XiaoNet.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import time\n",
    "import math\n",
    "from datetime import timedelta"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from xiaonet import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAK8AAADDCAYAAAD0pEFUAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAAPYQAAD2EBqD+naQAAIABJREFUeJztvXts21mW3/m5fL9JkaLEhx7Uy7JctvzoKldVT08nmSDb\nE0wn2J38kUGjF0kG2ezODgIkm50AWQQNDHYRJJldYBAECRaLAQZBZzIDLDbAAMlmGukgPT2DzpbL\nLtuynpZkvSiJL5Him+Ljt3/I9xZlq8qSLYmi9fsAhCyZ/P1+PPzy/s4959xzhaZp6Oh0I4ZOX4CO\nztuii1ena9HFq9O16OLV6Vp08ep0Lbp4dboWXbw6XYvpJE8SQgSA7wBrQPU8L6gLsAEx4I80Tcuc\nxQF1+x7h5PbVNO2ND+B7gKY/jjy+dxLb6fY9P/ueaOTlcETghz/8IVNTUyd8yfvJ/Pw83//+9+Gl\nTc6INdDtC6ez70nFWwWYmpri3r17b39l7xdneXvX7fs6b7SvPmHT6Vp08ep0Lbp4dboWXbw6XYsu\nXp2uRRevTteii1enazlpnFfnPUXTNJrNJs1mk2q1SqVSoVwuUywWKRaLMgN4BJPJhMlkwmaz4Xa7\ncblc2Gw2rFYrRqPxwq5dF+8VR9M0Dg4OqFarJJNJ4vE4W1tbvHjxghcvXtBoNBBCqOcCOJ1OnE4n\n4XCYsbExRkZG6Ovro6+vTxevzvkhBdhqtWg2mzQaDfb398nn86yvr7OwsMDc3ByPHz/m8ePHHBwc\nvHYMn8+Hz+djfHycXC5Hq9XCYDDg8/kwm80ASvDniS7eK4SmaTQaDSXYRCJBIpFgZ2eH3d1ddnZ2\n2N7eJh6Pk0gkaLVaAK+NvLVajUKhwM7ODrOzs9TrdYxGI/39/ZjNZkwm04WMwLp4rxjNZpNarUY6\nnWZ+fp7Z2VlWV1dZWVlhb2+PYrFIqVTi4OCAZrN5ZAQVQig3o9Fo0Gw2qVQq7O3t0dfXxwcffIDL\n5UIIcfXEq2kapVKJYrFIo9FQfzcYDBiNRprNJvV6nVarhdVqPfKQtyud16nVauzv77O/v08qlSKZ\nTLKxscHKygqrq6vE43G2t7cpFotKtEajEZvNhtFoxGAwYDAcBqY0TaPVaim3Q7oN8/Pz9Pf3c/36\ndWKxGH19fef+vi6VeAGy2SxbW1sUi0WEEAghMJvNWCwWarWaEnZPTw8+nw+/34/f79fF+zVUKhU2\nNjZYXl5mfn6ehYUFdnZ2lK8rBww5MAghsFqtKoJgsVgwmUzKbajX6xwcHHBwcEC9XiebzfL06VMq\nlQq5XA6bzdYd4m21WtTrdRqNhvpGSr+qffQ8CZqmsbKywvLyMvv7+8DhqCtH10qlwv7+PvV6nb6+\nPoLBIH19fRSLRTwejxohbDYbNpsNk8mkvgBXmWq1Sjwe5+nTpzx69IhHjx6RSqWAL10B+W+j0YjV\naiUQCNDb24vD4cBut2M2m9XzyuUy5XKZXC5HKpVif3+f1dVVMpkMbreb6enpC3lf7yxe6T9lMhlK\npRLlcplsNksymSSbzZ7qWJqmkUwmSSaTlMtlJTqz2YzZbObg4IBKpUKz2cTtduN2u/F4PHg8Hlwu\nFw6HA4fDwcTEBBMTE2pENpku3Q3mQjk4OCCdTvPixQsymYyKIEj7SgFbLBbsdjvBYJDbt29z584d\n3G63Ei8cfkbZbJZsNsvKygqPHz8ml8tRq9XUKH7aQettORPxJpNJ9c3LZDJsbGywtLTE2traqY8n\nA+XSAEIINaLKUV3TNCVo+XA6ncqF+IVf+AV8Ph8ulwuDwaCL9+CATCbD2toamUyGer3+2t1Iumcu\nl4twOMxHH33EL/3SL9HT04PNZjvilsnohM/nY2dnh4WFBarVKgcHB8r9uAje+lPd29sjnU6zubnJ\n/Pw8i4uL5PN5CoUCqVSKeDyubk2nQbobMkwDqFt/+/qlWq2mJnJGo1HNknO5HMFgEIfDwf7+PkND\nQ4TDYYxGo3IjrgrSJ93f3yebzZLJZKjValgsFmU3+TAYDIRCIWKxGLFYDKfTydbWFs1mk76+Pux2\nu8qsBYNBzGYzm5ubOJ1O5U40m02KxSLb29usra3h9Xrxer1qsnfWvLV4U6kUMzMzPHv2jGfPnql4\n38HBAbVajXK5TK1WO/VxW61W+8JE4PBW9WqsUaY1231seeuamZmhWq2Sy+XQNA2fz6dSl1dNvIVC\nQd3mc7mcmoyZTCY1l7BYLFgsFiYmJrh79y6Dg4PkcjkeP36s5h5ms1kJ2OFwYDKZCAQC2Gw24MvP\nqFAosLGxwfPnz4nFYurudx6808j7/Plznj59qkbeV8Mqb4oAyG+yvPVbLJYTnVsKtl6vU6vV1AhT\nrVYpl8tsbGxQrR4ugZLf/t7eXgKBAFarVbki7zsyrGUwGPB4PITDYTWZtdvt6t9SxLFYjLGxMbxe\nL1tbW8zMzNBqtQgGg0ciOl83CZZx5HbX77x4a/EWi0Xi8TgbGxvk83k0TcNqteJ0OrFYLEdE/FW4\nXC48Hg89PT0EAgECgcCJzl2v18nlcuRyOZLJJIlEglwuR7VapVarUSqVSCaTPH/+HKvVSqlU4tat\nW9y6dUulMK+CeC0WC263m8HBQT799FM8Ho8SbPtDulSSra0tVldXmZ+fx+12c/36dWq1Gs1mE4D9\n/X0ymQzb29uUSiXgy8mfw+EgFAoxNDR0ri4DvIN4C4WCKuIoFosA2Gw2vF6vuq28KcsSDAYJh8NE\no1FGRkYYGRk50bmr1Sqbm5vE43GWlpbQNI16va6+9aVSiVKppNyXdDqN2WxmcHAQh8NxZSZx0h2w\nWCy4XC6uX79+RLR2ux2r1aqev7GxwdzcHPF4nJWVFebn5wmFQuRyOZVVA8jn82xtbbGzs6Pi8RKn\n06nEa7PZztVNe+tPcGBggJ/7uZ/D7/eTz+fJ5/P09/fT39+Py+VSLsHX4XK58Hq9+Hw+dVs/CfV6\nHZfLRSgUIhQKMTIywtraGouLizx//pxarabcCZn0WFtbY3l5GU3TCIVCRz609xUpHKPRiMPhUBGF\n9vqDdnEZjUYVLpOfjQyVtVotdnd32d7eZmFhgYWFBWZnZ9WkvL1MUroi5z1Bfifxfutb32J8fFxN\nBqLRKJFIBI/Ho4z0dRiNxiP+7ml83lAoRL1eZ2xsjL29PZaWljAajaRSKQqFgprA7e3tUa/XefHi\nBdFoVNWg+v3+t33rXYcUr9VqxWAwKJ//VWEZDAYlXrfbjc/nw+1243A4aLVaKkT29OlTnjx5wvr6\nOul0Wp1DZuXkiH7ertlbi9fr9TIyMkJvb68Kkcms10lH3rPA7XYTCAQwGo2qKmpzc1OlkmXEIx6P\nMz8/j8fjIRqNnvt1XSbk/MNkMqlJrsyKNptN9f+VSgWj0YjL5VIZtWw2y+rqKsVika2tLba2tlha\nWmJlZUXFjK1WK+FwWN0FvV7vhXz2b30G6UfJSUGtVsPpdCp/96ImRDJBEQwGuXbtGqVSCaPRSCaT\nUb54o9Fgd3cXg8FAb28vt27dupBru4yUSiWy2SyFQoFyuUy1WlV3P5kedrvdaJrG/v4+CwsL1Go1\nHA4He3t77O3tqdBbo9FQo/rU1BTf+MY3uHv37oXUNcA7iLc9s9VJ5HX4/X5GRkaUbzY7O6ue02q1\nSKfTFItFRkdHlaivIjISk0wm2dvbI5/PKz/V4/Hg8/mAL2PEuVyOeDyuRuZKpaJCZTLE1tPTw+Tk\nJN/+9rcZHR2lt7f3Qt7L+z/l1jlCLpdT5ZBbW1vs7u6qCZz0V2u1GvPz86rKrFarIYRQ0QbpEvp8\nPiKRCMPDw4yMjNDf34/X6z3x3OVd0cV7xdjf32d9fZ3Z2VkWFxd58eKFmrjJdHur1aJUKqkaBVkm\nKVP2cnLW09NDLBbjxo0bjI6OEgqF8Hg8F7aOrevFK1e5plIptre3SaVSrxWzw5fZJlk8UigUVITj\nKqWM6/U6lUpFuQR7e3tHUvGvpuElBoNBRSz6+/sJhUIMDw9z7do1rl27xsDAAE6n80LrqrtevHIk\n2draYnt7m52dHRUek0jhymUruVyOTCaD1+tVE5WrjCx6Om7JT3tZqtfrxe/3c/PmTW7evMno6CgD\nAwNEIhFly4vkUom3vWpMrmyt1+vq0V5pJlldXWVhYYH19XVVktlesyqR9RAyM7i2tsbQ0BBut/tK\npIol0reVReY2m03ZXBZFyZ/tYrZYLPT29hKLxbh16xb3799ndHQUv9+vJnkX/l46ctavQQo2n8+T\nTqdJpVIqOH5clZpc/dpeDC9L/9qR4o3H4/zsZz+jXC7z7W9/m3A4fCVSxRKfz8fw8LAqHG+P+Var\nVVUfIhuRwOEobLfbicVifPTRR1y/fp1oNIrX6+1opvLSfWqNRoNKpUImk2FlZYXnz5+rksvjQlxy\n/VW1Wj0yar86SsslStvb26qgJxKJ8M1vfvOi3tqlwOv1MjQ0RLlcVgKWoi0UCuzv79NsNpUd4fCO\n6HA4iMVi3L9/X2VS5UrhTtFR8crWQu3f+GQyqSZfGxsbbG5usr6+TjweV2WO7cj6YWnoN2GxWFQt\nhcz3XyVcLhf9/f00m03MZjPhcFi5Zfl8Xk3iEokEyWSSQqGgBL67u8v8/DwGgwG/34/dbr/QDjmv\n0lHxlstlUqkUmUxGlTiurKywsrLC9vY2mUyGvb09VSV2XH2oHFFPisvlYnBwkLGxMQKBwJXydwGV\nAXW5XEQiEcrlsrKhjEDs7u7y7Nkz5ubmVPVYuVxmZWUFOBwAIpEIbrdbrcroBBcmXjnbr9frSow7\nOztsbW2RSCRU2nF5eZmVlRVSqRSlUolKpaJy77KY5NUSPLvdrgLjQgjl+8oeBO3ittlsBAIBwuFw\n10/W2ida0i5f937as2Jut/u1NK50JRKJhKoMs9lsNBoNstks6XSaarXK0NAQ165dw+PxXGhS4lUu\nTLwyvpjNZnn+/DnLy8tsbm6ytbVFJpOhUqlQrVZV/rxcLquFghaLRS0CfLVOOBqNEovF8Pv9SuDL\ny8ssLy+TTqdfW44km2nIFbHd7jbICa7BYHjnInu5ABNgenqaUCiE1+tF0zSWl5epVCqkUim2trZY\nWVlRtdFut/us3s6puDDxNhoNSqUSiUSCx48f88d//Mesra2xtbWl1kkdhxSv0+lUo0F7PHFycpL7\n9+8zPDysCoJ++tOfqi+DLI2USPE6HI6ub1Qie49Vq1X13t8lciLTvk6nk/7+fuAwOZHJZMhms2xu\nbpJMJpV4ZdOXTnGu4m02myruurOzo+Krs7OzrK+vk81m1cRBrmINBAL4/X6cTqdaJNjT00NPTw8u\nl+u16v+BgQGGh4eV/yqEIBAIKHHKkUiOym63m/7+fqLRqGpU0m3IFSKFQoHNzU02NzcRQuByufD5\nfAwMDBCNRk8t5OPuQoFAgMnJSfL5PJVKRZWbFgoF1R6qU5y7eBOJBEtLSywtLbGwsMDq6qrqTihD\nMnI0NZvNyp8KBoPY7XacTqfq/Sqbi8gVqwB2u135vNL4x4lXrtNyu92EQiEVp+xG8coVIjs7O3zx\nxRd88cUXNJtNvF4v0WiUjz/+mP7+/jOJX/v9fiYnJymVSupLciXE22q1yOVybG5usry8zOLiIhsb\nGzQaDVWVJJeayFrg8fFxxsfH6e3tVbd3v99PIBBQXXGOC4zLtlP1ev1IK89XWxnJ4wWDQZxOZ9f4\nvO0N7vL5PGtraywtLbG4uMja2hoOhwOXy3Umy29e7Y8hJ9hSqM1mU/UqO02k56w5d/GWSiVSqRSp\nVIpcLkelUlHNiYeGhojFYkQiEbV62Ofz0dPTc6TJhUxjfl1YptFoqIKbbDZLPp8/MumTFVNypi1n\nyd0iXvhSNMlkkmfPnvHgwQOVVAiFQty7d4/bt28TCoXeOXwlo0M7Ozs8evSIzz//nO3t7WPb/HeK\ncxWv7OUqQ1dy4iQLxz/44ANu3brF2NiYWkbytjSbTUqlkooZFwqFI70DXl1c2KkZ8rtQr9cpl8sk\nEgnm5+d58OABTqcTl8tFMBjkgw8+4Bvf+MZbH789Qynbbm1sbPDkyROePHlCLpcDOBKWe28zbCaT\niaGhIT7++GN8Ph/BYJBcLseNGze4fv06AwMDhMNhle16F2TF2P7+vsoKycybnKgFg8EjXV66jXw+\nz+7urpr15/N5nE6n2tTkXaMnsnFLLpdT8fZHjx6xvr5OoVBQboPVasXj8eB2uzsasbkQ8Xq9XtV1\nJZfLcffuXe7evasalJxFG/h28RaLRTXSt1otTCaT6hgjfeluRPZLaBevTLbIRa/vQr1ep1gssrOz\nw4MHD/jJT37C5uYmu7u7FAqFI7175eS5UwkKOGfxGgwGJVDZIr5UKjE+Pq72LzgrpM+bSqXI5/Mc\nHBygaZryc4PBIOPj46pouhuR70Wu7JWJn2KxqNrKbm9vv9ZOQIq6vdRU1oS0b1+1t7enuqY/evSI\npaUlcrkcpVIJg8GgJtdjY2PKlp10v849SSGD58FgEJvNRr1eP5cQVaPRUKlN2bdBCKEmfOFwmKmp\nKWKxGB6P50zPfVHIVdKBQEA1AikUCuzu7rKxsUEoFFKrumXTEI/Hc2RElj2O8/m86hy5u7tLIpFQ\nHZBkh890Oq1KJl0ul+psND09zfT0dMdtea7ileEpo9GoGt6dNXLUkEmQzc1Nstms2qFGtqAKh8NM\nTEx09chrt9tV2NDj8WCxWFTv3Y2NDbxer4r3yuaCso8GHI68MiKTyWRIJBJK+PKxubl5pDWt2WzG\nZrPR29vL6Ogo09PTTE1NMTY2prJwneLS1fOelvX1dZ4+fcrc3JyqaUgkElSrVdWeXmacQqEQPT09\nXdvqSfqawWBQZRbll3d9fZ1arcbq6qoKLQYCARXPlkjxypFXVvPJh1zaLkNigUCAaDTK2NgY09PT\n3Lp1i8HBQex2e6fMoOh68a6trfHjH/+Yzz77TKWiZbGKHKUGBweJRCKEQiF8Pl9XxXbbkb10g8Eg\n0WiU4eFhNjY2SKfTqidbezun9mSMRIpX7itRrVZVTFcu/4Ev17DJDNvt27dVp01Zyddpul68shJN\n9iirVqvKXfF4PAwNDTE1NUUoFFLtPLuR9lJQn8/H5OQk5XIZq9V6ZOmTHDnhyzi77EkMqIIl2dO4\nPYkja0nsdruqJ7lx4wY3b95kcnKSwcFBtdDyMqTVu1687fXBMl0pSyd9Ph+xWIwPPviAcDjc9VVk\nEq/Xy9TUFE6nk3q9TjKZVDW3MsoihFBVdfILKxuHyFFWpnZl9lE22JOtsyYmJpicnGRiYoJIJKIq\n+07Se/ki6Erxyi4usiWRvP21JySkcEdHRxkdHSUQCLw3Cy3tdjuRSASHw6GW61gsFuLxOJlMRo2y\nckPsWq2mXA45Crc3GpEbYcvUfCgUYmpqiqmpKYaGhhgYGOjYCuGvoys/zXq9TiaTIZlMqgbHsn29\n0WhkaGiI69evc/v2ba5du0ZfXx8Oh6NrXYZXka1IZddyWY03Pz+vKr8AZZ96vU5vby+9vb2vFfSb\nTCYGBwcZGBigt7eXnp4e/H6/ilR4vd5Lm9TpSvE2Gg3S6TSrq6tHxCv3txgcHOSjjz7i9u3bjI+P\nX1jXwouivWWpjF3L23p73NVsNrO/v0+lUiEYDDIyMqIiEVK8VqtVRRHC4bDaAqwb6CrxysKR9h0d\nNzY2qFQq2Gw2NWqMjY0xNjamMkDdGl34Ktrfj2wiEg6H1dIdyfXr1/nwww9fG3nbJ1wmk4loNEpf\nX59K93aLvbpevOvr60q8fX19DA8Pq/RlNBrtaO79IpAZTNnkrn3Jk/R55e6WUpjtk61Xt7vtJteq\nq8QLRxvAaZqmurY4nU7GxsYYHR1lYmJCjSTvM+2liXLt2VWi68QLh75cb28vY2NjalTx+/0MDw8T\ni8Xo7++/ch/kVaTrxCt3tAkGg0xMTOBwOAgEAvT19RGNRq/cfhNXma4Sr5xIWK1WIpHIkWXxLpdL\nH22vGF0lXviy60s0GiUYDB6pXOumyYbOu3NS8doA5ufnz/FSuoM2G5xl5F6370tOZd/2Zc5f9QC+\nB2j648jjeyexnW7f87OvOMlSZiFEAPgOsAa83mf0amEDYsAfaZqWOYsD6vY9wonteyLx6uhcRjpf\n16aj85bo4tXpWnTx6nQtunh1uhZdvDpdiy5ena7lTMUrhGgJIZovf776aAohfnCW53vLa/zvv+I6\n60KIS906skvs2y+E+CMhxLYQoiqEWBdC/LYQ4t06KR53rrOM8woh2tfb/Arwm8A1QJbmFzVNKx/z\nOqOmaSfbSO0dEUJYgVdb9/w+UNE07Zcu4hreli6xby/w3wAPgMzL6/s/gf+sadrfPNOTnVWK85iU\n518D9o75+3eAFvAXgC+AGnAf+DfA773y3H8J/Pu23w3AD4AXQAl4CPzld7zOKFAHfvm8bHGV7fvy\nuL8BLJ61DTrp8/4j4O8AU8DiCV/zm8BfAX4V+AD4F8AfCCHuyycIIXaEEH//FNfx14E94A9P8Zpu\n4FLYVwgxAPzXwH8+6WtOSqdKIjXgH2ia9hP5hzct+hNCOIG/B3yqadqTl3/+HSHEnwX+FvDZy78t\ncXi7Oil/A/hXmqa9vr1m99Jx+woh/h/gFzmsVfi/gV8/3Vt4M52s5314yudPcmiIn4qjn4QZ+Jn8\nRdO0P3PSAwoh/hwwAvzOKa+lG+i0fX+Nw7nFFPCPgX/C4ZfjzOikeEuv/N7i9ehHe38mF4cjyp/n\n9W/+21Zi/U3gv2iatvCWr7/MdNS+mqYlgASwJIQoAj8SQvyvmqblTnusr+IyraRIAXde+dsdIPny\n3zNAAxjSNO3Bu55MCOHlcFZ85rezS8qF2vcV5BKXM+1DcJnE+5+AXxdC/FXgEYe+6DgvjatpWlYI\n8c+Afy6EsHF4K/MB3wKSmqb9PoAQ4qfA72qa9iZX4PscRhn+4DzezCXkQuwrhPhLL1/3kMPR/zbw\nW8B/1DQtedxr3pZLI15N0/5QCPFPgd/m8Hb2f3EY3hlue85vCCG2gX/Ioa+a5dBI/1vbocaAwAlO\n+avAH2jHxEXfRy7QvjXgfwSuvzzPJodx9N86szfzEr0YXadrOdHIqy9TOYK+DOh8Obl9T5gh0RcI\nvsUCwVNkoHT7voV9T+rzrgH88Ic/ZGpq6oQveT+Zn5/n+9//Pry0yRmxBrp94XT2Pal4qwBTU1Pc\nu3fv7a/s/eIsb++6fV/njfbV63l1uhZdvDpdy6WJ8+pcHtrDp5VKRT2q1SrValVtCyu3B+jURjW6\neHW+Ek3TSKVSrK6usr6+zvb2Njs7O9y5c4c7d+4QjUZxu926eHUuF3L0TSaTPH36lIcPHzIzM8Ps\n7Cy//Mu/jNvtxul0YjabO9ZaVhevjqLZbNJoNKhUKmp/t8ePH/PFF1+wuLhIOp1We+AVi0XK5TKN\nRufKoHXx6iiazSYHBwfkcjmeP3/Os2fPePr0Kc+ePWNzc1PtUyzFWyqVqNfrHbteXbxXHE3T1MbZ\nhUKBTCbD1tYWc3NzPHr0iJWVFeLxOPl8Xm1cKLdx7fQexLp4rziaplGv16nVaiQSCVZXV1lcXOTp\n06fMzs6SyWQol8uYzWa8Xi8+n49QKER/fz9+v7+ju2NeCvG2h2YajYba3Lkt938Eg8GgtiGV2zl1\ny8Z3lwVpV0073NeuUCiws7PDwsICT58+ZW5ujufPnx/ZzzkYDBKNRhkYGKC/vx+fz6eLF768fa2s\nrLC8vEwymaRUKlEuf1luK/cW9ng8TE1Ncf36dRwOhxKyzunQNI1Go8HOzg4vXrxgbm6OJ0+esLCw\nQDKZpNVq4fV6CQaDDA4OKptfu3aNUCiE3W7vqN0vzSfearWUeH/84x8zPz9PMpkkk/myKk5uPRqN\nRvnud79Lf3+/ErTO6ZEuw87ODjMzMzx58oSZmRlevHhBvV6n2Wzi8/kYHR3l1q1bfPzxx9y/fx+H\nw6HCZJ2843VUvNItqFQqpFIp0uk0z549Y25ujqWlJfb29igUCjgcDux2Oz6fD6/XS19fHx6PB7PZ\njNFofKMBNU2j2WzSarXU7kGdnGhcBsrlMnt7eyQSCebm5piZmWF5eZlEIkG1WsXlctHX18fk5CT3\n7t3j1q1bTExMEAqF1MbdnXbVOi7eVqtFPp9Xk4THjx+ztrZGJpOhVqthMBjwer309/czMjLCxMQE\n165dY2pqCo/Hg9VqfaMQ5e2xVqthNBpP9Jr3nUKhwPLyMnNzczx8+JCnT5+SSCQoFouYzWb6+/sZ\nGhri3r17fPrpp0xOTuL3+y/VXa6j4q3X61SrVZLJJAsLC/zsZz9jdXWV3d1dSqUSZrMZj8dDNBpl\nfHycqakpbt26xfj4OIFAAKfTqSZtxyEFW6vVKBQKFAoFLBYLXq8Xp9OJxWLp+K3vItE0jYODAw4O\nDtjd3WVhYYHPP/+c2dlZVldXqVarmM1menp6GB4e5ubNm0xPT3Pz5k1isVinL/81OirebDbL1tYW\n8/PzR1yFWq2G0+mkv7+faDTK9PQ009PTDA8PEw6H6e3txel0vtFlyOfzrK+vs7Gxwc7ODtvb2/h8\nPmKxGENDQ4RCIUKh0JWa7CWTSba2tpidneXRo0fMzMyQSCSo1+u43W5CoRDDw8N8+OGH3Lt3j9HR\nUTweT6cv+1g6Lt7l5WUVmllaWlIBc5/Px+DgIB988AGffPIJn3zyCb29vVgsFkwm04l8rnw+z9LS\nEg8fPmRxcZGlpSXC4TD37t2jVCphNBrp6+v72mO8bySTSZ49e8bDhw/54osvmJmZodls0mw2lXBv\n377Nhx9+yDe/+U1cLhcWy5m2WzgzLly8Mr14cHCgRoC5uTl2d3ep1+tqQtXT08Pk5CT3799nfHwc\nv9+Pw+FQk4WvQtM0arUa1WqV3d1dVldXmZ+fZ3Nzk2QyicPhoFQqcXBwQKPRODaO/L7RaDSoVqtU\nKhXW19eZn59neXmZVCpFo9FQJY5yJcft27cZGRnB7XZjsVgu7fzgwsXbbDapVCrqlv7s2TNmZ2fJ\n5Q67AMkUSRcNAAAP90lEQVT60EAgwPXr1/nkk0/w+XzY7fYTz3ArlQrZbJbt7W1evHjB0tIS+/v7\nlMtl6vU6rVbrvN/mpaLRaJDP59nb22NtbY35+XlWVlbI5XIYDAZ6e3uJxWLcuXOHjz/+mFu3buH1\nejue/n0THRFvNpslHo+zvr7O+vo6Ozs7NJtNhBB4PB78fj+jo6OMjY0xMjKiRHvSiVU+n1fH39ra\nYnd3V01UDAYDdrsdl8uF1Wq9EpO1Wq1GKpVibW2NFy9esLGxoeLn0j27ceMGN27cYHJykuHhYWWX\ny2yfCxfvwcEB8XicJ0+e8OLFC/L5PM3mYdNug8FALBZTPlc0Gn2reGI6nWZhYYHFxUWSySS1Wk3F\neJ1OJ+FwmMHBQbxe76UeWc6KSqXC2toan332Gaurq+RyOYxGo4qZS3dhYmICr9d7qQXbTkfEu729\nzdOnT4+IV2bKYrEYP//zP8/t27eJRCJqxD2Nb5pOp9UELZVKqXix0WjE5XIRCoUYHBzE7XZfCfGW\ny2XW1tZ48OABm5ubqkIsEAgwMjLC1NQUd+/eJRwO43K51Osuu4g74jaUSiXS6TSFQoF6vY7FYsHj\n8eDxeBgYGCAajRIMBrHb7QDHNpw47rjSNUilUuzs7JDJZKhUKhgMBnw+Hz09PUSjUQKBgHIbrgLS\n5plMhlKpRKPRwOfzEYlEmJycJBqN4vV61Zq0yy5aSUeiDcVikXQ6TT6fp16vYzab8fv9RCIRBgcH\niUaj9Pb2YjYfto9tF62sgXiVer1OsVikWCySTCbZ3d0lnU5TLpcxGAz09PQwNDTE4OCgSnB00wf1\nLjQajSPibTabOBwOBgYGuH79OtFoVC2o7KY7UUfEW6/XqVQqHBwc0Gq1MJlMWCwWXC4XTqdTxRYP\nDg4olUrk83kKhQLlcplqtcrBwcFrx63X65RKJYrFIgsLC+zu7lIsFmm1WthsNvr7+5mcnCQWi9HT\n03NpY5dnicwu5nI59vf3KRQKALjdbvr6+hgaGmJ0dFTd5botWXNprra9eEamMWV4R5ZJJhIJ5W68\nivxSyAmhzBpJP3dwcJDp6WnGxsYubcborJHu2dbWlioqDwQCKpoTi8UYHBzE5XJdqpqFk3Lh4hVC\nqJFW1iW0Wi0ajQYHBwdUq1XK5TL7+/skk0ni8ThffPEFDx8+ZHV1lXg8zt7e3onOZbVacTgcymW4\nceMGsVgMt/tS7xV4ZpRKJXZ3d5XNqtUqdrudSCSihBuJRN7pHO3zj4t2wS5cvBaLhUgkwu3btzGZ\nTFSrVfL5PLlcDiEEjx8/xmQyYbPZ2N3dZXd3l42NDTY3N0mn01SrJ28RZrfbiUajjI6OMjQ0RDAY\nVFmjq0ChUGB7e5utrS11t/J4PEq07/olfnXirGnahQq4Y+Kdnp6mWCyqW1oul6NcLmMymchms7Ra\nLba3t0kkEsp3kwXSJ8XhcBCNRpmammJ4eFiJt5smJe9CsVhU4s3n8wB4vV41KW4Pi70tUsCdmPhe\nGrehXq/TaDRIJBIcHBzQbDbJZDIqJilXrtrtdiwWiyomKRaLFAoFarXaa+eyWq0qiiHXW3XbpORd\naI/AyEmutKPL5TpVOWiz2TxSXlosFqnVaqpvQzAYJBgMqlrpixggOlaYUywWqVaraiSVGTAZ+5XP\nk1kxmQ3q7+/H6/Wqvlmbm5usra0dK15ZD9zb24vL5boyI66kPfYt7Ww0GtVgcJpJWr1eJ5/Pk81m\n2djYYGNjQ/nRQgju3bvH3bt38fv9F1YT0ZGRVwiBwWDAarXidrvxeo/uY61pmlpoabVa6evro6+v\nj2g0ytDQED6fj/39fXK5nCqsbkdm05xOJ36/n76+vq6dUZ+W9iTOwcEB5XL5SHMQo9GoFqx+lcBe\nXVncPoFOJBIsLS2xtLREMpmkXC4jhMBisdDf368+t4u4w124eE0mE319fVy7dk2VQEaj0dee53Q6\nCQQCBAIBenp61MPn82EwGFhfX6fRaBw7gjidTtxuN5FIhIGBAQYHB/H5fFdGvLJtUz6fJ5FIqOU9\np0GGLVdWVnj27BkvXrwgk8mwt7dHMpkklUopd8RkMjE3N4fL5aJQKDA5OYnD4Tind/glFy5es9l8\nxDeyWq0kk69vz9Xb26uiBG63W0UJjEYjxWKRZrNJOp0+tt7U4XDQ29urMnYDAwOYTKYr4+/K5U/7\n+/uq59hxbtVXITOZjUaDlZUVfvSjH/Ho0SPy+Tz7+/vU63Xl2mmaphJMMtog75LnzYV/mrIkUQjB\n4OAgJpOJgYGB157n8XgIhUIEg0FsNhtWq1Wlc2U9RPukrx2/36/KKf1+PxaL5UqkgYEjafT2NXwn\nidKUy2WKxSL7+/uk02nS6TSfffYZy8vL7OzsqD698vgSIQSVSkVlQS+q+V5HxCtH0HA4jM/nO7ZZ\nm9lsxm63K9G21yFIn9lkMh1bnxAIBLh27Rrj4+P09PRcuYmaFLB0H066YqRUKhGPx9nY2FBVeSsr\nK6ytranqPznavkqz2VQRo4sq9u9YqMxkMmG1Wl+brJ0UOSl7td5XCKH83VAohMvlujKj7qtIGxmN\nRtU+qx3ZEkBGJOLxOAsLC8zPz7OwsMDCwgLpdJpcLvdGt0OOxl9V9XcedLUT+FWilFEMGcu8SrTf\nlWw2m5ovyFu+RLoV1WqVVCpFKpVidnaWBw8eMDs7q9wGuXTqTXzVYHKedK14j1umIsNwVqtVVahd\nNfHCl+sA7XY7Ho8Ht9ut1g5K5KhbrVZJJBKsrKzw6NEj/uRP/oSZmZlTn7N9lL8oN60rxVuv10km\nkywuLhKPx6lUKpjNZjXKhMNh+vr68Hq9V6bgXNI+8trtdtxuNx6PRwlXLn3P5XKqSWE8Hicej7O6\nunqkN9ybkKO71+vl2rVrfPTRR0xNTeHz+c7r7R09/4Wc5Yyp1+skEgkWFhaIx+Oqf2xPTw/hcFiJ\n96rEdl9FpmdtNpsaeeXq7EQiwczMDKurqyoDl81myWazKu17UuSAIXuaffzxxyrCcxF0lXjlhKBe\nr1MoFI6sxpAF59euXVMVU1dt1IUvXSdAdcCJRCJKvPv7+2xsbGA0GlUYTbaSPWmUwG63q1j6wMAA\nsViMqakpRkZG6OvruzC7d5V42xvmtefsNU3DZrOp1lDRaLSjTY8vC16vl5GREfb29tjZ2VHxWJlM\nkKG0er1+qgiB3+8nGo0yNjbGjRs3lHDlChXd5z0GeZuTRTmyTFKKNxQKMTU1RSgUupKj7qt4vV6G\nh4fJZrPMzc1hMBhUN6GTIn3o9glZOBxmcnKS6elpPvzwQ27evIndblcLOC+KrhKv3KkmlUopH032\nZJDVUlar9WuLTq4SDoeD/v5+IpEIPT092O32I6ndk+B2u1Vdid/vx+/3MzExwfj4OLFYTN3lvq5b\n53nRVeKViwlfFS8cznzNZnPHDHkZkbv2ZDIZ/H6/aiVwmvStXHkRi8UYGRk58lOm3mX6XRfv17C/\nv8/Kygqzs7Ps7OxQq9XQNO3I6uOenh61rP2qIzOZ7dtOSZFJezkcDiVAh8OBy+VS8wUhhGozG41G\niUQiRKNRVaJ6EZVjX/v+Onr2U7K3t8fs7CyfffYZW1tbqtNOe7yxr69PFezofDUOh4NIJEIkElEN\nX2Qnod7eXjWKulwuPB6PSvo4nU61mqXTdNUnXCgUWFtbY3FxkVwuR6vVOjJquN3uCwuQdytyTtDX\n18f4+DgTExP4/X7VDX18fFy12YIvs3WXcQ7RVeJt75gjC0BkcU+n9wTrBoQQRCIRgsEgo6Oj3Lhx\ng4mJCRwOBw6HQ7XEap8zXIaNU76KrhRve4WULt7TEQ6H1T4TN2/eZGJiQtUjtBfXdANdJd522pdc\nX+Zb22XA4/Fw584dCoUCIyMjjIyMqD05nE5nV/TiPY6uFa9E3771zfj9fj799FPGxsZwuVxq8uVw\nOLrabl0vXpkuLhQKZLNZkskkdru9KxvHnRcOh0ONuO8TXX+frVQqpNNpNjY2mJ+f5/PPP2djY+NI\n7arO+0nXD02VSoVarYbJZGJ+fl5lkWRrJ533l64aed1uN8PDw0xOTqpdx2UmTYbPdP/36tBVI6/H\n42FkZIRMJkOj0SCTyag+Z7LQRBfu1aGrxOt2u4nFYtTrdbXIUi4Q9Hq9XL9+nYGBAXp6eq7k2rWr\nRleJ1+PxMDo6Sk9PD2NjY3z66adq1LVarfT29qrO39L31Xl/6SrxygWWg4ODnb4UnUtAV03YdHTa\n0cWr07Wc1G2wAczPz5/jpXQHbTY4yyog3b4vOZV9j9td8pjdJr8HaPrjyON7J7Gdbt/zs684yZJn\nIUQA+A6wBpx86en7iQ2IAX+kadrJ28t8Dbp9j3Bi+55IvDo6lxF9wqbTteji1eladPHqdC26eHW6\nFl28Ol2LLl6druVMxSuEaAkhmi9/vvpoCiF+cJbne1eEEH1CiMTLa+t8C5g30E32FUL8d0KIGSFE\nVQixK4T438/6HGddVRZq+/evAL8JXANkhfixbbeFEEZN006+nfvZ8bvAA+AvduDcb0NX2FcI8b8A\nfwv4n4GHgAs4+1LAs0pxHpPy/GvA3jF//w7QAv4C8AVQA+4D/wb4vVee+y+Bf9/2uwH4AfACKL00\nzF9+y+v7u8D/C/wi0AQs52WLq2RfIMhhlvCT87ZBJ33efwT8HWAKWDzha34T+CvArwIfAP8C+AMh\nxH35BCHEjhDi73/dQYQQd4C/B/x1DvPo7yOdsu8vAhVgSgixIITYEEL8nhAi/DZv4uvoVDG6BvwD\nTdN+Iv/wprVnQggnh4L7VNO0Jy///DtCiD/L4S3qs5d/WwK+MicuhLAD/xr425qmJd7TNW8dsy8w\nCtiB/wn4NaAM/BPgPwgh7mqadmbbY3ZyJcXDUz5/ksOijZ+Ko5+EGfiZ/EXTtD/zhuP8H8D/p2na\nv335u3jl5/tCp+xrePma/0HTtD8FEEJ8D9gCfg746Smv6yvppHhLr/ze4vXoR/sqSheHI8qf5/Vv\n/mkqsf4cMC6E+G9f/i5ePgpCiB9omvaPT3Gsy0yn7Lvz8qcqzNU0bVsIkQeGTnGcN3KZ1rClgDuv\n/O0OkHz57xmgAQxpmvbgHc7zXaB9t5VvcThx+YjD0eF95aLs+6cvf07ycsQWQoQAD7D+Dsd9jcsk\n3v8E/LoQ4q8Cj4C/AYzz0riapmWFEP8M+OdCCBuHhvFxKL6kpmm/DyCE+Cnwu5qm/c5xJ9E0baX9\ndyGEDOHMa5p2cPZv69JwUfadEUL86OVxfo3DydtvvTznnx73mrfl0mTYNE37Q+CfAr/NoeEEh+Gd\n9uf8xsvn/ENgDvh3wH/FYRG3ZAwInP8VdxcXbN9f4XAk/w/Aj4Es8F3tZSztrNCL0XW6lksz8uro\nnBZdvDpdiy5ena5FF69O16KLV6dr0cWr07Xo4tXpWnTx6nQtunh1uhZdvDpdiy5ena7l/wfLqX32\nAzPVUQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f3fc3435d30>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# load the complete MNIST dataset\n",
    "from xiaoloader import load_mnist\n",
    "dataset = load_mnist()\n",
    "\n",
    "import numpy as np\n",
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "images = dataset.train.images[0:4]\n",
    "labels = np.array([label.argmax() for label in dataset.train.labels[0:4]])\n",
    "plot_images(images, labels) # MNIST are 28 x 28"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "BATCH_SIZE = 250\n",
    "\n",
    "EPOCHS = 4\n",
    "\n",
    "M = 784 # 28 x 28 pixels\n",
    "\n",
    "N = 512 # hidden neurons\n",
    "\n",
    "P = 10 # number of possible classes\n",
    "\n",
    "steps = len(dataset.train.images) // BATCH_SIZE # 220 steps\n",
    "\n",
    "w1 = np.random.normal(0.0, pow(10, -0.5), (BATCH_SIZE, M, N))\n",
    "\n",
    "w2 = np.random.normal(0.0, pow(10, -0.5), (BATCH_SIZE, N, P))\n",
    "\n",
    "b1 = np.random.normal(0.0, pow(10, -0.5), (BATCH_SIZE, N))\n",
    "\n",
    "b2 = np.random.normal(0.0, pow(10, -0.5), (BATCH_SIZE, P))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def equal(y_predicted, y_true):\n",
    "    \n",
    "    total = 0\n",
    "    \n",
    "    for j in range(BATCH_SIZE):\n",
    "        if np.argmax(y_predicted[j]) == np.argmax(y_true[j]):\n",
    "            total += 1\n",
    "            \n",
    "    return total\n",
    "\n",
    "def eval_data(dataset):\n",
    "    \"\"\"\n",
    "    Given a dataset as input returns the loss and accuracy.\n",
    "    \"\"\"\n",
    "    steps_per_epoch = dataset.num_examples // BATCH_SIZE\n",
    "\n",
    "    num_examples = steps_per_epoch * BATCH_SIZE\n",
    "\n",
    "    total_acc, total_loss = 0, 0\n",
    "    \n",
    "    for step in range(steps_per_epoch):\n",
    "    \n",
    "        batch_x, batch_y = dataset.next_batch(BATCH_SIZE)\n",
    "        \n",
    "        total_loss += train_SGD({x: batch_x}, batch_y, [w1, b1, w2, b2])\n",
    "        \n",
    "        total_acc += equal(g.value, batch_y)\n",
    "\n",
    "    return total_loss/num_examples, total_acc*100/num_examples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def train_data():\n",
    "    \n",
    "    start_time = time.time()\n",
    "\n",
    "    steps = dataset.train.num_examples // BATCH_SIZE\n",
    "    \n",
    "    for i in range(EPOCHS):\n",
    "\n",
    "        for step in range(steps):\n",
    "        \n",
    "            batch_x, batch_y = dataset.train.next_batch(BATCH_SIZE)\n",
    "        \n",
    "            train_SGD({x: batch_x}, batch_y, [w1, b1, w2, b2])\n",
    "\n",
    "        val_loss, val_acc = eval_data(dataset.validation)\n",
    "        \n",
    "        msg = \"EPOCH {} ... Validation loss = {}, Validation accuracy = {}%\"\n",
    "        print(msg.format(i+1, val_loss, val_acc))\n",
    "    \n",
    "    time_dif = time.time() - start_time\n",
    "\n",
    "    print(\"Time usage: \" + str(timedelta(seconds=int(round(time_dif)))))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "beginning training\n",
      "\n",
      "EPOCH 1 ... Validation loss = 0.12801770217298866, Validation accuracy = 10.36%\n",
      "EPOCH 2 ... Validation loss = 0.12893329311275625, Validation accuracy = 10.36%\n",
      "EPOCH 3 ... Validation loss = 0.1286522747955876, Validation accuracy = 9.46%\n",
      "EPOCH 4 ... Validation loss = 0.1297848246554159, Validation accuracy = 10.02%\n",
      "Time usage: 0:02:04\n"
     ]
    }
   ],
   "source": [
    "x = Input(w1, b1)\n",
    "\n",
    "h = Hidden(x, w2, b2)\n",
    "\n",
    "# r = Relu(h)\n",
    "\n",
    "# h = Hidden(r, w2, b2)\n",
    "\n",
    "g = Softmax(h)\n",
    "\n",
    "cost = CrossEntropy(g) # numpy.mean(CrossEntropy) over BATCH SIZE\n",
    "\n",
    "print(\"beginning training\\n\")\n",
    "train_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Test loss = 0.1315336948251015\n",
      "Test accuracy = 9.38\n"
     ]
    }
   ],
   "source": [
    "# Evaluate on the test data\n",
    "test_loss, test_acc = eval_data(dataset.test)\n",
    "print(\"\\nTest loss = {}\".format(test_loss))\n",
    "print(\"Test accuracy = {}\".format(test_acc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def plot_images(images, labels, pred_labels=None):\n",
    "    #assert len(images) == len(labels) == 5\n",
    "    \n",
    "    # Create figure with 2x2 sub-plots.\n",
    "    fig, axes = plt.subplots(2, 2, figsize=(2,2))\n",
    "    fig.subplots_adjust(hspace=1, wspace=1.9)\n",
    "\n",
    "    for i, ax in enumerate(axes.flat):\n",
    "        # Plot image.\n",
    "        ax.imshow(images[i].reshape((28, 28)), cmap='binary')\n",
    "\n",
    "        # Show true and predicted classes.\n",
    "        if pred_labels is None:\n",
    "            xlabel = \"True: {0}\".format(labels[i])\n",
    "        else:\n",
    "            xlabel = \"True: {0}, Pred: {1}\".format(labels[i], pred_labels[i])\n",
    "\n",
    "        ax.set_xlabel(xlabel)\n",
    "        \n",
    "        # Remove ticks from the plot.\n",
    "        ax.set_xticks([])\n",
    "        ax.set_yticks([])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def plot_example_predictions(predicted):\n",
    "\n",
    "    labels = np.array([label.argmax() for label in dataset.test.labels[10:14]])\n",
    "    \n",
    "    pred = np.array([label[1].argmax() for label in predicted])\n",
    "    \n",
    "    images = dataset.test.images[10:14]\n",
    "    \n",
    "    plot_images(images,labels, pred_labels=pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAPIAAAC4CAYAAADZq7GuAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAAPYQAAD2EBqD+naQAAIABJREFUeJztnXt0VNd97z9bM6P3W0LoAZIQCI0ECMRLYGMBhmDkOHZw\nXD9I4nY1Se2ueDVNH2mT3uU0TeKsOvc2qy+3va3dmwY7sZPYxibUUEzAgMBG5iWhB6AHkkCjt0Yj\nzUgjzez7xzkaD4MQkiwY6Wh/1po10t579vn95ne+Zz/PGSGlRKFQzG5Cgm2AQqH49CghKxQGQAlZ\noTAASsgKhQFQQlYoDIASskJhAJSQFQoDoISsUBgAc7ANmAxCiCTgAaARGAyuNWMSDmQDB6SUXUG2\nZUajYjm9zCohowX+1WAbMQG+CLwWbCNmOCqW08hsE3IjwJ49e8jPzw+yKTdTXV3Nl770JdDtVIxL\nI6hYThezTciDAPn5+axevTrYtozHTOwqzjRULKeR2SZkhQKv14vD4eDdd9/l0qVLuFwuMjMzuffe\ne2f6ReGOMaeE7PV6cTqdmEwmQkNDcbvd9Pf343A4cDgchIeHExERAUB4eDiRkZFERUUhhAiy5QoA\nKSVSSnp6eqipqeGNN97g1KlTDA4OUlJSwoIFC5SQ5wIjIyM0NDQQHR1NWloaNpuNqqoqzp49y7lz\n58jMzCQrKwuA7Oxsli5dSl5enhLyDMLj8VBVVcVrr73GuXPn6O7uJi4ujoULF5KQkBBs84KG4YUs\npcTtdlNZWcnhw4eprKzE6/USExODw+Ggvb2dtrY2Ojs7qaysJCYmBoD4+HiSk5OZN28e9913H6tX\nr2bx4sVB9mZu4/V6OXPmDAcOHODo0aPExMTw0EMPsWHDBjZu3EheXl6wTQwahhey1+uloaGBI0eO\n8Nprr9HY2MjQ0BBms5nQ0FCEEEgpMZlM9Pb20tvbi5SSmpoaRkZGsFgs2Gw2BgcHmTdvHhEREVgs\nlmC7NSfxer3U1dVRUVFBY2MjmzZtYtu2bXz2s58lPT2d8PDwYJsYNAwvZI/Hw6FDh3j33Xc5f/48\no09EsVgsJCcnk5iYSExMDJGRkb4utNfrpbKykubmZgYHB/nNb37D8PAwq1atIisrSwk5SEgpcTgc\nOJ1OLBYLS5YsYdmyZSxatCjYpgUdQwt5YGCA69evc+bMGerq6gCIjY1l/fr1bNmyhfT0dOLj44mK\niiIiIsInZCklhw8f5v3336esrIyhoSGuXLnCL3/5Sx599FHy8/MJCwsLpmtzFrfbzcjICGFhYeTm\n5pKenq7mMDC4kIeGhujq6uLq1at0dHRgsViwWq1s27aNXbt2ER8fT2RkJKGhoTd0s6WUREZGEhYW\nRl9fH1evXuX69escPHiQgoICUlNTSU1NDbZ7c4qRkREGBga4evUqdrudlJQUlixZQkpKSrBNmxEY\nWsher5fBwUEGBgZwu93Ex8ezefNmSkpKbjkxIoRACMGaNWuIiYnB7Xbz5ptvUllZyenTp6msrCQv\nL08J+S4zNDREe3s75eXltLa2snbtWrKzs0lMTLztZ71eL4EPmRyNs1Fac8MLeXh4GK/XS0ZGBsXF\nxezcuXPCWwIzMjJ48sknuXz5sq8lqK2tpa6ujqKiojtsvcKfpqYm3n77bWw226Q/29bWRnt7O319\nfQBERUUxf/58EhMTffsGZjuGFnJXVxfl5eXY7XaSk5NZtWoVixcvnvB6Y1RUFEuXLiU3N5e0tDTs\ndjuNjY20tLTcYcsVgbS1tfHBBx/Q3d1NUlISS5YsGVeE9fX11NfX09nZybVr12hra7tJyCtXrsRq\ntfr2DsxmDCtkr9fL1atXeeedd+jo6GDlypUsWrRoSksU2dnZZGdnU11djc1mo6Oj4w5YrLgVo7u5\nKisrGRgYYNWqVZSUlBAfH+/Lh096YE6nk6NHj7J3717Onz9PX18fQ0NDeDwePB4PISEhREREUFpa\nyqOPPsqCBQsICQmZ1d1swwq5t7eX9vZ2enp6MJlMJCUlsWDBginNNi9cuJDMzMw7YKViIgwPD+N2\nu30beZYuXcqmTZuIjY31lfF6vb5NPXv27OH8+fM0NzfT399PaGgoycnJLFy4EJvNRmdnJ/39/Zw+\nfZqsrCzuu+8+4uPjZ/VKhGGF3NDQQG1tLZ2dnSQlJbF48eIpt8ije64VwaGxsZH6+nqcTieLFi0i\nJyeHefPmAdpylMPhoKGhgTNnznD8+HEOHz5MX18fERERFBQUkJeXR25uLllZWfT09FBdXc1vf/tb\n7HY7zc3N9PT0EB0drYQ8E6mqquLChQv09fWxdu1ali1bplrVWUp1dTWVlZU4nU4yMzPJyMjw5fX1\n9XHlyhUOHjzI/v37KS8vJzQ0lNTUVF/LXVJSwvLly0lISEAIQVlZGV1dXXz44Yf09fXhcDgYGRkJ\nooefHkMLua6ujtjYWO655x4KCwuDbZJiiozOOgshSEpKIi4uzpd36dIlfv3rX3PgwAGampqIiIgg\nNzeX0tJStm3bxpIlS3z7BUYJDw8nNTXVUFs6DSvkgYEBPB4PqampqjWe5djtdt+Ms8ViwWw2++5J\nrq6u5tixYzQ2NhIfH8+KFSvYuXMnxcXF5OXlERcXR0jIjc+YDAkJISwsjOTkZNLS0pg/f/6s7laD\nAYU8ujNLSklYWBiJiYlkZGRMaOOAYmbidDpxOp03pI2MjNDU1ERVVRW1tbVYLBZWrlzJI488wq5d\nu0hMTMRkMt2yTq/XS0JCAunp6aSlpY1bdjZgOCF7vV6GhoYYGRnB6/UG2xzFNDA0NMTQ0JDvAg3a\nJFdZWRkXLlxASklRURFf+MIX2LVrFzExMTe1wv54PB4GB7Un+JjNZszm2S8Dwz3X2m63c+7cOZqa\nmnA4HNNS57Vr12hubga0btl4J4li+hkVsP86r8fjob29nd7eXt8WzMjISBISEjCbzbdcE3Y6nbS0\ntPDRRx/hdrt9N8vM5jVkMKCQHQ4HNTU12Gw2BgYGPlVdIyMj9Pf3+7ZlCiGYP38+ycnJ02StYiL4\nC230PSQkhMTERN8YeHQH1+i947fqjTU0NHD27FkuXrxITEyMYfbMG07I04nT6aShoYELFy74hLx5\n82Y2bNgQbNPmPJGRkezatYstW7YQFhZGU1MTx44d4+DBg7S3t+N2u8f83Lvvvsu7775LeHg4Gzdu\nNMye+dk/OLhDdHV1cfnyZY4dO8bly5cRQrBkyRKKiorIyckJtnlzguHhYQYGBrDb7TdNdoWEhJCc\nnMyOHTuQUrJv3z7q6urYs2cPtbW1lJaWsn79eiwWC9euXaO2ttb3mCC3283OnTtZu3Yt6enpQfJu\nejGckE0mExEREYSGhk5pJnJ0MqWxsZGysjIOHDiAzWYjJSWFzZs3k5+fT1JS0nSbrRgDKSUej8c3\nPpZS0tfXR39/P16vF4vFwooVK0hMTEQIQXl5Od3d3dhsNnp6euju7qanp4fy8nLKyso4ceIEIyMj\nWK1W3wMi/NekZzOGE3JYWBgpKSnExMQQGho6pTq8Xi/nzp1j//79HDt2jOjoaIqLi3n66afJzMyc\n9RMjswWLxUJCQgLz588nPj4em81GY2Mjzc3NuN1uQkNDCQ8PJycnhz//8z+ntraWpqYmFi5cSHR0\nNA0NDbz55pv8z//8D5cuXcJkMlFaWsquXbt47LHHZv2Skz+GE3JMTAzLli0jPT2dhoYGXC4X/f39\nDA4O3nInj9vtprOzk+rqampra2lvb+fEiRNUVFTgdrvZvn07n//857FarcTExCgh3yVGJ7k+85nP\nMDg46NsbvXfvXvr6+li3bh2pqam+2xmbm5uprq7m+PHjtLe3c/36da5evYrT6aSgoID77ruPLVu2\nsGrVKsM9d81wQh7dfhcbG4uUko6ODk6fPs3IyIjvUbeBuFwu36b7ixcv+h6POzw8TGZmJps2beKe\ne+5Rs9VBYsWKFQwODtLd3c2RI0e4ePEinZ2dNDY2kpaW5hNya2sr9fX1NDU1ce3aNVwuF8nJyRQW\nFrJhwwbuv/9+li5dasg4Gk7IowghcLlc2Gw2Xn75ZVJTU285HnI6nVy6dInu7m7fRoGYmBgyMzNZ\nt24dK1euNMwyxWwkJiaGjRs3smjRIlwuF4cOHaKyspKKigpfmcB15tGlwpKSEnbv3s2mTZsM3Zsy\nrJBLSkpwOBwcOXIEu91OT0/PLbtTow92Gx4exmw2ExERwdq1a9m6dSulpaVkZ2cbYvfPbCY0NJSU\nlBSeeeYZiouLqaiooKenh8uXL9Pc3ExhYeENQl26dCnLly8nLy+PrKwswzzS51YY9uxctWoVQgji\n4uI4dOgQ9fX1dHd3j1k2LCzM92jc5ORkFi9ezMqVK1m9ejVFRUWGmhSZrZhMJiIjI1mzZg3p6elY\nrVbsdjv19fVcv36dgoICoqOjfULOyclh0aJFJCQkzIn4GVbIixcvJj09ndWrV/v2Xl+/fn3MsomJ\niWzdutV3Fb///vsNfwWfjQghCA8PZ9GiReqh9AEYVsigdcfmz5/Pc889x+7duxkaGhqznMViITEx\nkaioKKKjo6e8bKVQBAtDC3l0c0hBQUGwTVEo7ihqr7VCYQCUkBUKA6CErFAYgNk2Rg4H7amKMxE/\nu4zzVLc7h4rlNCICf9xqJiOE2A28Gmw7JsAXpZSvBduImYyK5fQy24ScBDwANAKDwbVmTMKBbOCA\nlLIryLbMaFQsp5dZJWSFQjE2arJLoTAASsgKhQFQQlYoDIASskJhAJSQg4wQIk8I4RVCLA22LYpP\nhxAiTI/ljrt97AkLWTfQo78HvjxCiOfvpKETRQixSAjxnhBiQAhxXQjxwynU8Qs/v4aEELVCiL+8\nE/bqTHrp4NP4OVtiCSCE+JoQokIIMSiEsAkh/vckP/8jP7+GhRD1QogXhRAz5j5VIUSyEOJ1IUSf\nEKJLCPGvk7VvMju7/J918yTwPWApMPrslP5bGGmSUnomY9RUEUKYgfeAWqAYyAR+JoRwSSl/MImq\nJPA28AwQATwM/INez9+PcdwQQMq7tJY3DX7O+Fjqx/sO8AfAnwEfA9HAwilU9THwIBAKlACvABbg\nm7c47l31E3gDiAK26O//Bfwj8NUJ1+D/64UTfQG/C3SPkf4A4AU+A5wFhoD1wM+B1wLK/guw3+//\nEOB5oAEYQPvyH56kXbvQNhfE+aV9A2hHXzOfYD1j2XsUeF//+1mgFXgUqAHcQIqe94d6mgu4CHw1\noJ57gfN6/kngMcADLL3bfs7wWM7TfdwwlXPUr54fAWUBaT8F6vS/d47lp573GHBOj9Ul4Nv+3y9g\nBU7o+Rf8vrMdk7CvSI9/vl/aI/o5lTjReu7UGPkF4I+BfLRWYyJ8D/gC8PvAMuAl4HUhxPrRAkKI\nViHEt8apYwNwRkpp90s7ACShtTifBhfaFR20FjseTTxfBlYAPUKIrwB/gdaCWNFO5heFEL+j2x8L\nvAOcRgvgC8CPAw8UZD8DCVYsd6J95/lCiBohRJMQ4jUhRNpUnAggMJZwo581QojtwL8Bf6unPYfW\nQ/sz3f4QtFh2A2uBPwJeJGCYJIQ4KYR4aRxbNgBtUkr/TecH0XrL6ybq0J24aUIC35ZSHh1NuN2T\nC4UQUcCfAhullOf15JeFEFvQulYf6WmXgPG2y6UCbQFpbWhdxlQmfiL62yaAUmAr2tV9lFDgK1LK\nK35l/xp4Tkq5T0+6KoRYhXYC/BL4PbRW5lkp5QjaCZMD/F3AYe+6n7cgmLHMQRvW/AlaL8eJJqr3\nhBBFUsop/WaufjF5HE2Eo4zl53eBv5FS/lxPahRCfB/4DtrF9yFgAVAspezRP/M88GbAIRsA2zgm\n3RRLKaVLCOHgxiHQuNypu58+nmT5PLS9rcfEjWeKBa37CYCUcvM02DZRHhNCfE63AbTu2At++f0B\nIk4EMoA9ASe7iU8CaQXO6iIe5SQB3GU/b0ewYhmif+ZZKeUJ8N1o0YI2PDk2CZvW68Iw66+30S4Q\n/gT6WQisFkL4zzmYALPeGluB+lER65zkk3kGAKSUuydh55S5U0IO/D1TLzfPkPs/mzYa7aq4jZuv\n0pPZUG8DcgPS5ut1j3dVHIv30LrObuC61AcvfgT6GK2/P402BvZnhOllOv28HcGKZav+7utySimv\nCyH60Cb3JsN5PpmLuCbHnsjy+alfgKLQutr7AwtKKb2365lMAhta7HwIIcKBGCYRy7t1P3IHsCog\nbRXa5AxABdrJnimlPP0pjnMS+IYQIs5v/LgD7YS6PMm6+qWUDZMo3wx0AjlSyrduUaYaeCRgVnTj\nJO2C6fVzstytWJ7Q3/PQW3IhRCoQC1ydZF1Dk4mllFIKIc4BeVLKf7pFsWogRwiRKKUcfc7yRia/\nlHgSmC+EyPcbJ+9A+w4n/v1NcSZwvJlODxAakP4wMAw8gdaSvAD0cuNM54/RrsJfRBsfrUabQHjS\nr8wxtHHpreyy6F/wXrQJqM+iieuvJunfTTOzAfnPoLXSgelfB+xoY7pc3YavAF/X8+PQxPYfaF2z\nh4E6Amat75afMzmWepn/Ruvyrtf9fA8oZ3IrEDfNWk/Qz8+h9SC+gzbZlQ88BXxXzzehjfP3AcvR\nlo7O6nXt8KvnF8Dzt7HxMJqg16Atj9UB/z6pOE428FMJvp73Qz24nfqXe8OShV7mm/oJOqiX3Yff\n8gNwHfjWbWzL1k+AAb2OHwTk56F1D9ePU8eUhKznfVkPqAut9XofeNAv33/56SPGWH6aDj8NEss4\n4P+hzQy363FJ9csP02P5+Dh1TEnIel6pLrAB3YYy4Gm//Hw+WX6q9KvLX8hlwEu38TMJTfB9aBf6\nl4DwycRxzt2PLIQoBf4TWCylDBz/KWYRQoh8tBY7T0rZHGx7gslc3GtdCnxfidgQlAL/PNdFDOoJ\nIQqFIZiLLbJCYTiUkBUKA6CErFAYACVkhcIAKCErFAZACVmhMABKyAqFAVBCVigMgBKyQmEAlJAV\nCgOghKxQGAAlZIXCACghKxQGQAlZoTAASsgKhQFQQlYoDIASskJhAJSQFQoDoISsUBgAJWSFwgAo\nISsUBkAJWaEwAErICoUBUEJWKAyAErJCYQCUkBUKA6CErFAYACVkhcIAKCErFAZACVmhMABKyAqF\nAVBCVigMgBKyQmEAlJAVCgOghKxQGAAlZIXCACghKxQGQAlZoTAASsgKhQFQQlYoDIASskJhAMzB\nNmAyCCGSgAeARmAwuNaMSTiQDRyQUnYF2ZYZjYrl9DKrhIwW+FeDbcQE+CLwWrCNmOGoWE4js03I\njQB79uwhPz8/yKbcTHV1NV/60pdAt1MxLo2gYjldzDYhDwLk5+ezevXqYNsyHjOxqzjTULGcRtRk\nl0JhAJSQFQoDoISsUBiA2TZGVihuYHh4GKfTSUdHBy0tLdTV1XHmzBlcLhcxMTEsWbKE+Ph4FixY\nwKpVq4iOjsZisQTb7GlHCVkx65BSMjIygs1mo6WlhaamJlpaWmhoaKCmpoaTJ0/idDqJi4vDarWS\nmJiI1WrF4/GwYsUK5s+fH2wXph0lZMWsQ0rJwMAAR44cYf/+/XzwwQf09PQwPDyMlBKv14sQAofD\nQXl5OQAXL16kp6eHZ599VgnZKNjtdpqamjh+/DjV1dXYbDaklAghiImJIS8vj3vuuQer1Up0dDQt\nLS0AZGZmYjabCQlRUwvBpKWlhVOnTvHGG29w/vx5enp6cLvdSCkJCwtj4cKFdHd3093djZQSgM7O\nTo4cOcKuXbuCbP2dYc4IWUqJx+PBbrdTUVHBsWPHOHLkyA1CBoiNjSU3NxeXy0V3dzcmk4nq6moi\nIyPZvn076enpREdHB9mbucloS1xdXc2+ffsoLy+nvb3dl5+amkpeXh7r16/n8uXLnD17lpaWFjwe\nD4ODgzQ1NTEwMBBED+4cc0bIXq8Xl8tFTU0Nb731Fq+//jq9vb14vV5CQ0N9LfLQ0BAVFRVERkZS\nXV1NW1sb1dXVZGdnEx8fT0lJiRJyEBi9ENtsNk6fPs3+/ftxOBwACCEwm82sXr2axx9/nAcffJBT\np07xq1/9ir179+JwOPB6vYA2OTY8PGy4Ca85I+S+vj5qa2t56aWXOHXqFA6Hg6ioKIqLi1m/fj0W\niwUhBMPDw9jtdhITE+nq6uL06dO4XC4iIiJISEgw3AkwWxgYGMBms7F//36OHz+Ow+HA4/EQGRnJ\nwoULeeKJJ3zDoZiYGNasWUN8fDzr1q3jzTff5MMPP8TpdPKzn/2MoaEhvva1rwXbpWnF8EKWUtLf\n309FRQX79u3j+PHj9Pb2+pYjtm3bRnFxMWaz2dcid3Z2Ul1dTVVVFV1dXaSmplJQUMCiRYuIjIwM\ntktzkoGBAerr6zlx4gSVlZW43W5CQkJITU2luLiY0tJSn4gBUlJSiI2NZcGCBdhsNtra2qipqeHM\nmTPExsZitVopKCggKSkpyJ5ND4YXstfr5dq1axw+fJhXXnkFl8tFVlYW99xzD0888QSFhYU3zGIO\nDQ3R0dHBoUOH+PDDD+nv72fbtm2UlJRgtVqD6MncxuFwcOXKFaqqqrh27RpCCEJDQ8nNzWXbtm0s\nXryY2NhYX3khBJGRkWRlZbF+/XqamppoaGigt7eXM2fO8NOf/pTnnntOCXm24PF4OHbsGKdOncLl\ncpGbm8v27dt56KGHyM/PJyEh4YbyQ0NDXLt2jcbGRt8kWHZ2NosWLQqSBwqAnp4ezp8/T19fn29M\nnJOTw6ZNm9i+ffsNIg5kw4YNOJ1OPvroI1paWujv7+fKlSuGmvgytJBbW1s5c+YMhw4doqamhvDw\ncO677z7uv/9+CgsLiY2NxWz+5Cvwer309PRw/Phx6uvrkVKSkZFBXl4emZmZQfRkbuN2u+no6KC6\nutonPovFwvLly1m2bBmpqanjfj45OZmlS5eydu1anE4ndrud/v5+RkZG7ob5dwVDL4g2NDTw6quv\ncvz4cVpbW4mNjWXLli2sW7eOxMREn4illL6ljYaGBg4ePEhTUxOxsbEUFRVhtVpve7Io7hz9/f20\ntbXR0NCA0+nEZDIRExNDUVHRhHtKo7FMTEz0bRrxeDy+2ezZjqGF7HA4qK+vx+VyYTKZiIyMJDU1\nlfj4+JvKejwePv74Y9555x0uXLhAb28v6enpPPbYY2RlZQXBesUobW1tNDU10dHRgdvtJjExkWXL\nllFUVMTChQsnXZ/H42FgYAC73W6Y7rWhu9YejweXy4XH40EIQUhICKGhoTcsIQ0ODtLV1cXly5f5\nzW9+w+HDh30bQVJSUigsLBxT+Iq7h81mo7W1leHhYUDbYffAAw+wePHiKa3pu91uuru7sdls9Pb2\n+ma6ZzOGFrLJZCIsLIyQkBCklLjdbpqbm4mLi/OVsdvtNDY2cvToUY4ePUpVVRUAiYmJpKenk5GR\noZacgkxXV9cN2y2zsrLYsWMHqampU1rXHx4epre3l87OTvr6+qbb3KBgaCFHRkaSkZFBc3Mz/f39\n2Gw2fvKTn9zQwvb399Pd3U1ra6tvpxDAggULWLJkCdHR0TdMiCnuPqO7sUZJSEjAarUSGho66bpG\n50OMhqHP0OzsbJ566im8Xi/l5eV0dHRQVVV1w1V8eHiYqKgotm7dSm9vL/X19Vy9epW8vDxWrlzp\n2yiiCB6dnZ10dX3yRFqz2Ux4ePiU6jJqLA0t5PT0dEpLS+nt7SUiIoLKykoGBwd9t7mFhoYSFRVF\nRkYGJSUl1NbW0tPTQ0dHB7m5uVitVnWn0wygt7cXu90+bSIMCQnBZDJhNpsxmUzTUmewMbSQQ0JC\niIqKYvfu3dx7771cvHiRxsZGBgYGsFgsZGRkYLVaiYqK8u3HvXTpEklJSaSlpZGQkGDYK/hcxmQy\nERERQXR09JRb9pmGoYUshEAIQXR0NNnZ2cTFxbFs2TKGh4cxmUxERUWRlJREa2srJ0+epL6+npCQ\nELKyspg/fz5RUVHBdkExTdjtdj766CM6OztJS0tj8+bNrFixQm3RnG1ERUURFRXFggULbkgfvdvp\n0qVL9PT0kJmZSVFREenp6YSFhQXJWoU/6enppKenU1tby8jICN3d3dTU1JCZmTnhFQWHw8HFixcZ\nGBhg2bJlfP7zn2fp0qWGWHoCg28ImQgDAwN0d3fj8XiwWCykpaXxwAMPqE0gM4iSkhK2bt3q21J7\n8eJFXn75ZVpbWydcx8jICA6HA4vFQk5ODjt27CAlJeUOWn13mfNCrq6u5v3338fpdJKRkUFhYSG5\nubnjbsJX3F3S0tLYsGEDjz/+OGlpaVy/fp2ysjJqa2vp7u6+7eePHz/Ovn376OrqYt68eWRlZRER\nEWGYiS6Yw0L2er309/dTWVlJWVkZTqeTBQsWsGzZMubNm2eYSRAjEB0dzeLFi9m+fTtWqxWz2Uxd\nXR2XL1+mo6Pjlp9zu910dnZSVlbGiRMn8Hg85Ofns3z5ckwmk6EmMueskIeHh2ltbaWqqoqKigqG\nhobIyMggNzdXbQCZgcTFxZGfn09xcTE5OTk4nU7q6up8t5qObvIY/Xv0JpiqqirKy8upq6sjNjaW\nkpISNm7cGGRvpp85e8aOjIxw7do12tvbGRzUfqdrdG1RMfMIDw8nPT2dnTt34nQ6aWpq4oMPPsDt\ndtPT08PmzZtJSEjwTYbZbDYqKip46623OHv2LFFRUWzYsIH8/HzDzFT7M2fPWo/HQ1dXl29bZmxs\nrG921EhjJ6NgNpuJjo4mPz+frVu3+h40cOHCBVwuF3a7ndzcXCwWC2fPnqWxsZGamhqOHz+OyWSi\noKCA+++/n+zsbCIiIoLtzrQzp4Vst9txuVy+ZyEvX758Rv5Wr0JDCEFcXBybN29myZIlvPLKKxw7\ndozf/va3nD17ljVr1pCWlsbevXtpbW1lcHAQs9lMYWEhmzZtYuvWrYZ8OD3MYSEPDg5SXl5OY2Mj\n0dHRFBcX37TGrJiZhIWFkZaWxlNPPcXGjRu5fPkyJ0+epKqqigsXLrB8+XJWrlxJTEwM+fn5WK1W\ncnJyyMhNc2K3AAAHmElEQVTIMOzegDkr5NHbGj0eD2azmXnz5qmdXLOE0V15eXl5ZGRksGTJElJT\nU6mrq2NwcJDCwkLCw8N9O/pSUlKIjo727fQzInNWyGazmfT0dBITE+nu7sZisaix8SxCCEFYWBhh\nYWEkJSVRUFAQbJOCypwVclhYGPn5+Vy9epXW1lZyc3PVk0AUs5Y5K+Tw8HBWrVpFVFQUfX19FBUV\nGXJZQjE3mG1CDgdtW+V0MTrBNTQ0RF1d3aeqy88utS3s9kx7LKeT2RZLMZseeyKE2A28Gmw7JsAX\npZSvBduImYyK5fQy24ScBDwANAKDwbVmTMKBbOCAlLLrNmXnNCqW08usErJCoRibOXvThEJhJJSQ\nFQoDoISsUBgAJWSFwgAoIQcZIUSYEMIrhNgRbFsUnw4hRJ4ey6V3+9gTFrJuoEd/D3x5hBDP30lD\nJ4oQYqcQ4pQQwiGEaBFCfH8KdfzIz69hIUS9EOJFIcSMuZFVCJEshHhdCNEnhOgSQvzrRO2bRbFc\nJIR4TwgxIIS4LoT44RTq+IWfX0NCiFohxF/eCXt1Jr0MNB1+TmZnl/8PBD8JfA9YCozeTtJ/CyNN\nUkrPZA2bCkKIdcA7wF8Bu4FM4P8KIaSUcrIn58fAg0AoUAK8AliAb97i2HfNT503gChgi/7+X8A/\nAl+dwGdnQyzNwHtALVCMFsufCSFcUsofTKIqCbwNPANEAA8D/6DX8/djHDcEkPIurctOm5/+zzia\n6Av4XaB7jPQHAC/wGeAsMASsB34OvBZQ9l+A/X7/hwDPAw3AAJqQHp6kXf8HOBqQ9hhgB8ImUc+P\ngLKAtJ8CdfrfO8fy0+945wAXcAn4Nvp6vZ5vBU7o+Rf8vrMdk7CvCPAA+X5pjwBuINEgsdyFtlEk\nzi/tG0C7//c5gXrGsvco8L7+97NAK/AoUKN/hyl63h/qaS7gIvDVgHruBc7r+Sf12HuApXfbzzs1\nRn4B+GMgH+1KMxG+B3wB+H1gGfAS8LoQYv1oASFEqxDiW+PUEcbNu4QGgWhg5QTtuBUutNYZPuk+\n+ftZI4TYDvwb8Ld62nNoLcGf6faHoPUYuoG1wB8BLxLQHRNCnBRCvDSOLRuANiml/0blg2g9rHVT\n9O9WBCuWG4AzUkq7X9oBIAmt9/BpCIxlPJp4vgysAHqEEF8B/gItdla0C9OLQojf0e2PRYvlabQL\n6wvAjwMPdLf8vBM3TUjg21LKo6MJt7uZWwgRBfwpsFFKeV5PflkIsQX4A+AjPe0SMN52uQPAHwgh\nHkXrTmWgdbMB0ibnxg32rQceRwvcKGP5+V3gb6SUP9eTGvUx+nfQgvwQsAAollL26J95Hngz4JAN\ngG0ck1KBNv8EKaVLCOHgxm7zpyWYsbzJR/1/oedN9KLib5sASoGtaL2uUUKBr0gpr/iV/WvgOSnl\nPj3pqhBiFdqF+ZfA76E1Es9KKUfQLuQ5wN8FHPau+Hmn7n76eJLl89D2th4TN54pFrQuCwBSys3j\nVSKlfFcI8b/QxrOvo115X0DrEnonadN6XRhm/fU28CcBZQL9LARWCyH8xzYmwKy3xlagflTEOif5\nZGw66sfuSdp6JwlKLKeZx4QQn9NtAG2Y9IJffn+AiBPRGoE9ARcuE59cYK3AWV3Eo5wkgLvl550S\n8kDA/15uniH3/6n5aLSr/zZuvnpNakO9lPJFtC5QKloXtgD4IVA/mXrQxj6jY55rcuxJHp+f+kkb\nhdYN3T+GXd5pfMyMDbjhKXJCiHAghvFb8qkQrFjagNyAtPl63ZP18T20rrMbuC71gagfgT5G6+9P\no50H/owwvUyLn3frfuQOYFVA2iq0AT1ABdoXlCmlPD0dB5RS2sB3u1ydlPLiJKsYklI2TOJ4Ughx\nDsiTUv7TLYpVAzlCiEQp5ehvnWxk8ksWJ4H5Qoh8v3HyDrTvcFq+v3G4W7E8CXxDCBHnN37cgXZx\nuDzJuvonE0ugGegEcqSUb92iTDXwSMBM/lSefD89fk5mJnGCM50eIDQg/WFgGHgC7erzAtDLjTOd\nP0abPfwikAOsRpsMetKvzDG0scyt7DKjtYjL9NffoLUCE54R1uu5adZ6gn5+Tj/ed9Amh/KBp4Dv\n6vkmtDHTPmA52tLRWb2uHX71/AJ4/jY2HtZPgjVoy2N1wL8bKJYWNLHsRZuA+iyauP5qkv7dNGsd\nkP8MWisdmP51tNWOP9T9XAF8Bfi6nh+HJrb/QOtmP6zH4IZZ67vm52QDP5Xg63k/1IPbqQvlhiUL\nvcw3dacG9bL7gA1++deBb41jlwk4AvSgdZeOAVsDyoShdQ8fH6eeKQlZzyvVBTaA1rUvA572y8/n\nk+WnSr+6/IVcBrx0mxgkoQm+Tz+hXgLCjRJLvUw28N/6d9kK/CAgP0+P5fpx6piSkPW8L6NdaF1o\nPZH3gQf98v2Xnz5ijOWn6fBzIq85dz+yECIfbQInT0rZHGx7FFNHCFEK/CewWEoZOM6dU8zFvdal\nwD8rERuCUuD7c13EoJ4QolAYgrnYIisUhkMJWaEwAErICoUBUEJWKAyAErJCYQCUkBUKA6CErFAY\nACVkhcIAKCErFAZACVmhMAD/H5kIHmv7M1e3AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f3fc1119e80>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_example_predictions(g.value)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAALgAAAC4CAYAAABQMybHAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAAPYQAAD2EBqD+naQAAIABJREFUeJzsvWmMXPl57vc7tZ46te9Ld1f1vg3JJmfI4WzUiJKssWFP\noMiGbNiAIiSwkXzJDbIBCYILXCC5QIJ8SAIjCXAd4ULGVZQP/mBbkSwbY0kjaShxJHKmm+xm7129\n1L5XndqrTj5wzv/OyDcy2TXSyM16gEYPa7qrz/+c57znfZ/3ef8laZrGGGNcVBg+6QMYY4xfJsYE\nH+NCY0zwMS40xgQf40JjTPAxLjTGBB/jQmNM8DEuNMYEH+NCw3TeX5QkyQ+8ARwB7Y/rgD5GyMA0\n8B1N04qf8LH8WuMiX8tzE5zHJ+TfjPD7vyr8EfD1T/ogfs1xYa/lKAQ/Avj93/99jEYjDoeDdvvx\nze/z+ZBlGUmSaLfblEoler0eFouFwWCA0WjEZDJxdnaGJEk4nU4sFgtTU1N4vV729/d5+PAhg8GA\n3/md32EwGNButzEYDBgMBgqFAtvb21y+fBmfz8dwOKRcLtNqtRgOh5hMJsrlMt/4xjfEcY7xC3EE\n8OUvf5kbN27Q7XYJhUK4XC663S6Hh4ecnp5SKpU4PDwkFArx6U9/mvfff592u43T6SSVSuFwOAgG\ngxSLRdxuNw6Hg7OzM2RZRlEUut0uqqoC4PF4WF9fB+D69evMzMxQr9d5+PAhdrudXq+Hqqp4vV4y\nmQx/+Zd/KY7zaTAKwdsAuVyOaDRKOBwmnU7T7/eZnp7G5XJhMBjodrt0u12q1SqFQoHJyUkkSaJU\nKjEYDMRiA4EAk5OTGI1G3nvvPV5//XVmZ2dpNBq0Wi00TaPf72MymZidneXKlSv0+32KxSKpVAqb\nzYbZbKbX6xGPxzGbzR85zjF+IdoAoVCIubk5wuEwiqLQaDS4f/8+Ho+HSCSC0+nk5OSE4+Nj7ty5\ng9frZWFhgYmJCX784x9jMpmYnp7m5ZdfplarUSqViMViDAYDOp0OnU6HUqmEx+Ph9u3bvP7665RK\nJer1OtevX6dSqVAqlYhGo5hMJhqNBicnJ3S73Y8c59NgFIID4Ha7sdlsHB8fI8syPp8PTdPY3Nyk\n0+kQi8Ww2+1YrVYURSEYDNJut8lmswSDQVRVpVQqUSwWMRqNxONxZmdnGQwGZLNZpqamqFQqGI1G\nbDYbpVKJTqdDIBAgn8/TarUwmUxYrVbxxDg8PKTf74+6tGcOnU6HVqtFOp1GlmVarRaqqtLv95Ek\niXg8zuHhIZIkiWBkMpno9Xrk83k8Hg9er5f5+XnOzs6oVqsiElerVWq1GisrKwSDQTY2NrDZbPT7\nfRqNBjs7O7TbbSRJQlVVTCYTkiSxsLDAcDg895pGVlFcLhdGo5FcLofRaMTj8dDpdDg9PeXk5IRW\nq0Wn08FkMhEIBOj3+3S7XaxWK16vF5fLhaZpHBwccHZ2hqZpzM/PYzQaKZVKuFwuQqEQbrebTqdD\nr9ej1+vRarUolUp0u108Hs/jxRgMWK1W6vW6SJfGeHIMh0NqtRrJZJLj42NKpRJGo5HBYECz2aRc\nLlOtVpEkiUgkgslkQlVVyuUysiwjyzKapmE2m7FYLBgMBkHuUqnE2dkZDocDt9vN+vo6yWSSRqOB\n1Wolk8mQy+WQJIlGo0E6nSaVSjEzM0M0Gj33mkaO4P1+n0qlIlKIZrNJv9/H6XTi8/lwu910u11s\nNhuxWIzvf//7mEwmVlZWKJfLeL1eZFlmc3NT5NF+v5/JyUlUVaXX6xGLxej3+3z9619naWmJWCwm\nTpjVaiUUCrG9vY3JZCIcDuPz+Ugmk6Mu7ZmDLMtUKhURjd1uN7Isi6Dx9ttvY7PZsFqtFItFisUi\nw+EQj8fDG2+8Qa1WY39/H1mWabfbdLtdWq0W3W6XTqdDuVwmmUzSbrdpNptUKhVcLhdLS0uUSiVK\npRKqqtLpdNjb2+P09JTPf/7zmEznp+nIBHc6nbhcLlE8lEol+v0+/X4fm82GyWQimUySSqVotVos\nLi7S7/fJZDL0ej3C4TCJRAJVVZEkiUwmQyqVwm63Yzab2dvbY3t7m06nw5UrV7Db7QyHQ1qtFoqi\nkE6n2djYIJFI4HQ6RYGj5/djPDkKhQImkwmDwUCxWGQwGHD16lXq9ToGgwGz2YzVaqXVapHL5Vhb\nW6PRaHB6eipqreFw+JEndbFYFALDxMQEsizjdrv5zGc+w97eHpubm9y7d4+5uTmCwSBTU1Pkcjlm\nZ2eJx+NChDgvRk5RzGazILLB8PjtNE3DbreLiKAXfJlMhsXFRZaWlrBYLCiKgsFgoN/v4/F4GAwG\n5PN5ut0u9XqdQqFANptlZ2eHo6MjLBYLAIPBAIPBgMfjQVEUOp0ODocDg8FAuVzGaDQiSdKoS3vm\n0O/3xXVrNpvU63URUADC4TCyLANgNBrxer243W5MJpNITRwOBzabDbfbjcfjwWg0omkasiyzurqK\n0+lE0zT8fr9IR0qlErVajXq9jqqq4mk/PT1NLpej1+ude00jR/B2u83Z2RknJyfMzs4SjUZpt9v4\n/X7a7Tabm5viMXPv3j1isZg4KY1Gg8PDQ+7fv08qlaLb7RIOh7l9+zbvvvsuGxsbzMzMYDQahTQ4\nPz9PIpHA5/MxGAwIBoO88sortFotkskknU4Hj8cjLsQYT45AIMDLL7/M1tYWlUqFbrdLpVIhnU7T\nbDax2WykUikajQZms5mDgwOGwyFut1t8KYpCs9lElmUcDgcOh4NqtYosy7z44ouk02mOj4/Z29uj\nUqkwPT3N8vIyJycnHBwckMlkuH37NjabjbOzM05PT0da08gENxgMhMNhIpGISCssFguaptFoNCiX\ny/zwhz8kkUhw9epV1tfXxUkDRL4+NzdHLpejXC7z7W9/G4PBwPT0NJFIBE3TUBSFXC6HpmmEQiE+\n97nPsb6+Tj6fp9PpUCwWkWWZ5eVlyuXyuMg8B+r1Otvb25jNZiKRCO12m+3tbQaDAQ6HA7PZLArD\n2dlZkdI4nU7u3LlDNBolHo+LekgvUBOJBKFQCI/Hw+npKb1eT+T3kiRRq9WwWq1Eo1GcTietVkuo\nY4qifLIR3Gw2i4MF6PV6dDodNE3DarUyMzNDr9fDYDCQSCT4yU9+ws7ODqVSiampKRwOB4qiYDQa\nqdfrVCoVGo0GDocDq9WKyWSi2+2iKAqXLl0SqUitVhOFTKfTIZPJYDQaiUQiWCyWkQqTZxXtdhtV\nVQkGg0L+lSRJnMtMJoPVahVp6XA4FFLhcDik2+2K99Cv+YdT0VwuJ6L7xMQEp6enQpSQJAm73Y7D\n4cBisQhN3OFwjFRPjcwCi8WC0Wik3++LO7ZUKuF0OgmHw1y/fp1sNovD4cDlctFut8nn8xQKBdbW\n1nC5XNTrdQaDAcPhEJvNxtraGrlcjmKxiNVqJZ1OoygKb775Jul0mkwmwze+8Q1sNpuQpjKZjLjz\nb9++TaVSGXVpzyS8Xi9WqxUAu91OPB5HVVWSySQbGxvcuHEDj8dDuVxmYmKCTqdDvV5nbW1NBBq3\n2y0aOFNTU9TrdTKZjMjx9fz64cOHFAoFLl26RLVaBR6LFsFgkGq1yv7+PvF4HKPReO71jExwXVKa\nnZ2lXq+LR0u9XqdUKvH++++jKIp4TAUCAWZnZ7FarTSbTUwmE7IsY7FYaDQaDAYDWq2WyKPL5TK9\nXo9qtcrf//3fMzExIXRvXYPXG0omk4mpqSm63S52u33UpT1zCIfDOBwOstkssizT7Xb54Q9/SDAY\nxOv18sYbb2AymWg2m0JJ0dMHvTnkcDioVCr0+32huOiRvFwu89JLLxEMBvnud7/LD37wAxwOB5/7\n3Oc4OTkRN4BeAyQSCWRZ/nBX+qkxMsFbrRbVapXj42PMZjNmsxlZljEajXS7XdGSz+VyvPvuu9Rq\nNfr9Pi6Xi36/T6vVEgRutVoYjUba7TayLIsTYzKZaLfbFAoFQqGQ6IzqkabdbhOPx0XOpj/2xng6\ndDod0uk0e3t7KIqC0+nEarUyGAxQVVUEr1arRT6fZzgcCp9Ju92m1+vR7XaFN0l/wqbTafL5vFDG\nNE2jXC4zGAwEZwwGg2jN61G82+1SKBQ+WYLr+XAymWRxcZFYLMZwOBSSUqVSoVAokE6nef/99wFQ\nFAWPx4MkSXS7XWq1GplMBkmScDgc9Ho9Ya4Kh8Mir5dlmcFgQLfbRdM0ut0ug8FA6Ol600l/zzGe\nDoVCgWazyenpKSaTiUQiwe3bt2m1WhwdHXHv3j0RSCqVCpOTkwQCAbxeL2azWQQpn88nmnwWi4W9\nvT2Oj4+xWq28++67eDwenE4nCwsL2O12VFXFYrHQbrfJZDJ84QtfwOl0cnp6Kgx558XIBNfv1GvX\nrolCw2w2k8/nRWTXF+vxeGi321SrVcrlMqFQCKfTyXA4xOl0ivfSO1pWq5WJiQny+Ty1Wo1CoUCp\nVMLv9xOJREilUrhcLhKJBLu7u7TbbaxWK51OZ0zwc2BlZYVarSZSFU3TePvtt4VS8hu/8Rsi9dTb\n+roKUqlU6HQ6GAwGWq0Wh4eH1Ot1qtUqdrudK1euiJsG4P79+yII7u3tsb6+jsVi4aWXXsLlcjEY\nDJAkifn5+ZGkwpEJLkmSMM3oon6pVEKSJCHoO51OUYgqiiJOmMViwWq1YjQaRd5nNpspFAr0ej1c\nLhfVahWTyYTP5xOpTKlUwu12EwwG8fl8WK1WDAaDyOf1x+kYTwddtfD7/eJ6hcNhQV49EEmSJHJv\nXUnRzXA2m01YZmu1GlNTU8II1+v1yGQyyLLMlStXhKnK6/VSqVRQVZVmsynIbTQaRepyXnwsXhRF\nUUgmk8KAUygU8Hg8hEIhVldXaTQaZLNZstms6H7ZbDZUVRUdsFQqJbzihUIBg8EgTFwWi4VEIoHN\nZmNzc5N6vU6tVuPy5cu4XC4KhQJut1u8l24bGOPpsL+/z4svvojRaKRWq2GxWLhx4wYPHz4knU6T\nzWaFWCBJEp1OB4vFgs1mE7m3x+MhkUhgtVqRZZkbN25QKBQ4PDykVCqRTCYJBAJ85StfYXt7G6PR\nyPXr14VNemdnh1gshsvlQpIkWq3WSD2NkQmuaRqBQABVVTEYDMiyTCQSwWq14na7mZycZG9vj1qt\nxsnJidC19bt3YmICu91OIpEQxcwLL7xALpdjMBiwsLBAOp1mOBxy6dIlWq0W29vbHB4eYjQa8fl8\nIoLo1szl5eWxF+UcmJubIxQK0Wg0yOfzwl+vS3uRSITBYEC/38fv9xOLxcjn87z//vs0m01h0AIE\nQSuVCoFAAJvNJuYBhsMhX/va11heXmZyclK4F/VC9MGDBwQCAfE0WV5ePveaRia47gfWczJdHnI4\nHPT7fTY2NoS4Pzc3hyzLIpXQrbSNRgObzUY2m6VUKrGysiLSHUmSxOBEoVDA4XAwOTmJ1WoVyoos\ny9jtdnGCm83mqMt6JhEIBJBlmUwmg9/vx+v1MhgMMJlM2O127HY729vbqKqK3+8XTT7d16+rXY1G\nQ4gAd+/eJRgMiubN3NwciqKwubkpRIF0Ok0ul6NarQr/udFoRFEUkaaeFyMT3Gg0Uiw+ngNttVqC\n4HqkPjg4ELlZIpGgVquJE5bP59E0jU6nI6K3PvGhd8vq9TrwONfP5XK43W6cTidut5ujoyMqlYqw\ndLrdbiYmJrh79y6dTmfUpT1zsFgsmM1mKpUK8XicSCQicm+DwSAaao1GA7vdTq1Ww2azMTc3RyqV\notfrifTGbDbT6XR49OgRp6enQh93Op1EIhFhqstms1QqFVqtFmazGafTSSgUIhgM4nA4ROv+vPhY\nUpSzszMmJiZwuVwMh0PS6bQwxX/mM5/h0aNHJJNJCoUC4XCY4XDI/v6+GHEKBoNks1kREXZ2dnj+\n+eeZn58nl8tht9uJRCK43W58Pp9o6WcyGbLZLIqiMDU1RafTYXd3V+TlYzwdSqUSy8vLVKtVHj58\nSC6XI5FICP17MBjg8XjweDxMTU2JyOr3+7lz5w6hUIgrV66wtbWF0+nE7/eLiRxN01BVlUwmI2qo\ndrvNcDhEVVWRu3s8HsER/UnwiRLcbrfj9XrpdDpCkNf1UF2k151o169f5/T0lE6ng91uF/laOp0W\n0ziapuHxeIQeq3c1+/0+P/3pTwkEAlitVmw2G1NTU0KlMRgMnJycUKvVWF5eHslD/KyiWq2Sz+cx\nGAwoioLFYqFarTI3N4fJZGJnZweXy0U4HGZ2dpb79+9zenpKtVrFYrHQbDY5ODggHA4zGAwol8tY\nrVYajQbdbhe3243BYMBut3Pjxg329vYolUrYbDZcLhfRaJTJyUkAstksuVyOqampkcxWI/vB9Ska\n3SSly0Z6t+vo6IhSqYTZbGZ6elrkZrIs43Q6aTab7O/vi5au0+lkZmZGLFIfbshkMmxsbPDee+9x\ndHSEwWDA5/MxPT3N6uqquBj5fF7kb2M8HWq1GoeHh6KeURSF4XCIoihC9dILP5vNJmZrj4+Psdls\n9Ho9Tk5O8Pv9wOPrp3/Agn69vF4vwWCQubk5UXzq3dJer4emaTSbTTHDORgMxHucByNHcN31tbu7\nS6vVIpFIsLi4KIYQ5ufn+dGPfkQ2m+V73/se9XqdTqdDKpVCVVXq9bqwwU5PT7O0tEQ0GqVcLlOv\n10XKonteLBYLvV6Per1OMplElmXx2NTzRVmWR7rrn1UUi0VarRZra2v0ej2azSbxeJx79+5hMpl4\n8cUXKZfLnJ2dkc/nP7IdSL/fF0Gr3+9TLpcpFAoEAgH8fr9o6U9MTGCz2Tg6OiIYDGIymdje3ubk\n5ISTkxNyuZx4kutNQH3Q5TwYmeChUIhwOEwsFiMYDBIIBHA6ndRqNVqtFo8ePaJYLIppDX3Gcmlp\niVwuB4CqquTzeRqNBo1GQ+RpetHa7/dxOBxEo1EymYwoTufm5oS9MpPJAI+7cfqeHGM8HfTtOcLh\nMLu7u1SrVaampggGgxgMBhFZDQYDLpeLGzduUC6XSaVSRKNR0cTTg04ikSAQCAiPSrfbZXd3F7vd\nzuzsrFBkVldXRa6tqzH6VNeonqKPJQe32WxMTk7i9XpRFAVJkuj3+6LrWK/XxRR9s9kU0x+6DChJ\nkhhDK5fLItXRx+DcbjfD4fAj09rBYBCn0ykq8WKxSCAQIBAIjDXwc0L3futat25hNhgMDAYDMWSs\nEzAWi4kdDuLxOL1ej9PTU9rtthhk0VPMWq0mBoqDwSDLy8t0Oh3h/IxEIjSbTSwWC06nU9RxuiHv\nvBiZ4LpHJBAIiIJSVVWKxSLNZlO0cCcnJ3nllVc4Oztjb2+Pv/3bv+XWrVskEgkURSEcDlMul4XE\n5PV6iUQixONxKpUK2WyWdDpNNBplaWmJtbU1vvOd73D37l2Ojo5YW1tDlmWq1SqdTkc8HcZ4cvh8\nPux2O5ubm9jtdubm5jAYDGxubtJoNJienmZhYYF+v8/Ozg6NRoOzszO2t7f5rd/6Lfb29vjzP/9z\nEomEaPbpTTrdqbiwsCB6HlNTU5TLZe7fv8/09LTYa0UfqtDT2VEUsY+l0WO32zk+Psbv92O32+l2\nu3i9XtHsWVhYoNPp8K1vfQufz4ckSbhcLjHQMBwORb4djUbJ5XJkMhmq1arYEuLs7IxQKIQsyxSL\nRb797W9jNBqZnp4WQ8v6Cfd4PGIAeownx6VLl2i32xwdHbG0tMTMzAyPHj0SI2a657/T6eDz+Wi3\n23g8HrGTVTAY5Etf+hJGo5GTkxMymQyTk5OiodNqtQQnDg4OhOd/cXGRer0utgF54YUXCAQCSJLE\nN7/5Tcrl8rnXNDLB9SLi+PiYer0ucmi96AgEAoL0xWJRpDDwuMr2+Xy4XC5sNhuhUAifzyfyr+Fw\nKMbSADFCpf+NyclJbDYbTqfzHxSW4yLz6dHr9UQXWlEUEUnD4bCYtj87O2MwGBCNRkV7PhQKiU2Z\nfD4fiqJQqVQol8vCxqz79/UAVy6XBQ/sdjutVot6vY7RaKRcLgvPPzy+7ufFyASfn58nFAqRTCbZ\n29vj4OCA7e1tGo0G8Xic3/7t3yaZTGIymXjhhRdENyybzQpZMBaLCROWxWIR0V/P9fRJet1Ur0/u\n1Go1UfkPh0OCwSCxWIxmsymKzjGeHO+88w5Xr17lypUrYgB5ampKTPdkMhkx3K1/6d3GYrHI9vY2\nu7u73Lp1SygnesSWZVnk6c1mE6vVSrfbJZ/Pk0qlRKe0Xq/zzjvvCIKvrKygqirf/va3z7WmkQn+\n1ltv4fV6Rf5sMpmYnJzE4XDg8/mEb0HXpUulEgDRaJRisSg6mPqjTN8NSVVVjEYjbrebdDpNoVCg\n3+/z2c9+FoPBwFtvvSVSIkVR6Pf7ol2sj7KN8XTQdw0DxNCDblzTI6we4SVJolgssrW1RaFQEL83\nMzNDIBAQvvHd3V1hfQWYnZ3F6XQKD5HZbBb/z2Kx4HA4ROdUb+Jls9lzr2lkgudyOTweD9FoVOS9\n+iT0h01V3W5XVOMul0tswqm3aiuVCiaTSYyo6QPEuqZarVaFlORwOPB4PLhcLux2O0ajUeycpXtY\n9LnNMZ4cHo9HbL4Jjx2B5XKZfr8v3KJ62qCrHPV6nXw+LzZ/MplMYmwR4OTkRPQmZFkmFovhcDjE\nLle6aqM7QfVBFX0nLa/XO9LwysgEd7lcXLt2jVAoxMOHD9nb26PT6bCxsYHFYhHehHq9jtPpZGJi\ngmg0KiLzYDBAURRSqRSapgk9Vd/zu9PpCM+5JEns7+8zMTHBrVu3xJa8rVaLRqOBJEkEg0GCweA4\ngp8DmqZxfHzM0dERt2/fZmFhge3tbQwGAzabDYfDgaqqYipLt2nopNYHI9rttghW+n6TdrudUChE\noVDAbrdz+fJlBoOBGG5QVZXDw0PS6TSBQACDwUCv1xu5I/2xzGRub29zfHwsiolUKiV0VH3jGKPR\nSLVa5cqVK2KgVHcPlkol7HY79+7do9PpMDk5yeTkJAsLC0IXbzQaYs9pTdOEFl6tVvnZz36G2WwW\nBczBwYFIhcZ4ciiKwvz8PCsrK8Iq6/f7xRZ6RqORWCxGq9XiwYMHOBwOAoEAoVBITGkNh0Pxs/C4\nERiLxfB4PJycnAjnqcViEYJBLpcTs736Jk9ms5lWqyV2TTsvRiZ4o9EQvgF9pyKz2czs7CyaplGr\n1YQ/Rd+ks16voyiK+JlisUg8HqfdbpNOpzGZTLhcLjGI2uv1MJvNBINBsbOSLMvU63WazSZ2u51w\nOCxG4/RJ/TGeDiaTifn5eaFbd7tdLBYL9XqdcrlMt9tlZmZGDBXrk/f6E1efC9DHzgCuXbsmvCmq\nqnJ8fEy322Vubg54PMlfrVbFBp+hUEjsjxONRsVA8rnXNOpJ0aUdvRUrSRJer5d4PM5gMOC9994T\nEdrv9wuZaXV1lRs3biDLMvfv32diYkJo4vqOpbp2qk/bB4NBDg4OxJ7Th4eHuN1ulpaWxGaO+rDD\neGzt6aHvM+hwOIRooCtdenTVPSWXLl0SY4V69NWbbNFoVIywXbt2TQQuXS7u9Xpi5G0wGGC320km\nkzgcDpaWlsTk1/LyMsfHx4RCoXOvaWSCR6NR/H4/2WwWm82G0WgUqYf+uNG9wO+//774SItWq0U4\nHGZ+fp5ut8v+/j4ej4cbN26Iqlmf8dP33Xjw4AGJREJMfIdCIfEZMgcHBwSDQSYnJ0XhOcbTQfcO\nvfjii8IfpCiKMM3pW+wVi0VOT0/J5XKEQiEWFxdJJpOYzWYmJiaEF9xgMLCzsyO6kVtbW+KjUO7c\nucPBwQFOp5NPfepTQnHb2dkRrsWNjQ3Ozs4+2ZlMfaJdvyN18V4v+vQJjtPTU+7cucPS0pLwFtdq\nNRRFERvW60rKh/cr1JtDmqZRrVaFvqp3Q/UiVH+MFQoFlpaWxlP158BgMCCVSrGxsUE6nRZebf0J\nGgqFKJVKVKtVut0ukUiEaDSKx+NhcXFRFIWRSARAKCy6H3x2dla04/UtPgKBAM899xy7u7scHR2R\ny+VYXFzEaDSSTCZpNpuf2OabMiAaNvqidXOOnq64XK6PFB+dTodsNsujR4+wWq1MTk4SDAYplUo0\nGg2SySTRaJRQKCRavo1GQ3yAkcFgEIrMhz+cyuFwUCgUyOfzorP24eMc4xdCXMtiscijR4+EMUpX\nSPRJqnK5LAx0KysrWCwWisUiTqeTcrksUpFWqyVmbPWBiMuXLwtZWLfD2u124R06PDyk1+sxPT1N\nuVxmb29PbPH34eN8Kmiadq4v4A8B7Z/A1x+ed43PytdFvpbSeaclLvKn4z5ruMjX8twEH2OMfwoY\ne0rHuNAYE3yMC40xwce40BgTfIwLjTHBx7jQeGKCS5I0lCRp8MH3n/8aSJL0z3+ZB/qkkCRpRpKk\nv5EkSZUkKSVJ0v/wSR/Trxv+qVxLAEmS/liSpA1JktqSJGUkSfqfn+b3n6aTGfnQf/8B8C+ARUD/\nfIl/5y7lkiQZNU37lezjIEmSCfgbYBu4CcSBP5ckqaVp2n//qziGfyL4tb+WH/y9/xb4E+C/BH4G\nOICpp3qTc3a+/gOg9O94/Q1gCPwGcB/oAC8C/zfw9Z/72f8D+NaH/m0A/jlwCKgfLOjfe8rj+vd5\n3Khwf+i1fwbk+EDzH3/9k7mWwQ+u5UujrO+XlYP/S+A/A1Z4HE2fBP8C+F3gPwSeA/534P+RJOlF\n/QckSUpLkvRf/4L3eAm4p2nahzeU/g7g53GEGuPp8Uldy98EWsCKJEmPJEk6liTp65IkRZ/m4H8Z\nHwesAf+Npmnf11/4xz4lS5IkO/BfAC9rmvb+By//X5IkfZrHj6i7H7y2A/yiVm0E+PkJ1SyPH70R\nnvwCjfEYn+S1nAVswH8O/CdAE/gfgb+RJOmapmnDJ1nAL+vzrn/2lD+/xGO/wQ+kj55BM3BH/4em\naa9/DMc2xtPhk7qWhg9+5z/WNO1HAJIk/SFwCrwK/OBJDuaXRfCfH6cZ8g8Vmw9/uqeDx9His/zD\nu/ppzD91IbtrAAAgAElEQVQZYOHnXgt/8N7jjVLOh0/qWqY/+L6lv6BpWkqSpBqPxYMnwi+L4D+P\nPHD15167yuPiD2AD6ANxTdPeHeHv3AH+mSRJ7g/l4Z/n8YneHeF9x/i3+FVdyx998H2JDyK/JEkR\nwAUkn/RNflWNnr8HXpUk6fclSVqQJOlfAvP6/9Q0rQz8b8CfSpL0R5IkzUqS9LwkSf+pJEl/oP+c\nJEk/kCTpP/oFf+f/5XHl/jVJki5LkvTbPK7m/9cnzdnG+EfxK7mWmqZtAH/7wfu8KEnSZeBfA/f4\nt+T/R/ErIbimaX8F/E/A/8Lju1Hisdz04Z/5rz74mf8O2OQxWT/PY4+yjjkeKyL/f3+nB/wWYAF+\nDPwZ8H9qmjZu9nxM+FVdyw/wBzx+IvwN8BZQBn5H+0BHfBKM/eBjXGiMvShjXGiMCT7GhcaY4GNc\naIwJPsaFxpjgY1xojAk+xoXGmOBjXGiMCT7GhcaY4GNcaIwJPsaFxpjgY1xojAk+xoXGmOBjXGiM\nCT7GhcaY4GNcaIwJPsaFxpjgY1xojAk+xoXGmOBjXGiMCT7GhcaY4GNcaIwJPsaFxpjgY1xojAk+\nxoXGmOBjXGiMCT7GhcaY4GNcaIwJPsaFxpjgY1xojAk+xoXGmOBjXGiMCT7GhcaY4GNcaIwJPsaF\nxpjgY1xojAk+xoXGmOBjXGiMCT7GhcaY4GNcaIwJPsaFxpjgY1xojAk+xoXGmOBjXGiYzvuLkiT5\ngTeAI6D9cR3QxwgZmAa+o2la8RM+ll9rXORreW6C8/iE/JsRfv9XhT8Cvv5JH8SvOS7stRyF4EcA\nf/Inf4LJZGJycpLBYEChUGBvb4/p6WkikQiyLHN0dISmaUSjUR49ekQul6Pb7WKz2XC73ciyzOHh\nIYFAgNnZWSKRCLVajU6nQzgcZmNjg8FgwOc//3nu3btHJpPBZrPh9/sBaDabJBIJDAYDlUqFbDZL\nu93ma1/7mjjOMX4hjgB+8zd/k3A4zNTUFNVqlWq1CkAgEGA4HHJ8fMxgMGBqaoqbN2+Sy+XY39/n\n8PAQr9eL1+vF5XKhaRqtVotyuczp6Sl2ux2/34/D4aDRaNBsNtE0DbPZjMViwWKxUCwWsdlsTE5O\nUqlUcDgcBINBCoUCxWKRr371q+I4nwajELwNcHp6isvl4ktf+hKFQoFqtcqbb75JuVym1+vhdrsx\nmUwcHh7y3nvv0Ww26XQ69Ho9FhYWeOGFF5ibm+PBgwdUKhWGwyGdTgev14vFYqHX6xEMBtE0TRA+\nkUiwsrLCxsYG3W6X5eVlzGYzqqoyGAyYn5/n5OTkI8c5xi9EG2Bubg6fz0ckEsFsNhMIBFhdXeWt\nt94im83y3HPPcXp6Sr/fp1gs4vP5CIfD3Lx5k0KhQKvVQpIkQqEQBoOBfD5PNpulXC5jMpm4fPky\njUaDUqlEo9HgzTffJB6Pk0qlqNfrAJhMJvb396nVapTLZTqdDhMTEx85zqfBKAQHIBgM4nK5ePvt\nt1EUBUVRyOVyJJNJms0mPp+PfD6P0WgkkUjQarWo1+t0Oh0WFxdRFIVisYjf70dVVUqlEuFwmM3N\nTVRV5fd+7/ewWCycnp7y6NEjHA4HVquVs7MzNjY2aLfbzM/PE4lEODo6Ymtriy984QsMBoNRl/bM\nIRgMcvnyZRGIXC4XkiQxMzOD1+tFURQsFgvD4ZBarYaiKBgMBvr9Pna7nU6nQ71eF9E6Ho/z2c9+\nlkajgSRJmEwmwuEwDoeD/f19dnd3OTk5IZfLEQwGcbvdDIdDyuUyrVYLq9VKsVik0Wice00jEzwc\nDhOJRMhmsxiNRjRN4/T0lFarRb/fp1KpYLPZ8Pl8uN1uTk5OGAwGOBwOfD4fqqpydHREPB7HaDQi\nSRJut5tKpUImk8HhcBAKhajVahweHtLv92m1WmSzWVKpFGazmX6/T7lcZn9/n/v373Pz5k263e6o\nS3sm4fV6aTQa9Pt9APr9PtFoFLfbjaqqeDweJEnC6XQyGAxotx8HVbfbTafToVKp0Gq1sNlseDwe\nXnvtNXK5HOl0mmq1SigUIhQKMRwOKRaLtNttjEYjqqoyHA4ZDoc0Gg1sNhuBQIDT01PS6fS51zMy\nwQGmpqb4whe+wNbWFu+++y7b29u89NJLOBwO7t69y+rqKm63m0wmI6LD7Ows6XSaUqlEKpXCZrNh\nNpvx+/10Oh2mpqaw2+08ePCAQCDAxMQEg8GA+/fvU6vViEajhEIhFhcXefnll/nOd77D6ekpzWaT\nhw8fEo/HP46lPVPY2dnB7XbjdDrpdDqYTCZkWaZUKlGv17HZbMRiMdxuN7FYjN3dXXEDRKNRDAYD\nxWKRSqVCo9Gg1WqxsrLCwcEBDx8+ZDgcMjExwcLCAgsLC6yvrwPw/PPPc//+fXZ2djg9PUWWZeLx\nOLOzs5yenn443XxqjExwo9GIxWLhW9/6lrhzb926RbFY5Pj4+PEfMZkYDofU63Vee+01ADRNE8Wm\n3++nVqthNpsxGo00Gg1cLheDwYB79+7x6quvEo/H0TSNiYkJUqkU9+/f59VXXyUQCLC3t0ev12Nl\nZYW1tTWuXbvG4eHhqEt7JnF4eIiqqoTDYZE21Go12u02vV6PUqnE8fExGxsbguiyLLO5uSmKe7vd\njtVqpd/v8+Mf/xhVVZmamqLZbAKP67azszMCgQCBQIBWq0UymaTf7/PKK6/g9Xoxm83UajVarRZe\nr/fc6xmZ4Jqm0e/32djYwOfzEY/HCYVC4pGjKAq1Wo1qtUqhUGB1dRWXy0Wr1QJAVVUkSUKWZWw2\nm8jxnE4n7Xabvb09Ll26JNKXubk5vF4vyWQSRVHo9XrkcjksFgszMzOEw2FisRg7OzujLu2Zg8Ph\nQFEUZFkmEong8XhotVo0Gg2RB3e7XRqNBpVKReTaJpOJWq1GpVKh2WxisViQZZlWq8XBwQHT09PM\nzMxQLBaxWCyUy2XW19d57bXXGA6HbG1tibTE5/Ph9/tpNpu0220ikQg2m+3caxqZ4PqCXS6XyNUy\nmQwrKyssLi6yublJMpmkXC6LSjmRSJBIJFhdXWVnZ4eNjQ1RwWuaRqlUot1uUywWKZVKlMtlcrkc\ne3t7GAwG7HY7zz//vHjETU9PY7VaMRqN9Ho9ksmk+FtjPDmCwSAvv/wygUCAwWBAsVhke3ub3d1d\nGo0Gfr9f1FyqqrKwsCAKw5mZGTqdDplMRhSeAGdnZywsLDA1NSVy936/T7vdptFocHh4yJ07d1hb\nW8NqtbKxsYHL5UJRFBwOB5///OfZ3Nw895pGJrjdbqdarTIYDEin07RaLfx+P0ajkVarJeSkYDBI\nvV4XxcbZ2RkTExO43W6uXbuGJEmUy2UMBgNer5d6vY7b7eb5559nOBySy+XweDziUTkYDFhbW6NQ\nKLC/v8/169eRZZlqtUqn0yEUCo26tGcO5XKZra0tjEYj169fZ21tjXg8jslkEimI1+slGo2ysrJC\npVJhe3ubYrHIYDAQURgep65Op5MbN25gMpk4ODhgMBjQ7/cpFApCLp6bm+N3f/d3yefz1Go1rFYr\nDoeDdrvNwcEBwEjp5sgEV1UVu91OKBRiMBigaRrtdltUx/F4XKQrfr+fw8NDFEUhFouRzWaBx5V7\nJpNhOBxisVhoNpsMh0Psdjs3btxgMBhgs9mYmZkReZ7b7cbhcIibZXV1FYPBQLvdxmQyYTCMbTZP\nC52wsVgMVVWp1+vkcjlarZYQABwOBy6Xi1AoRKlUEnq1/pr+JAWwWCxEIhHK5TLFYhGv10u1WqVe\nrxMIBLDb7eKG0XsgTqcTq9UKgM1mQ1VVkc6eByMT/OzsDJ/Px9ramnisHR8fI0kSU1NTvPHGG/zZ\nn/0Z+Xyey5cvI0kSa2trfOUrX+FP//RPBcmPj49xOBzY7XaOj4+JRCLMzs6yurpKrVbDbrdz8+ZN\n3nrrLfL5PJOTkxQKBSqVCvV6nYODA5xOJ8PhkMnJScrl8qhLe+ZQr9cpFotcu3ZNpCc//vGPhVJy\n8+ZNzGYzJpOJer2OpmkoikIwGGRpaUlo3AaDgbOzMwqFAoqiUCqV6PV6KIpCJpOh3W6zurpKKBTC\nYrHQ7/fpdrsYjUY8Hg/9fh+v18vU1BTpdPqTLTKnpqYIBAK0223W19dJpVJMTU2xv7/P3t4eDx48\noFqt0uv1ODg44FOf+hSTk5O8++67KIpCOp0mmUwSi8Ww2WwoisLq6irhcBhFUdja2sLv99Pr9fjq\nV79KqVRCVVUePnyI1+vF4/Hw5ptvsrCwwGAwIJfLoSjKuNFzDkSjURKJBM1mk3q9jiRJvPzyyxwf\nH9Ptdjk7O8PtdnN4eMjW1haJRIKZmRlWV1dJp9Ps7u4K9aPf74u0Ve8w9/t90cj5q7/6KxYXF7l0\n6RIrKyuUy2WhgddqNQqFAo1Gg9XVVWRZPveaRia4xWLB4XAgyzIGgwGTyYTH42E4HNJqteh2u6Jo\n2d/fZ21tDZPJRLPZJBQKkcvlODs7E8Wj3+/HYDAQDAax2WzU63VarRa1Wo2joyMajQYmk0n4I3Sl\nJpPJoGkamqYBCJ/KGE+ObreLoihUKhVUVcVsNjM5OSn8JT6fD7PZLNLQXq9Hq9VCVVUcDgeVSoXj\n42NOT08JBoOEw2GazSaVSkVIvM1mE6PRKNKZbrf7ka6pnlqaTCZMJhOPHj1CkqRzr2lkguv5mS7O\nu1wuAoEAwWCQ4XDIYDCg2WzSarW4d+8e+/v7OJ1OpqamxMmsVqsYDAYURSEQCIiToOfd77//PplM\nBr/fT6VSwel0cuvWLba2tigUCpRKJba2tnA6naysrGAwGIjFYqMu7ZlDPp8XgaLZbIqo63Q6iUQi\nzMzMiNpK7yCnUimSySRvvPEGmqbx4MEDUqkUoVCIiYkJzGYzxWKRfr/P4uIiOzs7+P1+vvzlL4tu\ns8ViIRgMYjQa6XQ6wOMOOcC/+lf/CpfLde41jUxwr9dLs9nk/v37lEolNE2jWCyyvLzMYDDgRz/6\nEUajEYPBwO3btwkEAtTrdQ4PD7Hb7ciyLIoaVVXp9/tcu3aNd999l29+85u0220mJiaYm5sTzsNy\nucz3vvc9JEkiGAwSi8VoNpuoqkq1WkWSJHGixnhyBINBAFwul2i46Q24ZDLJnTt3uHXrluh1rK+v\ni47lT37yExqNBvV6XUjA8Xic7e1tlpaWCAaDlMtlgsEgDoeDbDaLqqpCY9cNdEtLS+zu7pLNZqlU\nKrz88ssMh0P++q//+lxr+lhUFN1ApVsf9agNoCgKZrNZVMyyLIsvt9st0pgPe0cKhYKw0xqNRgKB\nAG63W6QndrudXq+HJEmiARQKhahWqxiNRux2O7lcbtSlPXNoNpvUajWCwSCDwUBck0AgIBpAvV6P\ndrtNLBZDkiQkScLlcgmrrK6CpVIper0ePp8PSZLEjRIOh3E6nWSzWdxuN3a7HbPZjCRJNBoN8vk8\nh4eHdDod4XfRO+LnwcgEz2azuFwurFarkH70rqXVamVpaQmA4XBIv98XOqfT6cTv95PL5cjn86JN\n3+v1+NnPfobZbGZxcZHhcEgwGKTX63F4eChyft2Q1W63OTk5wePx4HA4GAwGBAIBtre3R13aM4dq\ntSpUDd2v7XQ6Rbrx4osv8tOf/lSoWJqmIcsygUCApaUlLBYLmUyGn/70pzx48ABFUXjzzTd59OgR\nx8fHLC4uYjKZMJvNGAwGotEodrtdSJKqqrK5ucn29jZ+v594PE6tVuP09PTcaxqZ4KVSibt376Jp\nmigQCoUCqVQKj8fD7du3qVQqHB0dsb6+zo0bN/B4PGiaJiTAXq9HLBaj2+2yu7srup79fh+Hw8Hl\ny5ex2+30+30CgQBms1lEiEqlQqVS4erVq8zPz6MoCo8ePcJut4+6tGcOy8vLyLLM1tYWNpuNcDjM\n7OwsOzs77O/v8+qrrwrPt66OwOMcOp/PY7fbMRgMGAwGEokEExMTmEwmHA4HFouF9fV1IQqcnZ0J\n37+iKABCpHj11Vcpl8tsbm4yNzdHJBI595pGJvjCwgJOp5NGo8Hy8jJer1eYq/S0JJVKCdlHVVUh\n5mcyGXK5HP1+H7fbLey1/X4fg8EgHlulUklMiOjvnUqliEQiokt66dIlEUG63S7D4XDUpT1zaDQa\nRKNR3nvvPaFyeL1eJEmiWq1ydHSE2Wym3W5z7949nE4ngUAAi8VCrVYTT1eDwYAsy9jtdhqNBuVy\nWTxt9RtB73RrmoaqqsRiMWG6ev3113E4HMKyYTKdn6YjE3x6elpMZUQiEUF2WZZFV2swGAgfgi4F\nWSwWcrkcpVJJkFGSJJGva5pGt9tFVVXRmteNVb1ej3K5TDweZ2JiAo/HQyQSIZfLcXx8TKfTGfvB\nz4FqtSpqJkmShOSqX8NKpUI0GkXTNLa2tnjllVdwOp1UKhUx0GCz2bBarZjNZtGXKBaLdLtd3G43\n3W4XSZLweDxCXTMajSiKgqqqnJ6eioaP1+sdWSwYmeBbW1scHR1RKBS4d+8eiqJgtVqZnp5GURT2\n9/dJJBJYrVYxpVEulzk6OhJ3u9lsZnd3l3g8Lh6ThUKBer1OLBYTMpHZbBYRPx6Pk0gkADg4OGB/\nf18UnNlsdqyinAOzs7NYLBZef/110uk0+Xye4+Nj9vb2MBqNrK6u4nQ6RVSXZVmoKFevXsVqtVKp\nVIjFYhgMBnq9Hnt7e5jNZvF0L5fLItXc3NxEURRu3rwpUlWfz8fx8bGwYU9OTnJ0dHTuNY1M8IOD\nA9rtNv1+n7m5OWw2G3t7e3g8Hvx+P5FIhLOzM87OzoS11ul0inEn/W7v9/u4XC6cTicPHjwgn8+j\naRqTk5MiKng8no9M9+RyOQKBAIlEglKpBDyONvpw6xhPh2q1SqVSEaklgM/nY2JiAoPBgMvl4vDw\nkEajgaIotFotnE4nsViMBw8eIMsyfr9f6OOVSgW3281gMBBDE/o17/V6+P1+hsMh6+vrRCIRJEkS\nTT2Px4PdbhdTXOfFyATXDTkWi0VU3Ht7e6iqKvKnZDIpJrJ1Ud/j8aCqKt1ul06nQ6lUEpJRLpcT\nKky73aZcLmOz2cSYmx7hVVVFlmV8Ph8ej4dOp0Or1UKWZWH4GePJ0W63GQ6HGAwGQqEQTqeTRCJB\no9EQqcXBwQGNRoPFxUWRllitVo6OjoSUOBgMyGazZLNZPvOZzwjbs26i0m+gqakpEaw8Ho/ol+gK\njtFo5OHDh58swScmJuh2u/R6Pex2O9FolOeeew6bzSZSB92DYDAYmJ2dFeNkerGhaRrvv/8+nU4H\nSZK4ceOGmNdzOBwkk0l6vR7Ly8vcu3ePaDTKH//xH/P222+zvr7OvXv3uHr1Kp1Oh6OjI6xW69hN\neA7EYjEWFhaENBsOh/H5fGQyGfL5vFC29IC0sLBALpfjrbfe4vLly1gsFtFp1u3KoVBIyI5Xrlzh\nu9/9LmdnZ7z00ktiUNnr9Yq+xtLSkthRQdM0AoEAtVrt3GsameD9fp9Lly5x7do1IpEIw+EQt9tN\nsVgU8t/x8THBYJCbN29ydnZGJpPBarXy6NEjJiYmWFtbIxaLUa1WuX//vtC5g8Eg7XZbOBULhYLw\nEr/zzjscHR2JhoAsy6iqSq1W+1iKk2cRhUKBd955h5mZGUGwZDIpVKyDgwN8Ph+KotBsNvnJT36C\noigsLy/TbrcplUpCwUokEiwuLoraq9lsioac3hSCx1aPfD6PzWYjn89zcHDA2tqakJJjsdhIBB85\nzCmKwpUrV3jppZdEQ0aX+TweD0tLS4RCIbHfhqqqJJNJ9vb2OD4+Fl3LaDSKyWQSc5p6VzSfz6Mo\nCn6/n2q1KqL+w4cPKRaLDIdDZFn+SDdUz+vHeHq0222Rdh4eHrKxsUG/3xepptlsxufzYbPZqFQq\nAMzMzNDr9SgWi+LJq6eOVqtVKGdOp5O5uTni8TjVapVcLkcmkyGbzdLv94Vao++Q8PDhQ2RZ/mS9\nKAsLC6ysrHB2diZ2lcpkMrzwwgvMzs6KvOrs7Iy/+7u/IxwOYzQaOTk5Ed4Ug8EgipnhcEgkEsFk\nMpFKpfj+979PNpslEAiQz+e5ffs2Ho9HTJ6USiWy2SyDwUAY5n0+37jRcw4sLCxw/fp1NjY2KBQK\ntNttms0mCwsLANy9e1d4h+bn5wkGg5jNZrrdLna7HUmShGXj5OSE/f19Pv3pTxMIBLDZbKytrZHL\n5dja2uJnP/uZCErtdhtJknjuued48cUX+dGPfsSdO3c4OTkR5rnz4mNxE+p7YVQqFUqlktj4x2q1\nEo1GhQ6uN25qtZp4lOVyOU5OTpBlmVQqJXwHsiwLB9rk5CSyLFMsFtnc3CQYDBKPxxkMBqKw1Lf/\nKhQKzM/Pk8lkRl3aMwf9emxtbTE5OUkoFGJ/fx+v1ysKzvX1dR49esTzzz9PIpHAaDSK/kOv1yMc\nDgsdXZ/N1LcEMRgMPHr0iPX1dYxGI3Nzc1itViER7uzsoKoq6XSaYDDI9PQ0ExMTomN6HoycogwG\nA8xmsyB4r9djenpaTGvoHS6/38/ExITYjq3VaonvlUqFdDotiF8ul4WWfenSJSYmJsSolCzLQk6y\nWq243W48Ho/wOOgt/VEq72cVkiTR6/UwGAw0m02xN6G+bUQ4HMbtdostP2q1mpih7fV6GI1Gsc2E\nbqI7ODgQ9dDdu3d5+PAhp6enlMtlLBaLcC4ajUaazSZnZ2d0u11isRhXr16l1Wp9siNrFotFjBZV\nKhUUReGLX/wiExMTVKtVfvjDHxKJRJiamgIeD5DqhWK32xXbDmSzWcxmsyg26/U6Pp+P27dvU6vV\nxMDrrVu3aLfb/MVf/AXT09MiTanX64Lg9+/fHxP8HJBlmVAoxM2bN/nBD35AKpVieXmZ9fV1XC4X\nDoeDL37xi5TLZd577z12dnZ44YUXuHXrFicnJ9RqNbEPJTzuSezv77OysoLdbufrX/86fr9fCAz6\nvivb29tcuXJFKHKdTodYLEY4HObu3buf7NZtZrOZBw8esLy8jKqqFItFjo6OhOynuwv1Svz09BRN\n01hZWSEajaKqqiC8PijhdrtFR/To6Ejkavp7KYrC5z73OXK5HOVyGU3TcDqdwi/+4MED0fgZ48lR\nrVbJZrOEw2H8fj/1ep1wOEw2mxW7xN68eROv18v09DTJZJLd3V3gsW06EAgwPT1NLBYTu8LKskw0\nGiUcDotAZrVa+eIXvygMc6+99hrb29viBjGbzeTzeQqFApIkfbI6+IfzrMFgQCaT4f79+zSbTZxO\nJ5cuXWI4HIrRJFVVURSFqakpkdpYrVZ8/1975xfb1lXH8c9J/K+2E7tJE9txmpTihhi36lSVNkio\nBSEGe6n4Jw31BcHDXkAMCYGEmCZNMB4qIRBCgBBDCCRgb2gItIKE1K7tJihrm+ZP46ZpXLu2lyax\nk3i1Hcc+PNycozQtJU6apHXPR7qKfHRzfI/P79577u/P97a10dTUxNzcnNZZ8Xg85HI5nE4ntVqN\nubk5HfBRASB14qhgj8/n0xqIhvpQ5WjlcplQKKS9GvPz8xQKBex2u054CwaDJJNJbt++rYU4VeKU\nz+fTd9OVSgkqyqnSM5SMcltbG7t27dIBuvHxcWZmZpBSavtZLxs2cOX1mJ6eJpPJcOvWLaamprR2\noFqCFItFarWaFmXs7u4mnU4jhCAcDnP48GHi8Tjnzp0jkUgQiUR0lbzb7aZQKGgtPICxsTG8Xq+u\n/VSqsx0dHcRisQ0lyT+t+P1+nWoRDocJhUKcP39ey7GFQiGmpqaoVqvaV64KhA8ePMjs7CyJRIL9\n+/cTCAQIBAJIKYnH48TjcdxuNy0tLVQqFU6fPo2Ukvb2dmZnZzl06BBNTU0kk0nOnDlDIpGgvb0d\nj8ezoQLyR2Lg09PTDA0N0dTUxLFjx9izZw+Tk5P6AVS5CgF6enrI5/NcuHCBnp4eXC4X8/PzjIyM\nMDExQbFYJBaL4XQ6uXv3Lh0dHRQKBe0vV9FOKaUWXW9ubiaTyZDNZmltbdWV/Yb6SKfTDA4OMjw8\nzMDAAMFgkHQ6TTQapaenB6fTycjICOVyGZ/Pp6tzwFKXHRwcZHBwUJcRulwuLl++rFM2Ll26RDQa\npb29HZfLpdXLLl68yOLiIjabjWQyqaOoqsxtW6/g2WyWSqXC/Pw8LS0tWnixra0NsHJVKpWKfspW\n6ZUTExNaj7BSqTA6OqoVqXp7e/XAIpEI165dI5FIkEwm8fl8BINB+vv7dfGDkmpWrkGbzWZyUdaB\nequD1+u9J8FNraELhQKdnZ2USiU976r00Ov16iv8zMwMN27coLW1VS81qtWqjlH4/X4ymYxOqXA4\nHJTLZVwuF3v37qVUKmnNSpUDs142bOCpVIpsNkt3d7d+hYhaw+3YsUPL6drtdv1KE1W0oN4CUa1W\nSSaTOhfC7XaTy+Vobm6mr6+PsbExstksMzMzFItFWltbiUQigFVRlMlktOjQ/Pw8fX19JBKJjQ7t\nqaNWq9HZ2Uk0GqW7uxu73c7u3bu1nmAulyMcDlMoFIjH4ywtLeH3+/H7/bjdbmKxGG1tbVouWZWz\nFYtFqtUqsViMPXv26Dxzu92O3+8nHA7rVNl9+/bp/vP5PDt37tRJWuthwwbe1dXF+Pg4V69e5ejR\no3R3dzM7O4vb7cbj8ejbk9frJRAIMDc3R29vL3v37qVWq3Ht2jWGhobo6urSP+rk5CT5fJ7W1laG\nh4eRUnLkyBFOnDihs9WuX7+upSnUA0tLSwuRSIRKpYLD4djo0J46wuEwwWCQ0dFRhoaGWFxcJJPJ\n6Ajx5OSkLmRRIk12u107BlRB+PHjx1lYWNB5QcrtK6VkfHwcsAqZleyHz+cDYHJyUss39/f3EwqF\nSKfTuqRtPTwSN2E0GiWdTmOz2VhaWtJyACqfRAnF5PN5UqkULS0tep2ukt/VZrPZuH37tr59nT17\nVp+5O+gAAARcSURBVC9tVBAnl8tx69Yt7Ha7XgveuXNHV/TfvHlTV/Ub1o7dbsfr9erPDoeDzs5O\nXXw8MDCgSwKVLHapVNK5/aoYQgXxXC6XvrMrvUmllbO4uEhHR4f+31KpRLVa1TWd5XKZ6elpLem8\nXjZi4C6w1uCxWEw/TefzeWZmZvT7eVTt5dLSEqlUilQqxc6dO8nn8wwNDSGlxOfzaV28Wq3GxMQE\nwWAQKSUjIyMEg0ECgYD2oSuVWqWZ5/F4WFxc1MegHnhXHqfhobgAnZY8PT2Nw+Fgx44deDweFhYW\n8Hg89Pf3ay3IcrmsfeBKk1A991y5coW7d+9qt63yrgFazSqTyRCJRHA6nbo4Ql3IVP1toVDgwIED\nK9Mu6p9LJXdW7wacBOQTsJ1c7xiflq2R51KoBX+9NPLbcZ82Gnku123gBsOTgKnrMjQ0xsANDY0x\ncENDYwzc0NAYAzc0NGs2cCFETQhRXf67eqsKIV7ezANdK0KIzwgh3hFCLAghUkKI72/3MT1uPEFz\n+QEhxJtCiPeFEGkhxKv19lFPJHOlhu2XgFeAPkClej2wrkgI0Syl3JI3QgkhPgK8AXwPK3jRA/xK\nCCGllI/FpD0mPAlzaQPeBMaAo1hz+XshRFFK+YM1d7TOyNeXgdkHtH8aqAGfAi4BZeAI8EfgD6v2\n/QXwtxWfm4CXgZvA+8B/gBN1HtePgDOr2r4IzAHO7Y4YPo7bYzyXn8MKOvlWtL0ITLEcv1nLtllr\n8B8C3wSiWGfgWngF+ALwVSAG/Bx4XQhxRO0ghMgIIb7zkD6c3B+JKwFe4OAaj8NwL9s1lwPAu1LK\nuRVtp4F2rLvNmthwNuEDkMB3pZRnVMP/S1gXQniAbwEflVJeWW5+TQjxceAF4F/LbXHgYaHa08AL\nQojPA38GwljLFYBQfcMwsL1zGQTeW9X2HtYyKsgaT7bNMHCwbkn18CGsfIO3xL2/oB14W32QUh5/\nWCdSyr8IIV4CfgO8DhSxrkBHsG63hvrZlrl8VGyWga+WIqpxv8dmpXigF+tq8UnuP6vrSv6RUp4C\nTgkhgsAs8GHgVWCinn4Mmu2ayyywb1VbYLnvNcuWbZaBr+YO8MyqtmewHhgArgJLQI+U8t+P4gul\nlFkAIcRJ4IaUcvhR9GvYsrl8G3hRCOFbsQ5/Fuukub7WTrbKwP8JfE0I8TzwLvAVIMLyjyKlzAkh\nfgr8TAjhwhqcH/gYMCWl/BOAEOIt4LdSytce9CXLrqWvA/9Ybnoe+AZwYrMG9hSyJXMJ/BXLC/O7\n5WVnD5Zn5sdSyjUvN7ckkimlfAM4BfwEa8ACy920cp9vL+/zEjCCNcBnsXKUFR/Eeor+n18FfBY4\nh/Uw8wngOSnl3x/FOAxbN5dSygrwHOAA3gF+DfxSSllXsMfkgxsaGpOLYmhojIEbGhpj4IaGxhi4\noaExBm5oaIyBGxoaY+CGhsYYuKGhMQZuaGiMgRsaGmPghobGGLihofkvKG7T4R2m6xQAAAAASUVO\nRK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f3fb5db95c0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "images = np.array([img[:,0] for img in w1[0:4]])\n",
    "labels = np.array([label.argmax() for label in dataset.test.labels[10:14]])\n",
    "plot_images(images,labels)"
   ]
  }
 ],
 "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}