIU39kz6ofZvljxbNE6EDZzrQl1FxYlL8BCgSCiwpJlkFWIBUdn7aEwkBZYJPLhR5PZEXMCwT5RohQgeA843/5NY4+9UeQB2yhaZ9QxenMKuVpPCCKsuvJCwQCwamy54c/o+XujxDp6AAGktUDJHqTorPkhmupXnqXEbwkEIwEYjleIDjPRI4dA0nCXFxsbDNKdg5adi+cP5fxn/snCi+bldcxCgSCiwhdRw2H0VLJ6TNXWxK9yfRwJddfi338eFEeWDCiCEuoQHCeiR7rwFTgQrFajG1pn9DBAUhmt5vCy2ZjLS3N6xgFAsHFg6kglZw+HAHAnCFC065AySBJHS2eyPv4BII0QoQKBOeZ6LEOzMXFRrUkgHgwCGRHxgPEe3pRYzF0kZ5JIBCcIUZKpnDaEpoR/JiyhAY3b2HbA9/g2F+ezf8ABYIUQoQKBOcRXdOIHDmKxVOMlLHsFesOAGAudme1b3/kv9j+2fvRE8I6IRAIzox0mU4tFE6+L8hYjk9Va5MUhVjHcSKHj+R/gAJBCuEMIhCcR7R4nPEf+wigGwFIuqYRH0aEJvr6kEwmFIcjvwMVCAQXDYorezlecdiRFAVdVQ0Rmv7uiXV1jcgYBQIQIlQgOK8oViuTPv0pghs3GyU7E729RuoUS0awEiRTNCkOuwgWEAgEZ0zxlfOZs/pnxnK8JEmYCgqIBwLE0yLUnUwDlwgE0RKJrCwdAkG+uKCX49va2mhubiYQCGS9FwhGE1osnqxSIif/3NJWUCArYh6SIlR2OETdeIFAcMZYiotxz7s8Kyo+/TodFKnY7UhmM/HugHD/EYwYef3ps2rVKrxeL36/n9bWVh544AHcbvcJj1m5ciWQFJgej4eGhgbjGJ/Px9KlS422Xq+Xpqam8zV8geC0Ofi/a2n76c8Z9+EP4pw4AYCYv9vYP3g5Xu0PYXa7kWQhQgUCwZmhqyrRLj+JnoFE9IYITQUmSZKEuaiIWCCIHk+AbUSGKrjEyZsIXbVqFQBLliwBBgTkiURjfX19luisr69n/vz5tLa2Gm1Wr16Nx+PB6/VSW1t7/i5AIDgD+nfvoX/3HmSH3dgW7x4QoUOX40PYxlQjKRf0IoVAIBhBosc7ef1diym95SYKZ81AkmUjE0d6OR7A+9l6MJnQVWEJFYwMeROhDz74IOvXrzfe19bW0tLSQltbG16vd0j7QCBAc3MzbW1thrhcuXIla9asobm5mbq6OgDq6upyHi8QjAbChw4DYCsvM7bFspbj3Vntp//gu0gWs1iOFwgEZ4wRHR+NomsakiwP5A7tHRChRfPnofb1oSVE/XjByJAXc4vP5yMQCODxeLK2ezweGhsbhz3O7/fT1taW1R7I2hYIBPD5fFm+oQLBaCHSfghTQQGKcyDaPe0TKimK8WBIo9jtqZyiQoQKBIIzQ3E6QJLQolHQdGAgV2giwxKKJBEPBNAikZEYpkCQH0uo3+8HGOL/6Xa76RomPYTb7aY7Y9kSMIKO0lZQgLVr11JfX4/X62XZsmXU19dn7R9MR0cHx48fz9q2Z88eACKRCOFw+NQu6gyJiD/2LC72+QgdOozJU0xcVdGjUQAincl73uR2E4vHjbZqOEz3pk04pk4lHI2mKppculzs98bpIuYjGzEfA+SaC8VhJx4OEwmHUXQNKeUSpEWjhHt6kK1WDv7yNxx7/Enm/f7/oMQzpI8LFXFvZJPv+bDb7SdvlCIvIvREFsrTsV4++OCDrFixwlh+X7JkieFjCkmf0aVLl7J3795hA54eeughvvnNb57yOQWCM0WLxYgeOUrhvDlIJrOxPZ66583FRVntI+2Haf/RQ1R8+IOMueP2fA5VIBBcZCgOZ9LCqSfTwWUlrO/tw2K1Yk6tLkaPHhuRMQoEeRGhwwnC0xGgK1euZMGCBTQ0NAzbZsGCBQQCAVpaWoa1ht57771ZEfWQtIQuXrwYm812Wgr+bMjXeS4ULsb50EwmLv/pD4gcPoLN5UQ2J4WoGkymSLF6PFitVqN9OBZLbi8pvijn40wRc5GNmI9sxHwMkDkXJpcDPRbHYjJjslqxeQaCIOVIFKvVirO6EgCt039RzuPFeE1nw2icj7yI0LQvZyAQGCJIa2pqTnp8Y2MjJSUlQwRocXEx69atMwRnuu8Tidvy8nLKy8tPffACwRkim80UX1FLz5athgAFiKXcU3JVS4Kh9eQFAoHgdLny8d/Ss3XHgCU043sl3hMEwFJaCkD48OH8D1AgIE+BSbW1tbjd7qyAIkgGGJ3IfxOSfqB+v58VK1ZkbYNkXtDMyPh0/yJVk2A0EA8Gk+lQ9AHfTi2RIJ7KE2opK81qr/b1A2QlmBYIBIIzwex2o9is6KnAJEvGj950cKQ19R0UPXIUXdfzPUSBIH8Vkx544AHWrl1rvPf5fNTW1hqC0efzUV9fn3WMz+dj3bp1eL1empubaW5uZtWqVYZldXB6poaGBpYvXy5SNglGBbsbfsCri95DPCMlStzfDakve2tptghNW0IVYQkVCARnSd+uPQRaNqCnSgSbc4hQS4kHJIno8U70jCBJgSBf5C1P6IoVK1i1ahVr1qwBoLW1leeee87YP7jkZiAQ4NZbbyUQCBjHpEn/YmtoaDCS4Hd1dVFTU5NlMRUIRpL+1r1Iiow1I0do9Hin8XqwJRRJQnE5MRcJESoQCM6OfT9/hCNPPM38//svIBWYJMugacRSmWckRWHOz3+M2eNBi8WRLZaRHLLgEiSvZTtPJBAHR7rnStF0un0KBCNJaN9+rOXlKBnBR7HOgZRkltKSrPbVd92Jq3YuDu+kvI1RIBBcnCguJ5CswgYkqyYVu4l3+Q1LKICtsgI1GkNPCEuoIP+I2oB5pr9tH9u+8GX6W/eO9FAE5xEtkSB84CDWinKkzKCkzgFLqHWQJVRXNZBlJFn8WQoEgrMjXTVJzch9nfYLzRSh0Y7jdL/xFonQ+c2RLRDkQjzt8kzflq0c+/2TtP3s5yM9FMF5JNS2Dy0Ww1pVmRUZn16Ol8zmrGhVgM4XXsL/8qsgqiUJBIKzJB3gmLaEApiLk2ma4hkZZI7+6a+0/uAnRI4czev4BAIQIjTvlNx6M+ayUo4+9aesLwLBxUX/nlYgudQlWzItocnleGtpyZCKSEeefIaOPzyNpIg/S4FAcHakk9OrGRZOcyqNYSzDEpr2WY+0H8rb2ASCNOJpl2ckRaHsHYuIBwK0/27dSA9HcJ6oePc7uPKJx3BfMT9reyxlCR0SlASo/f0oTjuSLCyhAoHg7LCWlmCtKDei4yFzOb7bCPC1V1cBENq7P+9jFAiECB0BPLfchGyxcODX/4sq0mJclGjxOLJJGRLpHu0cXoQm+vpRHE4ksRwvEAjOkjEfeD8LfvcbnFMGCsKkl+P1eAK1P5mX2DamGoD+vSJO4WJmtOaBFSJ0BDC5nBRfcxX9u/bQ9fcXRno4gvNA+2/X0bN9J7JlIDJeV9WM5fgcltBQP4rTAWI5XiAQnAMkkwLqgCU0M1doeknelrKEhvcfREsk8jk8QZ6Idhynd/MW1Gh0pIcyBPG0GyGq378Y7xc/g7m0FF1VR3o4gnNIoj/ElvtXcqTxcWTrQN69aGcneupL3lZVmXWMlkigRaIoTqdYjhcIBGdN5MhRDv12HX2trca27KpJyRSIis2GY9JEZIsFLTL6RIrgzNF1HU1ViR4+jNrbhxaNjfSQhiBE6AhhLS+j9IbriXd1Ee04PtLDEZxDghs2gq5jHzcuK0do5PAR4/UQERqLUTR3Do6aSSIwSSAQnDXRjuPsfWgNfTt2GtvMnmLjdazLb7y+7EermLDs42ix0SdSBGfO0af/xGvvvJO+7TvRdW1IMOxoQDztRhBTYQE9m7ey5UtfRY1ERno4gnNE91s+ABzeSVn+nZHDAylQ0ktgaUwOB5O/8kUq3neH8AkVCARnjSmdrD7DumktLzdeZxo/JLMZLR5DE8+hiwY1FGbHv36bvh27RvWPCyFCRxBJkujbvYeOvzaz9z9Xj/RwBOeI7jdbkBSFgulTs7ZHjqQsobKMJaOUZxpdSyWrFyJUIBCcJelk9VokagSlmFzOpN85EMsQoeGD7Rz8n7X0bt+R/4EKzgs7vvltwvsPUHnH7VgrK0Z6OMMiROgIM+6eD2Ep8bDnhz+ld+fukR6O4CzRdZ1Ay3rsE8ZjLirK2pdejrdVlCObsivm9re2sW/1o4R2tyIQCARni+JMWkK1aBQy0jQZeUGPdRjbYp1dHP9rE/43W/I7SMF5oaP5b+z/xa8omDmDitvfiWK3j/SQhkWI0BHG5HBQc//n0cIR3v78l9AyviwEFx5qKEzxFfMpvGwWst2WtS9dkWTwUjxAuP0Q/pdfJR4M5mWcAoHg4kZxOkCS0KLRrFyh6SX5TEuoa8pkAHq3bMtqK7jwiB7vZPPnvoTidDL+Y/+IraL85AeNIEKEjgI8V19J2a03E2jx0fqDn4z0cARngcnpYPq//j8q33s7ssWStS/tE2odFJQEyRyhAOZUqT2BQCA4GyRJouyWm3BMHJ8tQlOiJNMSainxYC5207drt4iQv8BRnA4KL5vF2H/8IK4Z00a9e5fp5E0E+WDS5+8l+PYWQq17iQeDQ5ZyBRcGWiJBrMuPYrVkRSLGA0HUvj4gtyU0kdo3uJ68QCAQnCmX/+wHBHwbQRtIVJ5ejlf7+kj0hzClfESdNV6CGzYRDwZRHKN3+XakCe3bT/R4J4neXmSLBcXhwD52DJay0hGNPtdVFWSZ6JGjjP/4PShWKyaHY8TGc6oIETpKMLuczH34J8S6A/S37qVgxrRR7cchGEqiP8Qrde+m5IbrqXz3O7L2ZVYjcU6cMORYNWUJFSJUIBCcKyRFQVLknJZQgGhHB6ZJEwFwTZtKoMVHz+YtQ1LIXaqo4TBdr7yGqbCQotmzSIRCvLn0Hwm1Da0uNe0bDzDuIx9CcdjpeXsrRXMvRzab8zLOeE8vGz5Rj626iorb34UkS1gy0nGNZoQIHUWYi4qQzGb62/aye9UPmPLAl3HPuXykhyU4RY43/43+XXtwz59nRKCmCbXtM147vJOGHBvv6QHA7BYWcIFAcG5oe2g1vVt34P3McmObNSMzR7TjOM6UCC2+agGxzk4kk5AFvTt2ceDX/8Oh364j0dtLyQ3XMvn+z6NGIpTV3UK8249it6OpKlokSqyzC8lsIuDbiK4mWP+hj2MqKKD0lhupuvM9lC+69bxZl0MHDtLyoY/Tt30HpbfciB6PG6VYLwTE3TbKMDkcxLv8dP7tBfyvvcn8//4FpTdcN9LDEpwC7b99DGSZ4quuGBr9vncfAIrLhaW0ZMixRXPnoMfjKEXCEioQCM4N3a+/Rc+WbUz69DJjm7ViIF1P9MhA7uLCWTMxFxdjKnShq+qo9yU8HwQ3vc22f/4G3a+/BYB93FjK33UbRZfPRo1EUewOqu+6Y4iFU9d19HgcLRol2tlL1eL3EtiwiaNPPsPRJ59Bttkoq7uZub94COUcWUd1VWXfL/6LXd9ahRoKMeaDS6h87+1YK8pHZVL64RAidBRSfNUVTPv6V9n14Pd4a+k/UnP/55jy5S9eMl8Kuq4TOXSY8IGDeBZejZZIcHjdH9j/6K/QYjEUhwNTYQGOiRPxXHs11Xe+Z6SHTPRYB51/e4GiuZfjGD9uyP60JdTpnZjzC8Kz8CoKpk9BsdmG7BMIBIIzweRyokXCWcvxlhIPss2GFokQ2n8gq71itxML9BDv7cXidud5tCND/959WMrKQFOJ+f0EfJvwXHs1JddfR1HtHMyFhSddVpckCcliQbZYMBUU4P3cp9Hicfr37qfrpVcIvNVC/542erdsw1LiIbhpM7qqUXbLjZjP0AVry1e+xsFf/w/WinIm/tMnKbl24QW5kiZE6Cil5PprmVnsZvd3v8+eVT+k49nnmP/rR7CPHTPSQzvnaNEo/o2bCbT48L/xFoH1G4h1HMdUWMgVv/8/9GiUnq1bCe0/gGwyoUajqP39dD73PL3btlE4fRqKy0l3iw9bZQXFVy7I+y/Bfb/4Fbqq4rnmSiNJdBpd0wjt3w+AI7X0NQRNS/lvXRo/NAQCwfnH5HKhJ9RkrtAUkizjmDCevp27hojQ7jfepO2nDzPv0Z9T8c5F+R5u3tDicTr+2syB//pvOp9/kckr7sNz9VUkenqZ87MfYKuuOuuYDNlspmDqZAqmTka750PE/AHCB9sJtx9mz3/8iN6t20GWKZg+leJrrqb4yvl4rrkKe8ZSuq7rJHp7CbXto3f7TrrfbGHav/wziZ4eCmZNp+I976Live/GOX7ckGwsFwpChI5iimbPYs7DP2Hvzx6mZ9PbhI8cRbZakUxK0n9UvvAybOmqSn/rXoIbN1F6y02oskTvpi1sWPKhZANZwj5uHKU334hj0gQi+w8gO+yU3XwT5XW3QkpcaokE0aPHUMNherZtR7KY2f61fyV65CiOSROZeO8yxn3o7rxYFnVN4+gzf8I2phrPNVcPEZLh9kNG2hPnMCJ0xze/g2yxMPPhH5/v4QoEgksEpSD5gzjRH8ra7piYFKHhQSLUVl2FFonQ8WzzRSlC+3bu5sB//x+H1z1OrLMr6T515RVIJhNaypdysCvVuUA2m7FVJH1xtVgc7+fvpevlV+nbvpO+3XvoffRXHHj0V0z7l69Rtfg9SIrCi1ffhBoKDemraN5cLB431vJyJi77hFGe9UJFiNBRjtnlYupXv0yko4PoocMkenrZ/8gv6dmyjfJ31FF+Wx3FV87HUjx6I+EO//4Jut9aT3DjZnq3bkMNhQGY9vWvYp81k0QoRNWS9+H0TqRgxnTMbjeKzYZsOfESiLXEY7xWo1Em3/c5Opqew//Ka2z7ytfY/eD3mPDxe/B+7tOYClwn6OnskGSZeb/4OV2vvJrT37Pn7a3G64KZ03P20b+nDUtZKcjCEioQCM4NpnTVpPBgEZrM0BEPBIkHgsYybtIv1M2xPz/LjH/7xgUvcHRNo3fHTgpnzkCNRDj8hyfZ9/NHsFZVUrVkMSULr8ExcQKmAlfejDqyxYxrcg2uyTXoqkqiv59Q2z76WttQnE6CGzej6xquqVOQLGYUqwVLWRnW8jIKpk/FVl2FqcB1wVo+ByNE6AWCrbwcXdOIB3tQnE70WIyDv/ofDv7qf5L7q6uY+d1/p6zuZiSTifb/+W1SyNntKA570iIoSbgX1KJYrcR7egj6NqVqCuugaWjxBEgSFe+oA6BvTxvdb7yFnoijJ1T0RAJNVbGWlTJm6V0AHPtrM/6XXyXm9xPr9Kf+76Li3e/A+5l6tFiMtp/+nJ63t6I4nTi8k7CPH4d9TDWWslK0eBxzYQGTln/irJaiFasV9/x5uOfPI9rZxdGn/sixvzzL3od/Qfm734mrZtJ5EaLR451IskysswvnhAk5fYfSIlRxuYwv/8Go/f2YJk5AMgkRKhAIzg2lN99IPBBEtlqzttsnjDdeh/btp2huMguLpCiU1d3C4XWPc+yvTYx5/+J8DvesUcNherfvpOOVVwm8sZ6eN1uIdXVx5ZOPIUkyzhovUx74MoWXzcbsLkIZNC/5RlIUzIWFFM293PgMdFVFV1VmfOtfko1kCdlkumhdtYQIvYCQZBlLsZuaz32aCZ/8GH27dhFo2UBo334i7YcI7TtAoMWHGomy5f6v5uzjyqfWYa+sILh5Cxs/de+Q/YrDwbXNfwTgyJPPsLvh+0PaFMycjmvaVHRV5fC6xznyh6eS4zOZMBUWYCooIB4MEti4CS2WoOr9ixl3z4exjR2DYrUi26zIlmQy92jKV+lc/oFZS0uY8Il7GPsPSwm+vZXw/v2ovT0ceeqPaNEok+///LBi8HTo37uP195xB+XvXETZLTdhHzc2Z7ueLUkRWjhrRs7r1GLxZMCVy4EkLKECgeAcUV53M+aiAqLHu7K2OybmFqEAle95F4fXPc7+R39F9fvuGJVuX/FgkL5de+jbtZuieXMomDaVRF8/z82sRYtEko1kCceE8RRfcyWh1jaslRXYqqtxTq4Z1dHjl1psgBChFygmhx333Dm4585BV1W0WAwtGkMNhVFDYSavuA8tEkWNRNBjcdRYDIDwgXZixzuJ9/Qy9p6kH2b6D1JK/drq2b4DAHOJB+8XPmMkPJbk5B+HbLcR2n8ACSi99WZKrr8WU1EBisOJbFKQFBOSyYRsTv7vzPjCyyeK3Y7nygVosThRfxfdr79J77YdtP9uHWW33My4j36I8kW3npEPUPebLay/51PE/N1YPB4s5WU5vziixzuNNCiFl8/O2VeiP5Wo3lWApIy+L3yBQHDhIpnMoKpZ26zl5ZgKXCR6++jdtoOqxe819tnHjsFz3ULCB9vpb9uLa3JNvodskJkqau/qRzn2p7/St2t3Vt378Z/8KNXvuxMtFqPs1puQTGYs48finDYZV0UlitMx4hZPwfAIEXoRICkKit2eFc3nmDA0TdBg3HMuO+F+58QJMH/eCdtcCDWdZIsZe2Uls77/XfyvvM6RJ57iePPfON70HJYSD1f+4TEKZ+X21RxM746d7H3oEdr/by2y2cyke5dTestNw5ZH87/ymvE609qQiVGy0+W8pH4BCwSC88vxv7/I1q/8M1Xve2/WSo0kSRReNhv/q68n0wXpepZ1cPKXv0i0s5NETy9aLJYX/0Nd1wnt3Uf3Gy10v/kW3W+0YB83llmrvkUiHKbrxZcJtKzHNqYa17QpWCsqsFVU4KiZSDwYRLaYmbj8k8g2K/GU6LYI8TnqESJUcMmgWCyU3XwDpTdcS1/rXo7/tZnenbuIdR4nuDlB9xtvse+R/6Jg2hSsVZVYSkpAU3FMmED1B+5CC0fY/Nn7CW7YRMHM6Yz5hw9QNOdyo/ZyLjpffBlIVilxTZuas42l2M3Ez9RTOHOaEKECgeCcocVihPbuI94dHLKvaO7l+F99nVhnF5HDR7JSA5kLXEiSRPjQEY7/7XkmfOIeLB7PkD7OBl3TjKX+Y39+lre/8BViXQNuA5YSD45JEwhuehtd06i6607G3fNhI2hVtliG/74cZPkVjF6ECBVcckiKYuRvUyMREr19hPbuo7+1DbW/n2N/+mtW+6LauTgmjkeLJyh/122U3VZH0bw5WD3FJxSNsS4/PZu3AFByw3XD+iGZXC6Kr5iPfXxun1KBQCA4E9LR7Vo0MmRfYcZKWM+mt7NEaPrYgG8jux/8Hgf+67+Z+v++ypj333lGVlFd14kePUbP1m30bN6K/7XXCby5nqv/9DiyyUysO4CpsICi2jk4JiWzpNjHViPbbChWq/hxfhEjRGieyaxcIRh5FJvNyCVqHzuGyve8m0R/P3F/N4meHnRZxuSwk+jtQzKbKVl4NbLNdkqO7UeeegZ0HYDSm64ftp2uqugJ9aJJuSEQCEYH6cIZamSoCHV6J2EqKCDR20vXy69S8e53DGlTcv1Cxn/yo7T/32O8/dn72LbyaxRdfhnVH3g/41MxBf433iKWyhKS/FHfS6K3j5IbrqNg5nS0aIwXr7qBSEaJUMlkwjmlBv9rb2ApKcHiLmLWf3wnmclFLKFfUggRmmd2f+PfibYfYtyH7sZaVjrSwxFkICkKJpcTk8uJraL8rPpK9Ic48oenAXBNn4pr+rRh2x554mn2PvwL5j78U4rPQdS+QCAQwIAlVI1Eh+yTZJnSm67n6NN/ovut9USPdw55JkmSxLgP3U3ZLTdy5Mln6Nm0hYBvI4rLReGsGchWKzu/+R2633hrSP/jP3EP1XfdiZaIY580EdeM6diqq7CPG4tr6hTMhQXIdpsQnZc4QoTmkeDGzRxd2wi6TrDFh3vpEqZ+8C7DEnckGKKqyGG8Boz3g7dlts0kvX3w/4PZ1O5nzliP8f+J+hk8jsHjyTXO9LEdvRGj/1ztj/ZEqCy0Ze3Pda7hGHx9J3qfHk95ge20rudMxnLgl79GTUW9j/3Q3UiSlPOzOBIMkejtA03DVFhw2ucUCASC4TAVJL9T+nv60VWVo33RrO+g8nfdxtGn/wSaxrE//xXzne/L+R1VVVnJpPpPocXiJPp6ifsDHNy1j1KHGWXBlUy4fDa6piNbzCg2O4rTQZ+rkHhPD7LJzJSv3Dfgx5kj5dNw39vp//++8yg3T6scsn1Tu5/yAhsdvUlLb3lB8lnisSlGP2nSz6LhvofP5Dv/dPoY/Cw60XMrfV0nag/DP6cy22vxOP6XXiVysJ2Ka64GYH9XHxNKzl8Bl9NBiNA8YinxULLoFrqefQ4tEsH/3//D648/wZjb38HGsTP5/TGN+xbNoO14H09sOgjA4jnjWFI7Ht9BP99v2g7oLJ47jqc2HeLOOWNZUjuQ/sh30M8Pm3cwpdzF7o4+rpzo4c19fu6rm07tuAGh+dO/7+DVtk5KnBa6+mMs9JbyuZunD+nnyokeXt/bCUh8adGMrD4afQf4w8ZkyTdJkoxxNvoO8OSmdu6YM4Y/bDiIDiz0llJV5OAPGw8CutHeW+bih83buX12FVMr3fygeTu6rvO+ueOzrms40udKX9/g9+nruHNO0tfy8Q0H0AEJ+PJtM7Oux3fQzw+at6PpOhKc8hhyjWXc7rc58kTSCnqkbCy2sVPYnzGWdL/p8X2k7RBOwOwZvVWvBALBhYe52E3XJz/LX/Yc5eW/b+eNgz1ZzwPX1ClI4yegH9jPvv9bx297PNy6cJbxHTX4O1W2mLF4PDy1r8/4nn8qWMydcy7L+Sy6U4mypLbihGMc7ns8/fwqspvo6o/x+w0HCIYTyXNuOkRloY32wNCylrIEn79xCgA/fXEPmqajp/aNdTs42hPJ+T08+Hl6Ogy+hsEMfhad6LmVfj6nn1NA1vjSz6rhnpWNvgM8teEAn/OaKdm6gc7nXyLe3Q2SRP+XvsAmexlfavTxiYU1fPrG3MGy+UTSdV0/ebNzw6pVq/B6vfj9flpbW3nggQdwu91ndcyZ9DmYrVu3Mnv2bLZs2cKsWbNO/8JOg77uAFv+81H2/uJXFPV2Z+0LOt0cqprIsZJqjrsrCBQUo9rsfOxqL796vQ1Vy/6oTLLMqrvmUlXkMG7yRA6fU5MsGzd48/YjPPpq65A2n7y2hrrpVcP2o8gS99clheiRYIivPL4hazyKLPGVRTP4XlPuMciShJZxq8lSUryqmo4iA7qEmtqvyBL/cde8k/6qXPH4RhKahkmW+eg1k/j1a3uN9+kvqoSmocgSuqaTOSpZgi8tmmmI1R80bx9yPScbw+CxaPEYC3a8yZWbXwBdJ2ay8Lt3f5L+wuSXkqrpxmd2pCdizPM7X3mCyfu3ccNbL6FUVQJgt18Iya/OL+FwsryrmIskYj6yEfMxwHBzsb+rj4/94GkqDrXRb3ehmsxZzwPfQT9r/+cv3NGcrLx3tKSaP936Ib71waTFLPM7Nn1M5ndvmuGeRZnbc3Gi7/GzQU657GvDqJtc38PpbRV2E4m+PiPnthoJo4YjqKEQWjiCFoslYztSz6uecIw/bD1GTFbQzRZunz+JKWNKUhULbWzvjrL6jYOEFROYTJB67plkmY9ePYlfvz7w3LplWjnPbs/wnQVkObt9phZQJPjuO6dRHO4ltP8AHTtb2fTaJio7DmKNZ7tgmIoKUb76db7a5SauapgVmcbl14+4RTRvInTVqlUArFixAgCfz8fKlStpamo642POpM9c5FuE9vg2sv3AcZ5+8mVm7XiTMccPDts+bLXT6ywibHUQsToIW+1ELHYSJjOzJ5SyoKaS/b0xntnRQVySQTFxjbeU1/Z2kUh9tDoSiiyzcHIZr7R2prZLqT2ArmOSJBbWlPDank5UXUPSdSQdZBkkXUfTdBRJYvGcMXhLXby65xivt3Um+9B1ZOA9l4/hWDDMW3s7jX4BFE3FpCWQVRVFTWBSEyha6rWuoiSS+0xacrvXbcNlktATiWS5UDX5v5ZIgKYlk+ebFMIaHO1PoEoSCZOFsjI3+0IaEZOVqNlKzGxBtdqZM6WaVw71EFEsxE1m4iYLcZMF1WrlvfMm8OTmQ0MEaFpwD4eWSBDvDhDr6iJ8sJ22V1rofasFRyS5BI/FguWL9/Pjw8rAF0aqXyDri+/eDU/Anl3c5HsNvagQEA9WECJjMGI+shHzMcCJ5uLvr2/j0dV/oNfqJG5OBj8O/qF+y+t/ZGbbJgCkkhIm/eMHcc+fx7aEhR/9bdcQIbrOt5/HNww8t+6aN46ltROGCNDhLIOZDD4mc1zoOoqmImsqiqoOvNbUrO3JbVrq/0Rqm5bVPt3WosW5oswOoRB7D3ZgiUWwxqMUk0AKh9Dj8XP1sQxFlsFqo09SUs8iK0VuF0ejOlHFjKokRaoO6JJE8jkNJk1lfIGZY109KIkE1lgER7QfVywMicSwp9MBZcZMxty4kEOTZvDvhyz0KxbMisz3l9Ryw5QTW6nzQd5EaHFxMevXr8fr9Z5w2+kccyZ95mIkRKhJ09gakflh8w4cwS5qDu9hZuc+nIf2Y0n8//buPLyt8k70+Pcc7ZZsy/ISO/ESy85iAiRxHJawFIpJO6yBSdrpRldwh0s7hXYS0rkdhts7pUlv4U6H0rqdcmfodAkxLd1h4g5laaA0UQoEAsR2FmdPLMurdp37hyzZx7sdW5bt3+d58jg6R+foPa+P3/en826haU2D6BdVVKIGIzFVRVMUYoqKI8OC2dS3Vq+qQixGLBJBC0fQImFi4QixYDAZZA/Wlp3HH9bdworLV/Hcu6d1Qei1Sxfwh3fP6App9Z//iVCbl6tffo5Y35yjUrFKkDGY5Iee5Ee/0fLihSveS4+vkx9c/VF6jcPPwKFGo7x/9y9wt76t3242ozmdnA0rBI1mokYTLruFtt5wvGVJUdAUBUVRyMu0cq4rgBaLYUCjPNdOltkQf2oYi/X91NBiMbRYFDSNWDiCFokQ8Afp6vEng0mjFkWJRFBT11A7NygK1qJCwuUVNMacHCp0E7JncU2hhVcOt3F4QTmaLSNtAlBIUZ9Qj8eDz+fDNWiyW5fLRUNDQ/JJ5kSOqa2tnfA5Ac6cOcPZs2d125qamgAIBALJP+bpEgwGCIWCxGIaKwpyuefqCh59QWNf5lr2LVuLEouR03mOnE4v2V1enF3t2P1dWIN+bMFerEH/nAhSo6pKVDUSNRiIGIzEVCPZWRlkZFj6lg81ovQFgonlRBWjEUVV41Ma9T0d1SIRuroDnG7rwBQOYg4FMYeDmKIjfzscyKDFMAzKTy3Qw9CxpKNTrVYcK6rovriGpwK5hBSVk2/Hm1UMavypczSm0di3zagq3HN1BSsK7PR+4R4iXZ0EQyGismxnUmCYaWXmM8kPPcmPfqPlhWI2YwwGufPyMr699xSRYdqoVZORJf/w9xS/uYcTP3mS8Ln4pPGxUAjOnGHwPC4jNeAOHFoZOw6+CVxDqhqFo4pK0GwlZLYSMltYUODCmZuNwWHHYI/PjmKw2zFk2FCttr6f1uRP1WKOT9E34EmlFg4TCwZ5q7WNJ19pRgmHMEXCGKNhrNEw15VlU2RRiAb88Sb9QIA2byeHT3gxRkKYwmFMkRCGaCRxRhQNlL7erFE1Xk9GDEZiBiNFhS5chfmYnNkYndmYcnKwlZZgLV6I2jfbgHLMx6MvNBGJxfhjS3xsh8mg8E83X8ja4qxpjXUm8sUwJUGo1+sFGNJX0+l00jZghYSJHDOZcwI89thjPPjggxNI/fRaVezkqsp8nnv3DACaqmIrK6O5Y/gpgq5dWsDHqhfxo5eb+eM7JzH2NXFfXpLFTUtzIRoj/nBbAw2eeesU+1q9yRubvht71SIn65cv4L/eOcO+Y75kE0D8p8Lq0hxuuKAo/jemqKDAb946zd6j7Wh9711b5uLmixbyyzdP8eoRL6Ak911angsovHLYS8xgSP4RRVVDsslh8HVdd8nkpyf6j1ePJPMQoNhh5lybD3M4iDkc4tJCO+sXZxENBHjxreM0HzuHqe+P3xgNo2gaJVkWlubbIRqNB7qxGFo0hmJQUYxGVKMRxWSK/99qwexyYXLlYMrLxba4LLkG/RWD0nJ1ZT6AbttVlfmsKnYC8QFJJmc2ilEmZBZCTC1jVhbR3oNctMChq2sGuqoyn9WlLihdT977aul5twn/4cP4jx4j4usg6vdz4pSXrm4/St/TySyLAZfNhLcnSE8wXoZqioLdaibXYQFVjY+EV+I/FYMKihLf1rdPMRqT/1STkXe9fo52hYmpBqKqij3Dii8U63ttSNYlideJ/+dk2Tjtj/S/b+DPvmOuXFqIZjDwXNO5ZP1z7dICbjiPemew1UWFvI5Dl8fXLi2gepjPWAzsH1RXLMq2cbzDP+LrxPluGkeaB8cWAO+vKuTKivSaGjIlQajP55vwvrGOmcw5Ae6++242bdqk29bU1MSGDRuwWq3T3rQTDQQJmi0YYzEsFgueVi8vNp3TvWfwTTfQi03nyLFbeL61m4jVntze2KVyUc6iISO+n33dT2TB0Kl/TmkqZvMCnolFiBQ6h+w/G1JZkVeSPJ+n1csufxeRXGvyPbt6VGx+G7u6TESd+qB5V3vfwKOs8S319kLTWWoW543Zf2g4w+Xhse4Q9PWjBXg2rHJBWXwGgF+2WoguKx9ynvH0BZ1MWl5oOjvkfS82naNmcR6rFznpPXIUc0E+NruDcF+Pemli7Cd5oSf5oSf50W+4vLDk5BALBnn7ZMeQsikhUR4lyj7ryotgwIpKnlYvPxw0YHVI/80B28fTF3QwT6uXJ0cYXDuWozDmo9Tnj/YtXTrgAcjg6z5fw5X/I33GeOr+4WKB8daVw53/mQOnec/qzrRpigdISdvfSKPVRwsWxzpmMucEKCgoYMWKFbp/lZWVox4zXQZ3yF5ZrJ+iZ+CzQlWJB0mRWIyf7WtNHlO7vBCjqhKJxXik8W08rd4h5x7OwPOMtD9xvsHpHPiZP9vXmuzzqCr6UYnRmKbbNpzErmhM4+HGA8n0j9dYebiyOCeZ1ocbD/DNXW8NmWUgYbJpGCkttcsLMfSNbIzPAqAM/X29c4y37v8Kx3+6U56ECiGmnMkZH+z4w9+/Oa7yfrCxyv+x6qLxGKscnwqjlsOTLPMHGi2fBn/GWNdbnmvXvVbor0fHU08NPv8V7jwMKoSjMb7Y4OGFg6fP+3qnSkqC0ES/zeECxIqKikkdM5lzppP9J31DRgS+eaJD9x6N/uAzpsHgMWS3rFzEp6+o5N7a5bqbfafnyJAA1Kiq3L66BOMwEwWPtD8RuD3ceEA3kObTV1Ryy8pFuuMTUx5tWFWi2x7T4v9GCkRVdWJ/XAMNN6pycB6+eaKDW1YuSgaDA+NPg6pw++qSvj6bk0vDSGm5t3Z5vHlrkNWlLt3v6/FdrwPEVw8xyrS9QoipdUozAWAM9s+pOVJ5P1awNFL5P1JdNJ5ydDzl+FQbXA6fbyA6Uj4N9xnjud5DbT2616qqsGFVf101Wj01XFo+WF3Mpy8vx2RQ0y4QTUkQWl1djdPppKWlRbe9paWF2traSR0zmXOmi9OdAb7/YpNubrSBc1oODNgUReETl7mTgehAv3ztOCc7eqku0f9BDX7CmbgRN1WX8fHLhjZDA3z88nI2VZclz5Mw8NvjwHnifvHaMd3xiqJgUOJpGp4+Ck0E19FYvHXEoOj/uAaudDGckx29uj+0gXk4sICNxGL84rVjaIMyT1XgvtoqNlWXcV9t1bCB6FhpGCkt99YupyjLyiONbyfzLhEEP9L4NkVZ1mQ+m/zxwsZvtsVH4wshxBQ50tbN1xwX8eMbPkOPLf50bWB9MLi8TwRLJzt6hy3XEuX/r/cdxdHTibOzjbz20+z+7z3JuugL60pxhPxEI5HkuUYyWjl+PkZrgRuuHB543RM1Uj4BQ+rmwQ91Pn6Zvt5aX1WoO7cCybrjl68dT8YCiesYXE+NlpYLF2bztVtX6gLRI23dE77eqZayobhbt25lx44dydcej4fq6mqqq6uTr+vq6iZ0zFj709WCLCvrLyhK3iS1y4u4dWUxRlXlvtoqbltVmgxcNqwsobaqiPtqq1CVeICaCLBuXVmcnAR44M1eVZiFUVVZ584bciPWVhWxzh3vmJxrj0/Xsc6dR+3yoiHnWefO6/tj1veVLMrOYMPKkuQfeiKdK4tdyeu4fXVJMuxc587jtlUlyfQbVIXbVpVyX20VRlXhpgsXct/1VckAfMPKkjEniS/Kzkh+1uA8HFzAblhZwm2rS5PpUeifqD5xzYlAVOm7pvGkYaS0VJe4dNvuq63qu9b+31kinx2heJ+f3OLCMT5FCCEmpizXwQfWr6Ezp4DqEteoQVKivkiUUcOVa96X/8Sxu+q48ydf5xO/eJSP/rqev/ndD7jtpZ04u9rxHz9B3ouNfKLhEe7+6de562ePcPLee3nryw9wfOfPkq15Wl+QOVo5nkhPop7KtZt1DxiKncOXz4kVkz7/niXxMn3AvmJnxrDl8OD6dCKGy6eBBn7GhpUlbFhZ0n+9Vfp665PrKpP1c2LFpIF1RyIWGKmuHCstV1QU8M2N1ZgMKp9aVzHjE9XDDKyYlOjLOXh1o4aGBrZs2UJzc/O4jxnP/vGYqXlCveaMUdc5h7m/dvyRsx0UZlmxWCwzvnb8CV8P0XYfkTOnyY6GifT0kH/dNSiqSjQQIBYKYxplffex1iQebv87Db/k3He+y4pvfI2yT90hcx8OIHmhJ/mhJ/nRb7S8iHT38ObPf0sOEXpLy8dVbyTEQmEOPP5DKjbcQCwcIXDyFIe/+29YFxYSycohy5lJDyq5RQUsuOF9aJqGb99rtD3/Et3nvBh6ugi1eQmePUfO2jUsvutToCgcefwJelsOkVFRTuaypQQXlVK68gJMfYt1TNXa8d5ANHkt83XteIDu06dBUVhw+WUYMx1ptXZ8SoPQdDVTQaglP72mSpgJwWB8Rk5L39xmM8F//AQHtz9MT/MhYoPmTlv5g++gAL69Hg4/9n3M+XnkXFJD4c034Fhy/gPaTjz1NIe++2+sfvy7FN18g1SsA0he6El+6El+9BstL87+/jn+/IGPUfLxj1B6x0fGfc5wZxdv3f8/6X7nICUf/wjFH/oAlvw8jJmZ8fkyzfHp6hRlaLu3FoslV7yLRcJEenoJ+zowWMxEe/00/8u38b78J4InT+mOK/vMJym89UYMFgsdr+/HUpCPtXDBhLsqpUO9kk4GB6HpREZCiHlFi0bpeH0/5557noL1tZhdOYQ7OultOYTdvRhbSTHWRYsw57kw5+SQXbUcTdOIBYLk176Xzjff4vRvnuH0b57BeUkNlfd+DktB/qTTU3T7rWStXkXWiqopvEohhIgzZsWfLkZ7xz85eSwUTgagRbffQvn/qCOjtATVZBrX8YqqopjNYAYDNkxZWdiK+rscrfrut4j6AwTPnKHjtTfoeP0NuvYfIKPSTbSrG/+x47z5918GTUMxGLAsKMC6aCHWhQspueNDmJ3O+Cp20SgGCTRnNQlC55BoMEjY5yPs6yDa1U0sFCYWDmMtKsSxNP7UzvvKq4R9PkDpX/VBgcyq5diK4yMe217ajZZYo11V4z8NBhzLlmB0OIiFw/Q0NSeXtVRNpvhKE5kOVLN52G/GM0mLxeh68wBnn3uethdeItzuA8DkclH8wY3YK9xctfs5TJmOvhUxLEOuIXP5UhZ94HYivX7ad79My2Pfw/vibg5/73Eqv/R3GKzWYT55HGmLRlFNpuQqF0IIMZVM2dkARHvHP+jmyA/+PRmAVn31AayFUzuvpGIwYHTYMTrKsbvLWXjbLWjRKNFeP1G/n2BbG5V/fy89zS34W48ROH6CDs9f8P15L/m11xJu99F9sImmbQ9jynHGg9TCQqyLisgoKyXrynVTml4xfSQInYVikQg9B5vo3P8WqtlMwfrriIXCtPzrdzj33PND3p933bUUf/gDKMDR//cEPU0tQ95TfMeHyXvPVQAc/Po3iQ6zpNeS+7+EY2klYV8H++/bMmzaVn7vUcw5TnpaDtH6w59gyc/HXJCHpaAAa1EhGSXFWBcWnV8GTFDT//kXzjy7CwBL4QIWfWgTRbfchHPNaoxZmRP6dm9y2ClYX0t+7Xs59avfYs7LJXj6NOa8fIz2ifcpanvhJaI9veSsTe/BdEKI2SnRzzLqj6/rrowwTV9Cd1MzJxp+jn1JBUvu/9KUB6AjUQwGjJkOjJkOLAX5ZCVaoUIhYsEg0V4//tZjmPNy0UJhtEiU3PdcSeDkKQInTtL99rsAWIoKqSwtRjGaOPfnvfj2vUbmBcvJWnEBmVXLMEj3jbQiQegsEQ0E8O5+hXPPv4Rvz15igXifF2vxInIuXYtqNiWDqsQykAarFYPVSkZZGY5lSwBYsuWLRLq646MTtfgynmgx7JUVWIsK0TSNpV+5P75mcCwaX589FkOLRsm5bC3mnBwivb2U1X0aLRqFSIRoMEi0u5dIdzfmvFwURSHS00vvoSN07X9Ldx0ml4uVjz6MYjbT+cabdBx4G1tZCVluN7aSYgy2yT1RhHgTUsdrr+Pd/Qo9TS0s/cr9RHt7yV51MUaHncKbb8R52VrMOc7znpNTUVWKbr2JSHcPXe8e5PC361lw4/uxu4efAmskrf/5U2LBIOV333Ve6RFCiOEYk0FoL1o0OmYQqprNZFevYvHf3ol98dQtaTkZiqJgsFgwWCyYsrJ0AbFj2RIWbtxALBRCC4UJeb30tBwi0tmFaXEpka5ugufO0bF3H75X98TPZzDgqFpG0S03kX/dNTN0VcOL9PTiP3KUkNdLNBBEC4dRjEYM9gxy110208mbNhKEzhLeV/7Mu/+8HRQFx9IlONeswrm2huyVF2EtXIBqteC6/FIUk2nUACvRLD+azOVLx3yP65Ia3WtN0+LrrUejEI2RvWolpR/7MJGOTvzHT+BvPUbv0VZifj8Gh4NYKIT35Vc4+1+/153HnJdHwfuvp+Sjf4NqMtG5/00AjNnZfd0C4gWoMTMTY0YG0UCAI48/QfeBt+l+twktEgHAsqAALRjEUVmB65IaTK6cSTeZj8boiK81f/IXv6bjjf2sfPQRVLN53MdHOjqwLFggqyUJIaaFwWrF4LAnHyiMxehwsHTrl8i66MK0n7s4EaRisWDMdJBRVgrEB2ppmkb+mtUs/6f/iW+Ph/Y9Hnx/+jOdb7xJ75Gj9Bw+gjEjgxNPPY290k3O2hpMzuxpT7MWjdJ7+AhdB96h+2ATZXd9Ci0Upu2l3bT830eHvN9ckI9t0UIUk4m2F3fT/vKfyF69ktyr1pFRvjjtur9NlAShaSrS3c2xHz9Jzto1mFw5OJZW4v7c31LwV+uxV7jjK+xMINiZboqioBiN0BcAJ4ousyuHjHL9t+lYKEQ0EGTFN76G94436T3YRPDwEXoPHab36DFi4RDB02fQIhGaHv5X/EeODvm88rvvIueyS4gFQ5z65W9QTUYyL7yA7FUXk197LdkXX4TJ6ZxUE/lE5axdQ+knP8bRx5+g9Uc7KPvkx8Z1nBaLEe7oxLFsKYpxfF0ChBBiot7z6ot0vv4GWiQKo3Q/7zl0GE3TcFRWYM7LTV0Cp4GiKBhsNmw2G7ZbbqTolhuJRSKE27yEOzqIhcL0th7jxFNPx1sFFQW7uxznpWtxXbaWzOXLpiwIj/oDHH3iR3S9dYCeg83E+kbvA+Rfdw22khKyVlxA6afvwFJQgDEzE4PVSiwcRrVYsFe4ifp7iXZ30/3uQTrf2E/rEz/CurCIog03U/BX6zFmTH9dNx0kCE1DbbtfofnhbxFu9xE618aSLfdhKVxA/nXXzImRgKrZjGo2Y8rKRMl1kXvNVdhsNmLhMFF/gFgwmOwHVP63d9Jz+Ajh9vZ4ARqLocXi3QcsBfkY7Blc+rOfYC0p7uvo7hh3H8+pVPW/vsLp3zzDyZ//gsIb3z+uEfORzi7QNEw5TnkSKoSYNgarBVRDvKVqBFG/n9fvuY/MqmWs+eEPZv0TtuGoRiOWBQVYFhQAYK8o58o/PMOZXc9x7vkX6fDs4/iPd3D8pztZ9f1vY8nLI9wen4TfXr443t1shMA06vcTPHsO/9FWeppa6G5qpmD9dWReUEW0p4dTv/oNaBr2JRVkXlBF9uqLcdasIaN4EarVisFqoei2m0fMd03TyLroQpZ+ZQveF3dz6rfPcnbXf3Pose/hP3GSxZ/5xKzs7ypBaBqJBgK0fOs7nHl2F4aMDCq+8DlKPvERrEWF82JdcdVk6gsg+yeET/Sx1DQtHoD2TWurqOqYfZtSyWCzsfQfNvPG579E649+SuW9nxvzmPgsBWBynn8fVSGEGEnHa/s59/s/kDdKP0jv7leIBQLkXHYJ5tyhC5jMRQabjawLV5B14Qrcn/ssofZ22l9+lc439mPJyyXq7+Xkr3/HyYafA32Dp5zZGDMycK5dQ+nHPwpovPv1b9L+8p8GnVzF7l5M9uqLMecsoubH/4G9vAxjdhZGu33CT1kVRUmO81j41xtY+NcbCJw+w7Ef7SB71UpC3naiwdNEuzrJXnnxFOXQ9JOaL01EAwHe+MJmeg42kbXyYpY9+A+41lRjyJh932ymg6IoYDCQzt/Ni/9mE03f/Ba+P+8l3N2NyTH6pMCapmFfUklGWWna970SQsxex3c+xfEfP4lz7RosIwSY5178IygKhbfclFZdvVJFNRqx5udT1Nd0H/X7ifT0olosZCwupedgM8FTpwm1eYn6/YR9PqL+XkDBsbQS1WzGnJdHRlkJWRetwFG1DLPTiSEjA0OGbVqeLFsXFFB53+eIBgIET53mwD9+ldO/fZaiW29i8Wc/MyOtghMlQWiaUFQVW1kJzupVLLn/i9iKF6XVkz4xNsVgoPrf6wl5O4h0dI4ZhNrLF7N065ewV7hTlEIhxHxk7eseFPa1A0NHvEcDAXyv7sWxfCl29+LUJi5NGWw2DDYblquvJO/qK5PTRWmhMFo0ghbrW2xSgZxL1qAYTagm44wE8AarlYzFZVTc+3l6mg9x8ulf0XXgbaq++kDaP9WWKGeG9R45SrizC//Jk1Te+3mqvvqPZJSWSAA6S2VffBH2xSVosRixUHjM92uaJhPVCyGmlTkRhPYt1DGYb+8+YsEguVeuS84rKvQSI/GNmQ5MTidmV078X04OpuxsjPaMGX+C7Fx9MVf8/jcUf/iDdL9zkNfu/ju6DzbPaJrGIpHODOp6+11ev+c+Dm5/GNuiRWResCztv7WIscWCIU489XPa9+wd9X0nf/Vbju94ivhkrUIIMT0seXkAhH0dw+6PdPdgcmaTX/veWdGEK0ZmsFq5+F+/SdVX/5Gwr4OD33iEaCA49oEzRJrjZ0jg5CnefuB/E4uEKbrtFjKXLcWYOXrzrZgdNC3G6V/9jnBbe3zu1hH6Anlf2k3H6/sxzNKpNYQQs4M5EYR2dg67P+/qK8i6aAWZFyxPZbLENCq/+y5sZaWgQKCz+7wWgplOEoTOgLDPx8GvbSfS3c3yB75M8Qf+WgLQOSRz+TIyV1Th2/cXwt72EZ9uB06fwZKXK83xQohpZV1YROaKqhEX7Ij09GJ02DFlSVP8XFJ44/uJdHVz7i+vEQuP3T1sJkhzfIpFurppeegbhM6ew33PZym548MSgM5BCzfeRswfoO2l3cPu1zSN0NmzmPPzZrwfkRBibnMsqWDNfz6O6/JLh6ya1PbSbg49+l0ivX4MKVjcQ6SWMdOBvWoZ1kUL03K2HQlCUyzc5iXS2c2iD38A9z2flW+ec9SiTbeDotD20u5hl8qLdHYRCwSxFOSjmqUPlhBieikGA4rROGTC+rYXXqLtxT9iycudkxPUCzDa7dhKS9JyKkAJQlPMvsTNivpvsWTLF2UQ0hxmLSok59K1dL6+n+DpM0P2B8+cBcCyYIE8CRVCTLvWJ37Miaeejq8810fTNHyev2ArKU6uuy5EKkmf0BRTDAZyLrsE2yxcXktMzJLN99L5xptokciQfUaHnYL3ryd7zSqZjksIMe3OPfc83e8epPQTH0sOUvEfaSXc7iP3ynUYHfYZTqGYj6T2E2Ka5L3nKgredz2xaHRIk7y1qJCFm24j99JLZih1Qoj5JMNdTqSrWzdXqM+zDwDnZWtn5brjYvaTIFSIaWTMysR/9CiBEyd127VoFEVVUdN02gwhxNziWFIBgL+1NbnNt3cfqCq5V14xU8kS85wEoUJMo/ZXXuWdBx/izLONuu2v3XMvRx9/YsQpU4QQYirZK/uC0OP9X4iLbruFsk/dgbVwwUwlS8xzEoQKMY0K1l+HISODtt2vJJvko34/PQebiUWj0gQmhEgJu7scgOCpU8ltGSXFFN5yk0wTKGaMBKFCTCODzUbB+uvwHz5C1zsHAehpOQSahmNJpTTHCyFSIqN8MYs+uBH7kkq0aJTAyVOEOjowZWdhkAUzxAyRIFSIabboQ5sAONf43wB0v/0uAFkrqlCNMkGFEGL6GR12lj/wZZzVq4gGQ7z15Qc4sPUBGRUvZpQEoUJMs7xrrsaU66Ltjy8T6enh9O/+C9VmJedyGRkvhEgd1WZFMZnw7fXgP9pKzmWXYHLJfNVi5shjGCGmmWo0UvbJO+h66wAdf3md3sNHWHDD+7AVF8900oQQ80iozYvnY58mFgwCUHjrjdIfVMwoCUKFSIEl938R/5GjdB14m6qvPYhj6RJMWZkznSwhxDySUVaKyZVD8OQpnJfUkHvlOlmqU8woCUKFSAFFUbAWLyIaCmF3u7EUFsx0koQQ84yiqlT/x/fp+MtrFLzvepmaScw4CUKFSBHVaCRz6ZKZToYQYh7LWbOanDWrZzoZQgCzeGBSS0sLjY2N+Hw+3WshhBBCCJH+UvYkdPv27bjdbrxeL83NzWzduhWn0znqMVu2bAHiAabL5WLbtm3JYzweD5s2bUq+1+12s2vXrulKvhBCCCGEmEIpCUK3b98OwMaNG4H+AHK0oLGurk4XdNbV1bFmzRqam5uT76mvr8flcuF2u6murp6+CxBCCCGEEFMqJc3xDz30UDIABaiurmbPnj20tLQM+36fz0djY6Nu/5YtW4Y0udfW1rJx40YJQIUQQgghZplpfxLq8Xjw+Xy4Bk2I63K5aGhoYPPmzcMe5/V6aWlpSQaYieMHBqY+nw+Px4PX66WmpmbM5n2AM2fOcPbsWd22pqYmAAKBAH6/f9zXNhmBQGBazz/bSH7oSX70k7zQk/zQk/zoJ3mhJ/mhl+r8sNls437vtAehXq8XYEiA6HQ6aWtrG/YYp9NJe3u7blviCWhtbW1y244dO6irq8PtdnPnnXdSV1en2z+cxx57jAcffHCilyGEEEIIIabQtAehidHrE9032EMPPcTmzZtxu91AvH/pwCb+uro6Nm3axKFDh0Z9Inr33XfrBjRB/Enohg0bsFqtE4rgz0eqPme2kPzQk/zoJ3mhJ/mhJ/nRT/JCT/JDLx3zY8JBaENDAzt27BjzfVu3bqW6unrEgHAiAeiWLVuoqalh27ZtI76npqYGn8/Hnj17Rn0aWlBQQEGBTBQuhBBCCDGTJhyEDn4COZZEX06fzzckIK2oqBjz+IaGBnJzc4cEoDk5OezcuTMZcCbOPZHgVgghhBBCzIxpb45PPA0dOMgI4gOMxuq/2djYiNfr1Q1eamxspLa2FrfbnWyaT5wv8XkTFQwGgf4BStMp0UHYarVO+2fNBpIfepIf/SQv9CQ/9CQ/+kle6El+6M1EflRUVIzr81IyT+jWrVvZsWNHMkD0eDxUV1frXtfX11NfX588xuPxsHPnTjZt2pQclOTxeJKBayIQTdi2bRt33XWXbtt4tba2ArBhw4ZJXZ8QQgghhIjbv38/K1asGPN9iqZpWgrSw/bt25NN5oNXTGpoaGDLli3Jieh9Ph/l5eXDNq0PTG5iEvy2tjZyc3NHnO5pLD6fj+eff56SkhIsFsukzjFeiUFQTz/9NJWVldP6WbOB5Iee5Ec/yQs9yQ89yY9+khd6kh96M5EfafUkFBg1QBzcz3S4KZomes6JcDqd3HrrrVNyrvGqrKwc17eE+ULyQ0/yo5/khZ7kh57kRz/JCz3JD710zI+UrJgkhBBCCCHEQBKECiGEEEKIlJMgVAghhBBCpJwEoSmWn5/PAw88QH5+/kwnJS1IfuhJfvSTvNCT/NCT/OgneaEn+aGXzvmRstHxQgghhBBCJMiTUCGEEEIIkXIShAohhBBCiJSTIFQIIYQQQqScBKFCCCGEECLlJAgVaamxsXHYZVsTWlpadO9JvJ6L5tO1TgXJLwHz7z6Yb9c7Gqk/Zo+ULds53/h8Pp588knq6+vZu3fvuI7Zvn07brcbr9dLc3MzW7duxel0jnt/Opto2rdt2zZsoeB2u2lubsbj8bBp0ybd9l27dk1H0qfcRPNiPNc6n+4NgC1btgDxysPlcrFt27bkMbPl3pjMdUsZoTcX7oORSDnRT+qPfnMttpAgdBo0NjbS0tKCz+cb9dvYQNu3bwdg48aNQH+BkvjDGGt/OptM2ocrFHbu3EldXV3ydX19PS6XC7fbTXV19TSkfOpN9vc42rXOt3ujrq5OF2zU1dWxZs0ampubk+9J93tjMtctZYTeXLgPRiLlRD+pP/rNydhCE9Nm586dmtvtHtd7nU6n1tzcPOK2sfans8mkfefOnbrXzc3N2rZt23T7Z8O1DzbZvBht/3y6N9rb2zW3263t3bs3ua25uVkDtF27dmmaNjvujcn8zqSM6DdX7oORSDnRT+qPoeZSbCF9QtOAx+PB5/Phcrl0210uFw0NDWPuT2eTTXviW1lCfX09mzdv1m3z+Xx4PJ4x+/+ki/P5PY50rfPx3vB6vbS0tOjeD+i2pfO9MZnrljJiqNl+H4xEyol+Un+cn9lQbkgQmga8Xi/AkD4YTqeTtra2Mfens6lI+/e+9z2uv/76Idt37NiB0+mkpqaGO++8M+07lp9PXox0rfPt3nA6nbS3t+sqmURe1NbWJrel870xmeuWMoIh+2b7fTASKSf6Sf1xfmZDuSF9QtPAaN/Cxur7ke7f4KYi7cN1wN64caOuAqqrq2PTpk0cOnQobTvaTzYvRrvW+X5vADz00ENs3rwZt9sNpP+9MZnrljJibLPtPhiJlBP9pP44P7Oh3JAgdAwNDQ3s2LFjzPdt3bp10p2bR7rpEzfBWPtTaaL5cb5pb2hoSFYqo6mpqcHn87Fnzx7dk5DplOq8SBh4rfP53oD46Oiamhq2bds24ntm4t4YzWSuezaVERM1X++Dkcz1cmIi5nL9kQqzodyQIHQMg78xTYdEfwyfzzfkpqioqBhzfypNND/ON+319fXDBvc5OTns3LkzWWAkzp3KP55U5cVo15ooYOfjvdHQ0EBubu6QwCMd7o3RTOa6Z1MZMVHz9T4YyVwvJyZiLtcfqTAryo2UDH+apyY6gm3gSE9N0zQguW2s/ensfNLudDp1oxoTqqurdaP3EiNj033E42TyYqxrnY/3xq5du7T6+voh2zRtdtwbk7luKSOGmu33wUiknOgn9cdQcym2kIFJ0yzR8Xcgj8ejm68M4k2UA5syPR4P1dXVyW9xY+1PZ2Olfbj8SBjpm2ltba2umWXbtm3cdddd42p6mUmTyYuxrnW+3Rsej4edO3fidrtpbGyksbGR7du3J7/Vz4Z7YzLXLWXE3LsPRiLlRD+pP4Y3V2ILpS/qFVMoMe3Djh078Hg8bN68mdzc3OQUEQ0NDWzZskU3qTLEJ41NPBIfaVWD0fans9HSPlJ+QLxJoL6+fth+OolJdtva2nT5m+4mkxdjXet8uTd8Ph/l5eXDVi4Di7LZcG9M9j6QMmJu3QcjkXKin9QfcXMxtpAgVAghhBBCpJw0xwshhBBCiJSTIFQIIYQQQqScBKFCCCGEECLlJAgVQgghhBApJ0GoEEIIIYRIOQlChRBCCCFEykkQKoQQQgghUk6CUCGEEEIIkXIShAohhBBCiJSTIFQIIYQQQqScBKFCCCGEECLlJAgVQgghhBApJ0GoEEIIIYRIuf8PyVgKoe48sXQAAAAASUVORK5CYII=", - "text/plain": [ - "
" - ] - }, - "metadata": {}, - "output_type": "display_data" - } - ], - "source": [ - "map_latent_dist = opt_posterior.predict(xtest, train_data=D)\n", - "predictive_dist = opt_posterior.likelihood(map_latent_dist)\n", - "\n", - "predictive_mean = predictive_dist.mean()\n", - "predictive_std = predictive_dist.stddev()\n", - "\n", - "fig, ax = plt.subplots()\n", - "ax.scatter(x, y, label=\"Observations\", color=cols[0])\n", - "ax.plot(xtest, predictive_mean, label=\"Predictive mean\", color=cols[1])\n", - "ax.fill_between(\n", - " xtest.squeeze(),\n", - " predictive_mean - predictive_std,\n", - " predictive_mean + predictive_std,\n", - " alpha=0.2,\n", - " color=cols[1],\n", - " label=\"One sigma\",\n", - ")\n", - "ax.plot(\n", - " xtest,\n", - " predictive_mean - predictive_std,\n", - " color=cols[1],\n", - " linestyle=\"--\",\n", - " linewidth=1,\n", - ")\n", - "ax.plot(\n", - " xtest,\n", - " predictive_mean + predictive_std,\n", - " color=cols[1],\n", - " linestyle=\"--\",\n", - " linewidth=1,\n", - ")\n", - "\n", - "ax.legend()" - ] - }, - { - "cell_type": "markdown", - "id": "e78427fe", - "metadata": {}, - "source": [ - "Here we projected the map estimates $\\hat{\\boldsymbol{f}}$ for the function values\n", - "$\\boldsymbol{f}$ at the data points $\\boldsymbol{x}$ to get predictions over the\n", - "whole domain,\n", - "\n", - "\\begin{align}\n", - "p(f(\\cdot)| \\mathcal{D}) \\approx q_{map}(f(\\cdot)) := \\int p(f(\\cdot)| \\boldsymbol{f}) \\delta(\\boldsymbol{f} - \\hat{\\boldsymbol{f}}) d \\boldsymbol{f} = \\mathcal{N}(\\mathbf{K}_{\\boldsymbol{(\\cdot)x}} \\mathbf{K}_{\\boldsymbol{xx}}^{-1} \\hat{\\boldsymbol{f}}, \\mathbf{K}_{\\boldsymbol{(\\cdot, \\cdot)}} - \\mathbf{K}_{\\boldsymbol{(\\cdot)\\boldsymbol{x}}} \\mathbf{K}_{\\boldsymbol{xx}}^{-1} \\mathbf{K}_{\\boldsymbol{\\boldsymbol{x}(\\cdot)}}).\n", - "\\end{align}" - ] - }, - { - "cell_type": "markdown", - "id": "7cf8263e", - "metadata": {}, - "source": [ - "However, as a point estimate, MAP estimation is severely limited for uncertainty\n", - "quantification, providing only a single piece of information about the posterior." - ] - }, - { - "cell_type": "markdown", - "id": "e0703339", - "metadata": {}, - "source": [ - "## Laplace approximation\n", - "The Laplace approximation improves uncertainty quantification by incorporating\n", - "curvature induced by the marginal log-likelihood's Hessian to construct an\n", - "approximate Gaussian distribution centered on the MAP estimate. Writing\n", - "$\\tilde{p}(\\boldsymbol{f}|\\mathcal{D}) = p(\\boldsymbol{y}|\\boldsymbol{f}) p(\\boldsymbol{f})$\n", - "as the unormalised posterior for function values $\\boldsymbol{f}$ at the datapoints\n", - "$\\boldsymbol{x}$, we can expand the log of this about the posterior mode\n", - "$\\hat{\\boldsymbol{f}}$ via a Taylor expansion. This gives:\n", - "\n", - "\\begin{align}\n", - "\\log\\tilde{p}(\\boldsymbol{f}|\\mathcal{D}) = \\log\\tilde{p}(\\hat{\\boldsymbol{f}}|\\mathcal{D}) + \\left[\\nabla \\log\\tilde{p}({\\boldsymbol{f}}|\\mathcal{D})|_{\\hat{\\boldsymbol{f}}}\\right]^{T} (\\boldsymbol{f}-\\hat{\\boldsymbol{f}}) + \\frac{1}{2} (\\boldsymbol{f}-\\hat{\\boldsymbol{f}})^{T} \\left[\\nabla^2 \\tilde{p}(\\boldsymbol{y}|\\boldsymbol{f})|_{\\hat{\\boldsymbol{f}}} \\right] (\\boldsymbol{f}-\\hat{\\boldsymbol{f}}) + \\mathcal{O}(\\lVert \\boldsymbol{f} - \\hat{\\boldsymbol{f}} \\rVert^3).\n", - "\\end{align}\n", - "\n", - "Since $\\nabla \\log\\tilde{p}({\\boldsymbol{f}}|\\mathcal{D})$ is zero at the mode,\n", - "this suggests the following approximation\n", - "\\begin{align}\n", - "\\tilde{p}(\\boldsymbol{f}|\\mathcal{D}) \\approx \\log\\tilde{p}(\\hat{\\boldsymbol{f}}|\\mathcal{D}) \\exp\\left\\{ \\frac{1}{2} (\\boldsymbol{f}-\\hat{\\boldsymbol{f}})^{T} \\left[-\\nabla^2 \\tilde{p}(\\boldsymbol{y}|\\boldsymbol{f})|_{\\hat{\\boldsymbol{f}}} \\right] (\\boldsymbol{f}-\\hat{\\boldsymbol{f}}) \\right\\}\n", - "\\end{align},\n", - "\n", - "that we identify as a Gaussian distribution,\n", - "$p(\\boldsymbol{f}| \\mathcal{D}) \\approx q(\\boldsymbol{f}) := \\mathcal{N}(\\hat{\\boldsymbol{f}}, [-\\nabla^2 \\tilde{p}(\\boldsymbol{y}|\\boldsymbol{f})|_{\\hat{\\boldsymbol{f}}} ]^{-1} )$.\n", - "Since the negative Hessian is positive definite, we can use the Cholesky\n", - "decomposition to obtain the covariance matrix of the Laplace approximation at the\n", - "datapoints below." - ] - }, - { - "cell_type": "code", - "execution_count": 7, - "id": "4f96ede8", - "metadata": { - "lines_to_next_cell": 2 - }, - "outputs": [], - "source": [ - "import cola\n", - "from gpjax.lower_cholesky import lower_cholesky\n", - "\n", - "gram, cross_covariance = (kernel.gram, kernel.cross_covariance)\n", - "jitter = 1e-6\n", - "\n", - "# Compute (latent) function value map estimates at training points:\n", - "Kxx = opt_posterior.prior.kernel.gram(x)\n", - "Kxx += identity_matrix(D.n) * jitter\n", - "Kxx = cola.PSD(Kxx)\n", - "Lx = lower_cholesky(Kxx)\n", - "f_hat = Lx @ opt_posterior.latent\n", - "\n", - "# Negative Hessian, H = -∇²p_tilde(y|f):\n", - "H = jax.jacfwd(jax.jacrev(negative_lpd))(opt_posterior, D).latent.latent[:, 0, :, 0]\n", - "\n", - "L = jnp.linalg.cholesky(H + identity_matrix(D.n) * jitter)\n", - "\n", - "# H⁻¹ = H⁻¹ I = (LLᵀ)⁻¹ I = L⁻ᵀL⁻¹ I\n", - "L_inv = jsp.linalg.solve_triangular(L, identity_matrix(D.n), lower=True)\n", - "H_inv = jsp.linalg.solve_triangular(L.T, L_inv, lower=False)\n", - "LH = jnp.linalg.cholesky(H_inv)\n", - "laplace_approximation = tfd.MultivariateNormalTriL(f_hat.squeeze(), LH)" - ] - }, - { - "cell_type": "markdown", - "id": "1b21f9c7", - "metadata": { - "lines_to_next_cell": 2 - }, - "source": [ - "For novel inputs, we must project the above approximating distribution through the\n", - "Gaussian conditional distribution $p(f(\\cdot)| \\boldsymbol{f})$,\n", - "\n", - "\\begin{align}\n", - "p(f(\\cdot)| \\mathcal{D}) \\approx q_{Laplace}(f(\\cdot)) := \\int p(f(\\cdot)| \\boldsymbol{f}) q(\\boldsymbol{f}) d \\boldsymbol{f} = \\mathcal{N}(\\mathbf{K}_{\\boldsymbol{(\\cdot)x}} \\mathbf{K}_{\\boldsymbol{xx}}^{-1} \\hat{\\boldsymbol{f}}, \\mathbf{K}_{\\boldsymbol{(\\cdot, \\cdot)}} - \\mathbf{K}_{\\boldsymbol{(\\cdot)\\boldsymbol{x}}} \\mathbf{K}_{\\boldsymbol{xx}}^{-1} (\\mathbf{K}_{\\boldsymbol{xx}} - [-\\nabla^2 \\tilde{p}(\\boldsymbol{y}|\\boldsymbol{f})|_{\\hat{\\boldsymbol{f}}} ]^{-1}) \\mathbf{K}_{\\boldsymbol{xx}}^{-1} \\mathbf{K}_{\\boldsymbol{\\boldsymbol{x}(\\cdot)}}).\n", - "\\end{align}\n", - "\n", - "This is the same approximate distribution $q_{map}(f(\\cdot))$, but we have perturbed\n", - "the covariance by a curvature term of\n", - "$\\mathbf{K}_{\\boldsymbol{(\\cdot)\\boldsymbol{x}}} \\mathbf{K}_{\\boldsymbol{xx}}^{-1} [-\\nabla^2 \\tilde{p}(\\boldsymbol{y}|\\boldsymbol{f})|_{\\hat{\\boldsymbol{f}}} ]^{-1} \\mathbf{K}_{\\boldsymbol{xx}}^{-1} \\mathbf{K}_{\\boldsymbol{\\boldsymbol{x}(\\cdot)}}$.\n", - "We take the latent distribution computed in the previous section and add this term\n", - "to the covariance to construct $q_{Laplace}(f(\\cdot))$." - ] - }, - { - "cell_type": "code", - "execution_count": 8, - "id": "4021889b", - "metadata": {}, - "outputs": [], - "source": [ - "def construct_laplace(test_inputs: Float[Array, \"N D\"]) -> tfd.MultivariateNormalTriL:\n", - " map_latent_dist = opt_posterior.predict(xtest, train_data=D)\n", - "\n", - " Kxt = opt_posterior.prior.kernel.cross_covariance(x, test_inputs)\n", - " Kxx = opt_posterior.prior.kernel.gram(x)\n", - " Kxx += identity_matrix(D.n) * jitter\n", - " Kxx = cola.PSD(Kxx)\n", - "\n", - " # Kxx⁻¹ Kxt\n", - " Kxx_inv_Kxt = cola.solve(Kxx, Kxt)\n", - "\n", - " # Ktx Kxx⁻¹[ H⁻¹ ] Kxx⁻¹ Kxt\n", - " laplace_cov_term = jnp.matmul(jnp.matmul(Kxx_inv_Kxt.T, H_inv), Kxx_inv_Kxt)\n", - "\n", - " mean = map_latent_dist.mean()\n", - " covariance = map_latent_dist.covariance() + laplace_cov_term\n", - " L = jnp.linalg.cholesky(covariance)\n", - " return tfd.MultivariateNormalTriL(jnp.atleast_1d(mean.squeeze()), L)" - ] - }, - { - "cell_type": "markdown", - "id": "6458ec70", - "metadata": { - "lines_to_next_cell": 0 - }, - "source": [ - "From this we can construct the predictive distribution at the test points." - ] - }, - { - "cell_type": "code", - "execution_count": 9, - "id": "73ba0f59", - "metadata": {}, - "outputs": [ - { - "data": { - "text/plain": [ - "" - ] - }, - "execution_count": 9, - "metadata": {}, - "output_type": "execute_result" - }, - { - "data": { - "image/png": "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", - "text/plain": [ - "
" - ] - }, - "metadata": {}, - "output_type": "display_data" - } - ], - "source": [ - "laplace_latent_dist = construct_laplace(xtest)\n", - "predictive_dist = opt_posterior.likelihood(laplace_latent_dist)\n", - "\n", - "predictive_mean = predictive_dist.mean()\n", - "predictive_std = predictive_dist.stddev()\n", - "\n", - "fig, ax = plt.subplots()\n", - "ax.scatter(x, y, label=\"Observations\", color=cols[0])\n", - "ax.plot(xtest, predictive_mean, label=\"Predictive mean\", color=cols[1])\n", - "ax.fill_between(\n", - " xtest.squeeze(),\n", - " predictive_mean - predictive_std,\n", - " predictive_mean + predictive_std,\n", - " alpha=0.2,\n", - " color=cols[1],\n", - " label=\"One sigma\",\n", - ")\n", - "ax.plot(\n", - " xtest,\n", - " predictive_mean - predictive_std,\n", - " color=cols[1],\n", - " linestyle=\"--\",\n", - " linewidth=1,\n", - ")\n", - "ax.plot(\n", - " xtest,\n", - " predictive_mean + predictive_std,\n", - " color=cols[1],\n", - " linestyle=\"--\",\n", - " linewidth=1,\n", - ")\n", - "ax.legend()" - ] - }, - { - "cell_type": "markdown", - "id": "6d10a99b", - "metadata": {}, - "source": [ - "However, the Laplace approximation is still limited by considering information about\n", - "the posterior at a single location. On the other hand, through approximate sampling,\n", - "MCMC methods allow us to learn all information about the posterior distribution." - ] - }, - { - "cell_type": "markdown", - "id": "b22b9996", - "metadata": {}, - "source": [ - "## MCMC inference\n", - "\n", - "An MCMC sampler works by starting at an initial position and\n", - "drawing a sample from a cheap-to-simulate distribution known as the _proposal_. The\n", - "next step is to determine whether this sample could be considered a draw from the\n", - "posterior. We accomplish this using an _acceptance probability_ determined via the\n", - "sampler's _transition kernel_ which depends on the current position and the\n", - "unnormalised target posterior distribution. If the new sample is more _likely_, we\n", - "accept it; otherwise, we reject it and stay in our current position. Repeating these\n", - "steps results in a Markov chain (a random sequence that depends only on the last\n", - "state) whose stationary distribution (the long-run empirical distribution of the\n", - "states visited) is the posterior. For a gentle introduction, see the first chapter\n", - "of [A Handbook of Markov Chain Monte Carlo](https://www.mcmchandbook.net/HandbookChapter1.pdf).\n", - "\n", - "### MCMC through BlackJax\n", - "\n", - "Rather than implementing a suite of MCMC samplers, GPJax relies on MCMC-specific\n", - "libraries for sampling functionality. We focus on\n", - "[BlackJax](https://github.com/blackjax-devs/blackjax/) in this notebook, which we\n", - "recommend adopting for general applications.\n", - "\n", - "We'll use the No U-Turn Sampler (NUTS) implementation given in BlackJax for sampling.\n", - "For the interested reader, NUTS is a Hamiltonian Monte Carlo sampling scheme where\n", - "the number of leapfrog integration steps is computed at each step of the change\n", - "according to the NUTS algorithm. In general, samplers constructed under this\n", - "framework are very efficient.\n", - "\n", - "We begin by generating _sensible_ initial positions for our sampler before defining\n", - "an inference loop and sampling 500 values from our Markov chain. In practice,\n", - "drawing more samples will be necessary." - ] - }, - { - "cell_type": "code", - "execution_count": 10, - "id": "1fd8042b", - "metadata": {}, - "outputs": [], - "source": [ - "num_adapt = 500\n", - "num_samples = 500\n", - "\n", - "lpd = jax.jit(gpx.LogPosteriorDensity(negative=False))\n", - "unconstrained_lpd = jax.jit(lambda tree: lpd(tree.constrain(), D))\n", - "\n", - "adapt = blackjax.window_adaptation(\n", - " blackjax.nuts, unconstrained_lpd, num_adapt, target_acceptance_rate=0.65\n", - ")\n", - "\n", - "# Initialise the chain\n", - "start = time()\n", - "last_state, kernel, _ = adapt.run(key, posterior.unconstrain())\n", - "print(f\"Adaption time taken: {time() - start: .1f} seconds\")\n", - "\n", - "\n", - "def inference_loop(rng_key, kernel, initial_state, num_samples):\n", - " def one_step(state, rng_key):\n", - " state, info = kernel(rng_key, state)\n", - " return state, (state, info)\n", - "\n", - " keys = jax.random.split(rng_key, num_samples)\n", - " _, (states, infos) = jax.lax.scan(one_step, initial_state, keys)\n", - "\n", - " return states, infos\n", - "\n", - "\n", - "# Sample from the posterior distribution\n", - "start = time()\n", - "states, infos = inference_loop(key, kernel, last_state, num_samples)\n", - "print(f\"Sampling time taken: {time() - start: .1f} seconds\")" - ] - }, - { - "cell_type": "markdown", - "id": "d0cf235c", - "metadata": {}, - "source": [ - "### Sampler efficiency\n", - "\n", - "BlackJax gives us easy access to our sampler's efficiency through metrics such as the\n", - "sampler's _acceptance probability_ (the number of times that our chain accepted a\n", - "proposed sample, divided by the total number of steps run by the chain). For NUTS and\n", - "Hamiltonian Monte Carlo sampling, we typically seek an acceptance rate of 60-70% to\n", - "strike the right balance between having a chain which is _stuck_ and rarely moves\n", - "versus a chain that is too jumpy with frequent small steps." - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "8c9a7f7f", - "metadata": {}, - "outputs": [], - "source": [ - "acceptance_rate = jnp.mean(infos.acceptance_probability)\n", - "print(f\"Acceptance rate: {acceptance_rate:.2f}\")" - ] - }, - { - "cell_type": "markdown", - "id": "6eae2e0a", - "metadata": {}, - "source": [ - "Our acceptance rate is slightly too large, prompting an examination of the chain's\n", - "trace plots. A well-mixing chain will have very few (if any) flat spots in its trace\n", - "plot whilst also not having too many steps in the same direction. In addition to\n", - "the model's hyperparameters, there will be 500 samples for each of the 100 latent\n", - "function values in the `states.position` dictionary. We depict the chains that\n", - "correspond to the model hyperparameters and the first value of the latent function\n", - "for brevity." - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "d9328818", - "metadata": {}, - "outputs": [], - "source": [ - "fig, (ax0, ax1, ax2) = plt.subplots(ncols=3, figsize=(10, 3))\n", - "ax0.plot(states.position.prior.kernel.lengthscale)\n", - "ax1.plot(states.position.prior.kernel.variance)\n", - "ax2.plot(states.position.latent[:, 1, :])\n", - "ax0.set_title(\"Kernel Lengthscale\")\n", - "ax1.set_title(\"Kernel Variance\")\n", - "ax2.set_title(\"Latent Function (index = 1)\")" - ] - }, - { - "cell_type": "markdown", - "id": "f875f762", - "metadata": {}, - "source": [ - "## Prediction\n", - "\n", - "Having obtained samples from the posterior, we draw ten instances from our model's\n", - "predictive distribution per MCMC sample. Using these draws, we will be able to\n", - "compute credible values and expected values under our posterior distribution.\n", - "\n", - "An ideal Markov chain would have samples completely uncorrelated with their\n", - "neighbours after a single lag. However, in practice, correlations often exist\n", - "within our chain's sample set. A commonly used technique to try and reduce this\n", - "correlation is _thinning_ whereby we select every $n$th sample where $n$ is the\n", - "minimum lag length at which we believe the samples are uncorrelated. Although further\n", - "analysis of the chain's autocorrelation is required to find appropriate thinning\n", - "factors, we employ a thin factor of 10 for demonstration purposes." - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "bc1730a3", - "metadata": {}, - "outputs": [], - "source": [ - "thin_factor = 20\n", - "posterior_samples = []\n", - "\n", - "for i in trange(0, num_samples, thin_factor, desc=\"Drawing posterior samples\"):\n", - " sample = jtu.tree_map(lambda samples, i=i: samples[i], states.position)\n", - " sample = sample.constrain()\n", - " latent_dist = sample.predict(xtest, train_data=D)\n", - " predictive_dist = sample.likelihood(latent_dist)\n", - " posterior_samples.append(predictive_dist.sample(seed=key, sample_shape=(10,)))\n", - "\n", - "posterior_samples = jnp.vstack(posterior_samples)\n", - "lower_ci, upper_ci = jnp.percentile(posterior_samples, jnp.array([2.5, 97.5]), axis=0)\n", - "expected_val = jnp.mean(posterior_samples, axis=0)" - ] - }, - { - "cell_type": "markdown", - "id": "43b956b0", - "metadata": {}, - "source": [ - "\n", - "Finally, we end this tutorial by plotting the predictions obtained from our model\n", - "against the observed data." - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "d8a65948", - "metadata": {}, - "outputs": [], - "source": [ - "fig, ax = plt.subplots()\n", - "ax.scatter(x, y, color=cols[0], label=\"Observations\", zorder=2, alpha=0.7)\n", - "ax.plot(xtest, expected_val, color=cols[1], label=\"Predicted mean\", zorder=1)\n", - "ax.fill_between(\n", - " xtest.flatten(),\n", - " lower_ci.flatten(),\n", - " upper_ci.flatten(),\n", - " alpha=0.2,\n", - " color=cols[1],\n", - " label=\"95\\\\% CI\",\n", - ")\n", - "ax.plot(\n", - " xtest,\n", - " lower_ci.flatten(),\n", - " color=cols[1],\n", - " linestyle=\"--\",\n", - " linewidth=1,\n", - ")\n", - "ax.plot(\n", - " xtest,\n", - " upper_ci.flatten(),\n", - " color=cols[1],\n", - " linestyle=\"--\",\n", - " linewidth=1,\n", - ")\n", - "ax.legend()" - ] - }, - { - "cell_type": "markdown", - "id": "b9c17a58", - "metadata": {}, - "source": [ - "## System configuration" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "id": "a84586dc", - "metadata": {}, - "outputs": [], - "source": [ - "%load_ext watermark\n", - "%watermark -n -u -v -iv -w -a \"Thomas Pinder & Daniel Dodd\"" - ] - } - ], - "metadata": { - "jupytext": { - "cell_metadata_filter": "-all", - "custom_cell_magics": "kql", - "encoding": "# -*- coding: utf-8 -*-" - }, - "kernelspec": { - "display_name": "gpjax", - "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": "3.10.9" - } - }, - "nbformat": 4, - "nbformat_minor": 5 -}