example.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "entire-abortion",
   "metadata": {},
   "source": [
    "# Laplacian Score-regularized Concrete Autoencoders Demo"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "historic-mattress",
   "metadata": {},
   "source": [
    "Let import some tools"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "still-address",
   "metadata": {},
   "outputs": [],
   "source": [
    "from pathlib import Path\n",
    "from torch.utils import data\n",
    "from scipy.stats import uniform\n",
    "from sklearn.datasets import make_moons\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "import matplotlib.pyplot as plt\n",
    "from omegaconf import OmegaConf\n",
    "import numpy as np\n",
    "import torch\n",
    "from torch.utils import data"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "opposite-fraud",
   "metadata": {},
   "source": [
    "Do not forget to import lscae"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "sunset-shaft",
   "metadata": {},
   "outputs": [],
   "source": [
    "import lscae"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "august-conversion",
   "metadata": {},
   "source": [
    "The default config could be found in src/config.yaml, but you can also pass these arguments as here:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "damaged-printing",
   "metadata": {},
   "outputs": [],
   "source": [
    "cfg = OmegaConf.create({\n",
    "    \"input_dim\": None,          # Dimension of input dataset (total #features)\n",
    "    \"k_selected\": 2,            # Number of selected features\n",
    "    \"decoder_lr\": 1e-3,         # Decoder learning rate\n",
    "    \"selector_lr\": 1e-1,        # Concrete layer learning rate\n",
    "    \"min_lr\": 1e-5,             # Minimal layer learning rate\n",
    "    \"weight_decay\": 0,          # l2 weight penalty\n",
    "    \"batch_size\": 64,           # Minibatch size\n",
    "    \"hidden_dim\": 128,          # Hidden layers size\n",
    "    \"model\": 'lscae',           # lscae | cae | ls\n",
    "    \"scale_k\": 2,               # Number of neighbors for computation of local scales for the kernel\n",
    "    \"laplacian_k\": 50,          # Number of neighbors of each pooint, used for computation of the Laplacian\n",
    "    \"start_temp\": 10,           # Initial temperature\n",
    "    \"min_temp\": 1e-2,           # Final temperature\n",
    "    \"rec_lambda\": .5,           # Balance between reconstruction and LS terms\n",
    "    \"num_epochs\": 300,          # Number of training epochs\n",
    "    \"verbose\": True             # Whether to print to console during training\n",
    "})\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "human-spelling",
   "metadata": {},
   "source": [
    "Read a dataset / build a demo two-moons dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "geological-bidder",
   "metadata": {},
   "outputs": [],
   "source": [
    "from scipy.stats import uniform\n",
    "from sklearn.datasets import make_moons\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "\n",
    "def create_twomoon_dataset(n=1200, d=10, noise=0.1):\n",
    "    \"\"\"\n",
    "    Creates two moon clusters in 2D, adding p nuisance features and d noisy copies of one of the original features\n",
    "    n: size of data (int)\n",
    "    d: number of nuisance dimensions (int), and number of redundant copies\n",
    "    noise: noise level (double)\n",
    "    \"\"\"\n",
    "    relevant, y = make_moons(n_samples=n, shuffle=True, noise=noise, random_state=None)\n",
    "    nuisance = uniform.rvs(size=[n, d])\n",
    "    data = np.concatenate([relevant, nuisance], axis=1)\n",
    "    scaler = StandardScaler()\n",
    "    data = scaler.fit_transform(data)\n",
    "    plt.scatter(data[:, 0], data[:, 1])\n",
    "    plt.show()\n",
    "    return data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "distributed-anderson",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAD4CAYAAADvsV2wAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/Il7ecAAAACXBIWXMAAAsTAAALEwEAmpwYAABGT0lEQVR4nO29f5QU15nf/X26pwZ6kEMPFrGltkbIxIE1QcwI1iJWTo7RboQtLGksJGGt9Gadd3OUzXv8JrAK2VFMBCjaiF1eB2XXfuNVvD7xRoo86IfHYNggO8Jn31cOWoNmRhgLsvqJ1GjXrKHZFdNAz8yTP7pvU11dt+pW1a3q6u77OYfDTHdN1e3qqqfufX58H2JmGAwGg6HzybR6AAaDwWBIBmPwDQaDoUswBt9gMBi6BGPwDQaDoUswBt9gMBi6hJ5WD8CLK6+8khctWtTqYRgMBkPbcOTIkb9i5oVu76Xa4C9atAiHDx9u9TAMBoOhbSCid2TvGZeOwWAwdAnG4BsMBkOXYAy+wWAwdAnG4BsMBkOXYAy+wWAwdAmpztIxGNLC2HgROw+cwKlSGVfnc9i8dgmGhwqtHpbBEAhj8A0GH8bGi3jo+aMoV2YAAMVSGQ89fxQAjNE3tBWRXTpEdA0RHSSinxHRMSL6ly7bEBH9PhG9TkSvEtENUY9rMCTFzgMn6sZeUK7MYOeBEy0akcEQDh0z/GkADzLzK0T0IQBHiOgHzPwz2zafA/CJ2r8bAfyn2v8GgydpcKWcKpUDvW4wpJXIM3xmfp+ZX6n9/DcAXgPgvCPvAPDHXOUQgDwRXRX12IbORrhSiqUyGJddKWPjxUTHcXU+F+h1gyGtaM3SIaJFAIYAvOx4qwDgXdvv76H5oSD28QARHSaiw6dPn9Y5PEObkRZXyua1S5Czsg2v5awsNq9dkug4DIaoaDP4RHQFgOcAbGTmvw67H2Z+gplXMfOqhQtd9X8MXUJaXCnDQwU8dudyFPI5EIBCPofH7lxuAraGtkNLlg4RWaga+6eY+XmXTYoArrH9/rHaawaDlKvzORRdjHsrXCnDQwVj4A1tj44sHQLwRwBeY+b/INlsD4B/XMvWWQ3gHDO/H/XYhtYyNl7ETTtexHUj+3DTjhe1+9bdXClWlnD+4nRsxzQYOhkdM/ybAPwfAI4S0UTttX8DYAAAmPkbAPYDuBXA6wCmAPwTDcc1tJAkctPFfkSWTr7PwgcXplEqV2I7psHQyRAzt3oMUlatWsVGDz+d3LTjRVd3SyGfw0sjN3fMMQ2GdoOIjjDzKrf3jJaOIRStCKimJYhrMLQrxuAbQpFkbrqIFcjWoiYf3mBQwxh8QyiSyk23F1+5YfLhDQZ1jHiaIRTOgKou2QOnlML5i9NNxVeC/j4LW29bFviYaZBrMBhagTH4htC45aaHNaZj40Vs23OsnoEDQDqrF1yozAYe85axo3jq0Mm6e8hk+hi6CePSMWgjrPbNlrGj2DQ60WDsVQgqszA2Xmww9mH3YzC0K2aGb9CGl/aNmD07VwBrli50NcKqFEtl3LTjRaWVxM4DJ6THMZk+hm7A5OEbImE34LIriQC8tWNdU7GWeE/HFWhlCFfM7UFpqiJ1JV03ss/zWAXjzzd0AF55+GaG34XoClq6GXA3RNqk2wpA13SjMss4O+VdgSvT5hH4+fNNsNfQ7hgffpehU2PezYA7sadNRnWbWBlCf5+ltK2bX94tlVTl74D0aPMbDFEwBr/L0Kkx72fAnTLCUQqkCvkcdt69AuMP34KC4n6KpXKDQXbKHMtwfq6x8SIe3D2ZCm1+gyEKxuB3GTIj7TSOKngZcAKwZunCBpeHTP2yN+tlfi9r5Yh9BSm02jg6gaFHXqh/tuGhAl4auRlv7VgnfXDYP5eY2c9IYl0m2GtoJ0zQtsuQCZABVeM7r7cH58rNgU83/zUAbH5mEpVZ92uIAOzaMNhg9LeMHcXTL78rNaBOclbWtdnIdQ/tQ5BL120/bjEIZ/D3/MVpz3RRI9xmSBteQVtj8LsM1UArcPkBUCpXmrJpclYW61cW8N9ePgmJvQdQrYbt6+3BqVIZ83MWzl+aRmXG+5rLEmGW2TMwumXsKJ48dNL3MzjHMv7wLQ2v2R9kquMTyB5GznGKB1yWCPfeeA0eHV4eaNwGQxCMWqahgbmW2tdemeH67NatWOnJQ97GHgDOTlXqgc5SuaJkTL1m/0JI7alDJ9FnZUDe3qCmsdjdO0Cji2fenB5lY58lUjL2Tx46Wf88M8x48tBJbBk7qj5og0EjZobfRQSZ3acFK0PYefcKDA8VXMcvZtkA8OAzk5jxewI5/sbupvKTcrD//fqVBRw8ftozRXPxQ/tdH15ZIrzx2K1KxzIYghJ7Hj4RfQvA5wH8nJn/nsv7nwHwPQBv1V56npkf0XHsTiGJHG+VNMq0UZllPPT8qxgeKkgzjB7cPYnZABOXcmUG2/cew4XKbEPHLlVuGJiP544Ufbt9yVYqqvELg0E3ugqv/guArwH4Y49t/j9m/rym43UUcbYLVKmEFWQIvi6aVlCuzGJsvCjNiAljQEWRVhgOvXm26ZhOCQkAIIJrYDkbxA9lMGhEi8Fn5j8lokU69tWNqGjQhCGoCyeNxl7w4O5JbVW5UZE9YISujwgAywY8w6ys/2Mw6CRJaYW/T0STAE4B+FfMfMxtIyJ6AMADADAwMJDg8FpH0NZ9qu6fdnThyNDpBiFUA9flEPLKQHWG7mX0AfgqfxpZZkMrSCpL5xUA1zLzCgB/AGBMtiEzP8HMq5h51cKFCxMaXmsJ0i7QrcR/0+iEa+ZHpxcFhXWNMIDpWYaVCf73GQJ6e/S4ZEylriFpEjH4zPzXzPxB7ef9ACwiujKJY7cDQdoFygTInjp0sqlSVvYgUdWj0UVcHut7b7zGVxtHRmWGccXcHuRzwc7FLCP0ysCNTn8oG9JFIgafiD5KVJ2OEdGnasf9RRLHbgecGi9ODRo7MgPBgJJYWM7KYutty3QNXYm4YpTPHSli/crw7pCzUxVcnNZnvMNgGrAbkkRXWubTAD4D4Eoieg/AVgAWADDzNwDcBeCfE9E0gDKAL3KaCwBagFu7QDe88sWdDwOvvrMbRycij1mVuILB5coMDh4/jUKAHHo7WaLYYxz9fRZKUxXk+yx8cGG6QYaCEKyBi8EQFV1ZOvf6vP81VNM2uw7d+fVrli6USgq4zRZlD5J5vVmcv6TX2Onep5UlgCHV6gGqD7ldGwYDF5QR4s+Hd0o5iGuhWCo3SFWYAK4hKYy0Qozo1lAfGy/iuSPuf+v0+QsJgutG9uGmHS82HHNsvIhLMbgydM7kM1StsvUy9kD1ISdcYkH88UksL5nhKuNQyOdMX11DSzAGP0Z0as/L9gdUfeR2n7/fg2bngRO+hlQQxIiWKzPaiopmGZhSCI6eOX8RY+NFDA8VMLH1Fjy+YRDzesMFcnVTKldcH/BeEtXOh7PBoBOjpRMjsh6qosdrUBaN7PN8X/RkFW4Dt/dfGrnZt7erIEuEr96zIpS7JMmrKkPA35pr4Vy56iuPUkUbB04JZS+JasBbptpg8MOoZbaIIPn1KvjNncVM3i+oq3r8GeaGDCKVMRTyOXx68QKl/etilquzaUY0yYS4cM7o/VotCpVS00rRoBtj8GMkSH69H2PjRaVZs9dMXBj6zWuXKOfGLxrZh9/aPYFiqYxCPof7VntXP58qlfHSG2eaXp/Xm8X9qwcaUk87DZnqtPMB63yI+mH8+wZdJCmt0HV4pUUGQfRUjUqxVMbQIy9g3fVXBXK5CHd/sVTGc0eKuGnxAlejDshdOfm+3qbGH14uKitLytr06aHZmWVlyfUBL7Kn/Nw7AlOgZdCBMfgxo5pfL8Ovp2pQzk5VAneKslOuzODtX5Rx/+qBQK0K3QxWv8TfniFgwy9fg32vvt/w/pyeTOhCqTk9GczOsnKwOigZck8fndfb4/n9b167RClGYgq0DDowLp2Uk0YBtFOlMh4dXh6oiYebwdp627Jqrr2DWa5W0a67/qoGl1iUqtiL07MAxVf1K3uOnPMRUXNWWff3WU0aP2HdgAaDEzPDTzlpXMrnbVo8Ms13J2uWNgvhiZnvg7snXfXlg6wgVIjLReR1DsSDTlaA53x9zdKFDSsbokYfvsnWMUTBzPBTTlJL+SD59hdrK46x8aKSsQeAg8dPu74+PFSQdqsKYuzDiqhFxcqS5zlYs3Sha13E5mcmsezh/46NoxMNrz956GSDG4tt8ROTrWOIijH4KccvhU8HOSuLbbcvQ06xubkoiAqSOeK1UslL1DtVi7gyBKxfWVDOenHu1cpSYNVMwbzeHs/zdvD4aVe3XGWWA8tQeGXreFVWGwwCY/BbhN8NKt7fNDqBuYqGOAjC6NmVOS8ElP0N4m6SrVTGxov44MJ00+tWlpTlj4XPf83ShUrbM9CQHrrzrhX1Kt2glMoVT7nkYqkcStjNa39u14pOCQ9D52J8+C3Ar4et833dxUT5nIVtty9r8gd7KXE6/z7I9lbGPTURkMs8TM8wnjp0EvNzFuZaGZydqnhW8ArlzMfuXF73iUPiW3dWvgqSVhENi1NoLa4WmYbOw8zwW4Cfxs72vcdizcyRZbuouI+sDGHb7cuUt8/nLOy8e4XU8Hjp+zOqM+gLlVk8vmEQuzYMerptTpXKdYGyt3asw657BrUVvqUJp2snaItMQ/diDH4L8LpBx8aLscsDlCsz2Dg60eRKGh4qYP3KQt13niXCTYsXNLo/bMZbpBTKIAATW2/xnGWqBKXts1WhNqmyL7fGMutXFrDzwAmpKy1sN7C4unrJsF9DuiU8DJ2LMfgtQHYjZoiwbY9rb/dYcPp6hfyyyI6ZYcYrJ89h89oleGvHOrw0cnOT8R4ekgdLVQyOalDabuCCSFbYZ/yb1y7Bc0eKnv2AZbUBXuSsrK/khG7s51anhIehs9Fi8InoW0T0cyL6qeR9IqLfJ6LXiehVIrpBx3HbFZmRm+GqaFaS2N0DYeWcoxgc5yxclpljN3BBWkLa8esHLHLiKzOsnCGUz1l47M7leHRYXRsnKs5zG/Z8GLoPXUHb/4JqR6s/lrz/OQCfqP27EcB/qv3flYgbMS0BwmKp7KlrIzJDZAYkqmaQXX7CGbAGGg2cs1Bp14ZB6XGc28oCzIxq3ORCZbZ+XNUagHlzLksnqMokRGVOT/M8LaqEh6E70KaHT0SLAHyfmf+ey3t/COBHzPx07fcTAD7DzO977bPd9fC92huOjRdjM/hWhnDF3B6tsQArQ57BVy+Ctnn0qkp1exi4zWbdto1Lp79gG+OWsaORtIpUEZ+lEPDhauh8vPTwkzL43wewg5n//9rv/wPAbzNzkzUnogcAPAAAAwMDK9955x0t41NFVw9aN4MDADkrg7lWNpbALAENY1ZVYlQln7MwsfUW/w1tBDHSfsg+j1uape7P7gcBuG/1AA4eP53ocYHw59PQmbRVAxRmfoKZVzHzqoULm/VX4kRnAYtM9KxcmY1k7HslAcV5vVnsqhUObapl4KgWIqkSJr6gs81jkPRDr5RE5xnMWdnQlbYCEQsIY+yjCroZvXyDKkkZ/CKAa2y/f6z2WqpIwjhFRSbve/7SDDY/M9nwsHruSLEuOSCCeY9vGMTbO9bh7R3rEgky6swRD5J+KNu2UPP7OwOcVWkJ/xoEr7RNhrocRMPfaVhkF0tlDG5/wUgrGDxJqtJ2D4AvE9F3UA3WnvPz37cC3cYpjqW9l3FwPgxE9albVSlQDTJufnZSWUUyTI667DyEyRF3C4rKsoG8tvUKcDqVKw8eP41TpTLyfRaYgdJUBVkiaVB3hhk5K6s1cGtlCCB/tU+xAnNWbhsMAi0Gn4ieBvAZAFcS0XsAtgKwAICZvwFgP4BbAbwOYArAP9FxXN3EbZx04GVs3PB6WAljsH3vsbqbKZ+z8PkVV2H0J+82GBgrS9h627LA4w1ipP0Ikg0UJnNI9iBwxiG8zn+h9qB46tBJLQHiLFWD5QDqzelVgs9GWsHghragbRwknaWjM8Ao9mc3plHJWVmsX1kIlAViD2huGTta15jPUlWczNl20D52HcFr3ftKEjFu1ZWauFa27TmmpZ5Cdu2pZgIRgLd2rIs8DkN7kUiWThy0Ii0zDuMU1HDIEGqOv7V7oqnDUoaqs0G7W0ekZ5amKshZmbqssZ37Vw9IjX43I8uyclLI5xquFUBPfUWWCF+9pzkNdmy8iM3PTCq1auzvs9DX29N2D1pDNIzBTwlRcrTFTF2WbtjfZ2HrbcsCzy6zRPVWhe06E9eB87Ofvzjtex6DpoPmcxbmzelBsVT2dc09LikoC5JuamUaJwAmfVMfab5XvAy+kUdOEFnXJyfOoJ/d5y3zyZdqbqOgfV+F0fGTbI5Cmm8OwP2z+yGLQ3jFTD6/4qqG1dTg9heUHs7286c6PSO4B/GNXz86cd4rcZO6PPxORiXbR6QJynRRvFITwzQ8F2mEOlNS7bRDc46g581Lq8YrwP/ckWLD5/bK4Nw4OoHFD+3HopF92GRrg6iKbFsjmRyduO6VJDAGP0H8sn0I1Rty4+gE/uLcBdy3eqBJodKtGbh4PczNfO+N1fKIuDTV2+HmUP2MOSuLxzcMuqqGCrzUP8uVGWzfe1kNteQTzBerrzBOVxUROkM42rn/gDH4CeInBSyafgDVm/3JQycbpHsBuVvo4PHTvjdzhi7PKrNEDQHbuDTVk745wvR2lX3G/j4rsAKlX4+As1OV+pjiNL6iHsCOkUz2RvXaaef+A8bgJ8jwUAE3DMwP9DdPHjrZcPHJDGWxVMbUpelqkY4LBODXbhzAW49Vq2zfeOzWBn9yXJrqSd4cYd1Hss++9bZldS19r1m9E68eAcDl5u9xGt8+K2MkkwMQ5Npp5/4DxuAnzKE3zwb+m2KpjM3PTmJsvOhpKM9OVQBy777E8A4ax6WpnuTNEdZ9FMdn9/p84qE9PFSIrOEjQ6TghnlgdSNBrp127j9gsnQSZGy8GKhK1k5lhrF97zFsvW2ZZ364V/m9nxslDk31KFr5QbN7oriPdH/24aGCNEXW/tDedrv39xmF7XuP+Z6/IMV4nUyQayftWWdeGIMfA/ZCK5Fvnc9ZOH9pOtJ+z05VGgxo0EKuVvkYncZU+EplfQLcJARUUt90SmPoYNvty5qKpKwMNc3+51qZWAz+2alKvcrb7fw560JE3AgAVl27oG2NWhhUr512TskEjEtHO3ZfIHA506JUriiLlPkh+rTKxMz6+6zU+hi9fKXOc+c8W37umVT6Vp3+tdrvY+NFDG5/ARtHJ2JvWi9wnr+nX37XdbunXj6Z+lRa3aheO36unzBJA0liZvgRcavQjKvFHQG4bmRfXcnxgwvNKwa7yFkaZ2h+N4zfuVMRg0vL5xb9ce0I15y9nWKS2M+fzL3I3Pw9dHrRluq14+X6aYfZvzH4EQhToRkFu3tDpsY4r/dyj9W0XGR2vLKMVPBzz6Spt6vss4ad0RMBPQ69JEGGgPk5C6Wpiqc0hP386VRe7QRUrh0v14/XZCYt16Rx6SjitlQLU9mqC9ltek6DSmOcyAy2StuQlrtnAqIzdpCzsth1zyB23r3CtahqloFL07N4a8c66Tlynj9RdOekz3I3C+2QZx43Xq6fdijIMgZfAZnfOenepSqk/aZ0u2G89N2FaWun1DeB22e1JC0q3RCG3f7Zh4cKmJXMys9fmsF9//l/4qHnjzbN7uf1ZjGnJ1Nvfzk2XsSjw8tx/+qB+nFEMd6/v/P69MVCEkDF/+6VkqlSc9JqH79x6SggW6oFXRLrxmko2+GmdPpK5+csTwGxXRLVyHbAzS985vxFpeD9/asHGjJlRIxDGBbZZOOlN864vj51aQaMZt/yo8PLpWmYaYmFJEEQ/7vM9ePX7CcNPn4t8shE9FkA/xFAFsA3mXmH4/0vAdiJy31sv8bM3/Tbb1rkka8b2SedgepuZ6eKMPbioVNow5vST3PeTX643Vk0sk9pu3zOwsXp2SbjIWQbdGjud+L5DYtMdjrIORobLzbUXgjJcnFP6jiGCl7yyJFdOkSUBfB1AJ8D8EkA9xLRJ102HWXmwdo/X2OfJrwaYovlnQ5yVqaeaimW2TIHgF1zx96rtZ3wioG0w2olTkrlimcAMHir9GbS5FtuNVH972LyYl+tXnA0HPJKWEjKxaPDh/8pAK8z85vMfAnAdwDcoWG/qUEmejZVK6R6aeRmLUa/XJlFaaqC+1cP4I3HbkUhn1NSSkyb+qQqXjfT+pUF7DxwIrX5zGEJ0wjejjhn960eiDyWNPmWW00QzSfnudoydhQP7p70lWbwiq8lVe+gw4dfAGCv4HgPwI0u260non8I4H8B2MTMrlUfRPQAgAcAYGAg+kWtAzFzdpbKn52q4KHnj+LwO2fqxj8qDNSrHYPMwNpttjY2XkRGEgPp77Pw3JFiKF9n2svet962DJufnfT04+esLOZaGdf0TWE0hN9dtYOaV1OdNPiWW4H9Wpmfs2BlqeF7cVtlup0rr+/Afl+6+fidxJ3GmVSWzl4Ai5j5egA/APBt2YbM/AQzr2LmVQsXumu/t4LhoQLmzWl+PpYrM3jq0Ent1ZJPHTqJ+QGEtdKenWNH3DRuxj5nZaWFPxtHJ7BoZB8WP7S/STbavl97NtXG0QkMbn8hNTPW4aECdt61oiHL4/7VA/Xf8zlLauwBoDR1qf5ZVDVv+vsszOnJNPxuz3hqh54FunFeK6VyBeDqufESRAuaip23reicGT4y4py86ZjhFwHYE3o/hsvBWQAAM//C9us3AfyehuMmjuyLiCNPh1EttHHOzKxMVQ7TbyaSZmQ3TZYI61cWfGetds0Xu9GT7bdUrqRqxirL8lBpnH7+0kx9Vfn9yfd9j2VlCR9cmG4o1lL1LbfbqjEIbtdKZZbR19uD8Ydvkf5d0HNydqqCoUdeqAdv7d+9LIgb5+RNh8H/CYBPENF1qBr6LwL4NfsGRHQVM4ur83YAr2k4buJ4pcPFwdmpSn12dq5cqbsogPZOmZPdNDPMeO6I+kz8qZdPNhh8r5sxbRWPQLP7aeqSmixHuTKj5MopSCpuy5UZPLh7EptGJ6qrSEkhRDutGoMS9iEXxgYI1y/QOOHwS+OMg8guHWaeBvBlAAdQNeS7mfkYET1CRLfXNvsXRHSMiCYB/AsAX4p63Fbg17EqDs5OVXBxeha7fFrrtRN5j8BlkOUyMxpcNX4GKk0zVjf3k063oEj1k1VezzDXXRlumdnttmoMileQdsvY0Xo/Yaf7MKwNcLa3BFqjq68lDz8u0pKHb8c+K0vyzIkb2G3ZL/Kz2+VhMLj9Bc9iqyDYc5hVXCJpqVuQLed1QQDe2rEu1HGyRPjqPSva5noKg+w+umFgvmvxmr0dqHNltujDORx686xSEebjCRQSxpqH320IaeK3dqzTln8v8ErZE7PTdg2w2VPZghr7eb3yGZV91i5mTF7nUdyUrZb8jXu1IWawsqb3Xswyd7SxB+Sza1lHOpmU9PmL0/izt9WMPVDN9GslxuBHQJeLh1CdQYw/fIv0ISJu4HYMsDndF0HIWVl84Qa58XEuzYeHChh/+BY8vmHQ94HcygelzKWQz1m+WRwqCHeMV1tLGZ3su7djn7wJd6nMcIvX3bJ7gvS5KJUrLc0YMwbfA79iFDFLyEnUBVXo77Owa8MgVl27oL78dt7sdn9qkk3BdRFWVVTMumRGiyDvHStuZj/D2aoHpUx1cdvty7SsIHceOIEtY8EF/tw6cnUTbkqk9td1KOS2cjVuDL4E1S72VaN/fagT2d9n1VPAZJ2enIGcVHZ18iGsUd28dolnK0eGf5ql34NQvJ90palKwG7z2iXVNFwHKvMLv4IgGVfM7el4d44XMslo8bqOCUKx1iylFRi1TAlBmhnsPHACjZnN/hDQ0JnKbdYgZrD246Wtq5MKYVLZ+vss3wBsf58l7Y0r8KpuFA/KVlWaKjVrcdh7K0vozWZQuRSPYF8poXaLaUUEZmWN3XWlZjuvr6QqxE2WjgSZQqbIflDZ1o+3a/vx+vtOUDRUyZ6xQ4CvbLKVJYDRUFAky1Zyaypvz9JJSsUwKHFn8rgxrzeLfF9v20wmksbZ+D0KcWXeeWXpmBm+BNUu9l7belGwuRJkmjJAuoOxqthXJX7niVAVBnvK46YiAqZnuSl/XLYC85tJpzUQHufxc1YG5UrzuvT8pRmcv1Q9brdo6gQhTBBchkrmne7zbnz4EoL4yr1S3zJo7nIk9rNl7Cg2jU54pnSlORgbBBFE9SJLhF0bBvHo8HJvZUGGa7EQEM5Ipi0QLuIJca69y5VZpUygdkj5jULQ2I3Oh7C4vmSToDge+GaG78CpoDfXyqA0VcH8nAUiYNPoBLbvPQbmav/YfJ+FDy54KGUSsOGXr8HB46cblskApI3IBWkPxobBq0uYPf9789oloZp8hDHSrShxl+Hn/tLZcEf1gdLqlU5chInd6PLh2+NHshafcUw4jMG34bwASuUKrAyhrzfbJIvs9rMbs1w17PfZKvUAKM3g7AqHnYLqamZ4qICvfPcozgcIToY10mkKhHul/Ym4g4prTCedssp0InOlbN97zPVaGBsvKsugWxnCFXN7UJqqamCtWbqwadIn4keyWGEcE46uNfhuUXGZgl7UjAhG1eivunZB3YiozJrSpvKog4JkhuR2gU8FOO9ZoqYgV5DMB6WMmQSQXRcE1F1ih98547s61EUnrjKB6rUhe2ienarUJ3LFUhmbRifwzOGTeOXkOeXVlV150+s69FLgjeN67LwppAKyHPs4Z02MxoIL1VlTO/tQ3fyjsurkuVYGm0YnGvyoXiJrdnJWtkn7RbWOIm34xRPGxot47kgxEWPv1M3vFMS1oQqj2hw+qCvtVC3f3us69GqfGgddafBlSzlZlZ0unN1vVGUZ2tGH6nahbxqdwMbRCczpydQbTQidnHJltumGkHl/+qyMr8Jgu2oO+SUL6Kj0VKWvtzOLsJI6h1fnc77XYdKFlF3p0vHSY9cZFHNydT7XsLzL91kgMKZq6XFJBm/ixu1CF5+tVK4gZ2Wl6Zflygx+a/cEZiUGv1yZxc98Mn7SmmopkC3z/eIJSY4/LedKN16fK+9T/6GKMNqbJIkHwpvg/L7tySE7D5zQHkvqyhm+1zLKXu6er/W51EHOymLN0oUNs96zU5W6sQfcjX3afaiytDY/Y1GuzODpl9+VuiZkxh6AUutHqThZxCbiOvBb5ruJegmSfPi340RDBa/7f9vty7QcY/3K6sNbdiwCmr7vXRsGcXF6FmenKrG5IbvS4Hsto+w328TWW7Dhl6+JrFxoFwFTWT1kiRJriBAFL8OlYixUJWWd2D1vsgfO5rVLXB/WH1yYbrkfP4q7SXbt3r96QGtznrRPNKLgd//3RRBDFIj2k5vXLnG1H86YHpCMG1KLwSeizxLRCSJ6nYhGXN6fQ0SjtfdfJqJFOo4bliCdZg4ePx0pQPa2bZamukSeZXad3aUNrws0zu5gIoNiy9hRbBydaHjgbH52EmPjxWrT+d5mj2Vlllvux4/ibrJfu0B1clCuzODg8dNYv7LQ8HpYOjVYK/C7///9nddHPoZwCw0PFaT2w/l9J+GGjOzDJ6IsgK8D+EcA3gPwEyLaw8w/s232GwDOMvPfIaIvAvhdABuiHjsKqml4UTJ3nJF21aINr9lxUiJLKnhdoE45BVl8IgwE4L7//D9dOxNVZhjb9x7D8FBB2t6v1b7pILIdbohz6ywaeu5IscFwhdXi6dRgrR2v+394qKAt9dWrsMrpmox6XaigY4b/KQCvM/ObzHwJwHcA3OHY5g4A3679/CyAXyGKOSVGE1FmSsVSucHNsGbpQiX3kGwpnbZUQ78UQuEee3vHOuzaMCg9l0HPsEiTkyFWAH4PzlahIzNDZfkfptsV0PoHYhp4dHg57ls9EGkfQ4+8gM3PTEofGs7bIYmMHR0GvwDA3v/rvdprrtvUmp6fA/Bht50R0QNEdJiIDp8+rU+oKCyqfmbZiRRGecvYUaX86XzOks480pZqGOQCHR4qYFZyLt1e9WprGGR8bnryDGDzM5MtM/o6mld7ra5EXCOsqqNKULwbiCqUdnaq0qDm6sQpRZ1EU/PUpWUy8xMAngCq8shB/163y0NWGerESw9fZKT4PTxExyMZaUs1DCpJEESHxMpmUMj3RnKpDQ8VsH3vMVf5C+HLb5XrImplr+xc5hX6CPhx/tJ0PQ7SzSTVd9hO3BXfOmb4RQD2NjEfq73mug0R9QCYD+AXGo7dQBwuD13BRy9jr/o0T5uqI+CdQugkyLkslSuhjX3eNkP1aujRzq4L2erqQmUmch1JZab1ge1WI2TL4+T8xeQzxnQY/J8A+AQRXUdEvQC+CGCPY5s9AH699vNdAF7kGDqvxOHyGB4qYP3KQuTUTBmFfE45I6cd2xvacVuyxsGyqz9U/9nrYdjOeeZu53L9yoKrxn0Y2vlhGBUxcZRN0m5avEDLcYRWVpJGP7JLh5mniejLAA4AyAL4FjMfI6JHABxm5j0A/gjAfyWi1wGcQfWhoJ24XB5RUzMFVpYaOtxb2WANo9Ok6hgW55J16JEXfBVHg/LjN87UXRKb1y7B5mcnG8470J7Nut3clfYeAzfteFHbsdJQoNYq/KQXfvzGGczrzQZScpURV6MTGVp8+My8H8B+x2sP236+AOBuHcfyIq60Jh2iav1uuvkhniJpUXVUQSWesvW2Za4GOQqiqMV+ruy+/HzOwrbbl7XNeQTctds3jU7g8DtnsOraBdolk1Pc+TRWvFQ0BQxoMfaCJFdTqQvaRiGORhZeebREwPy5Fs6VK562m1C9gZwR+1YHDuNEtbmEc9Wiy87Yb6J2ekjKkGkTPXnoJEb/7F3PbJAwyGoYOpmgKpq6SNK12FHSCnGkNW3bc0xqhMRM8a0d6zz90Qz5DdSpvtIg8RR74FcbJM+1D9rWLg14XSe6jT3Q3vGNsERV0cxZWTy+YTBQbCrpGFxHzfABvbO5sfGip3Le2anLDUqEMp7brScugLir6NJE2HiKahqsH8zA5mcnATSuKMK0tUsDYVrrZYmw+uP9gRp3AO2VCKCTqJMvu6yISmqsW9OeuOmoGb5utu055ruNPehy3+qBpmwecfO0e4ZNUFQaeUhFz1yKpcLgll6YtuI1VWQiXF7MMOPP3j6L9SvVDIp9VXz4nTNY/NB+LBrZh8UP7ceWseRdHUmjI1AtZEX8VHfdmvYkQcfN8IPgFlQEEDgAJrZ9dHh5PYAmC1S2c4ZNELziKV6zbADBtRY8KJbKuG5kX/18p614TRWZvouVJcAlPiSozDD2vfo++vssz2yoLBHeeOxWAFVROnuV7gxz/Xd7X+ZOYmy82JxUEYIMUT1DLGy7zTihGNLhtbFq1So+fPhwLPt2Gh3A/+aRQQB2bRjsWOMdFtlFLhP18nJ96SBnZTGnJ+Pqpivkcw0pjmkl7CTFz+DftHgB3v5F2TNwbn8opMWA6SKs0JwbOStbd9W04jwR0RFmXuX2XtfO8F0blodMDbSnARouI4uneM2y45x+lCszmGtlmrqatZNrTXZOh4cKWDSyT/p3frUOP37jjO+5F4VI7RoH8ULnCs/uIkzbeepaH77uJbzK/toxOyQOvPz7cfcVPjtVwVxbg4t8rnO03/MBRM/6+6yGmJLKg1Z8N+0aB/FCd/LEqVI5leepaw2+7i/Yb39pkzZuJV4B7LBdsIJgn+1enNYjRZAGtt2+TDngfXaqEjgF8d4bq5JZ7RoH8UJVulyVDJHURdTK89S1Bt/N6KjcK26bEJq1752k8WnfKrzqJeLS15HRSd/B8FABO+9e0XBe+zVJJGQArLq2qiGTRhE/VdxW2WPjRSXp8iB4TVxaeZ66NmgLNAeepi5NR9Z1EQEboDEjxysgREBHBL504BZMT4JO/Q7GxovapCtEYNvtO7IHKtOKbNyyQH4cJHGevIK2XW3wnVw3sk/LU76/z8KFymxgo9UON00SjI0XsXF0oiXH7sTv4Jf+7Z9oUdEkoF4N3Q5ZOnFM6KJQMFk66SJMNaMbYS+qpJXz0srwUEG7GJgqnfYdbBk76mnss0TKcRO7KyLt+kRumUStJC1pv13rw3dDV7OTKLRz4EsnrjGWhI7dbt+BLPvLWUDlhqztpJN2k5MOoovT32dpq+52I01pv2aGb0PMWLbtOVb36RFVdVnyOQvnL01H8oXKVDfttEPgSxdebgE3WeO/lbOw7OoP4dCbZ2PN5mmn78BtJrtxdALb9x5TUrzM+xRk1Yk3W1Y7qjP6nJXFJ6/6EF5640ws4yjkc1izdCF2HjiBTaMTLXd/RZo0EdECIvoBEf157f9+yXYzRDRR++fshpU67Kl6zJd7ze68a4Vynribps6nFy+o/32Gmk9+mmYCceOWprppdAKLHLPUCzZ3RKlcwSsnz+Gr96zwzD7JUPVB7UeG0DSza7fvQDaTPTtVgUrB+EXFWXA7tT0UkuZu5HNWU5ewH8dk7IFquudzR4qpSceOFLQlot8DcIaZdxDRCIB+Zv5tl+0+YOYrgu4/6aAtIC+x9spQkFHI5+qzV/HFO6Uc5vX24Fy50vInf9L4lbJ7ZU9kCErGzAshbQ20t76RrkQDFexB2zQju7bcJFB0Siq4IbtW4/Tpxxm0vQPAZ2o/fxvAjwA0Gfx2wq+oRFwsD+6e9HQrOL/Qm3a86CrlUCpXcP/qgY4VpZLh5ycvezTjDmPss0RSdcJ2MvBOdCUaqB4rzQgXoex8MNCkb6PjYenlqpVdq62KE0WNg32Emd+v/fwXAD4i2W4uER0mokNENBzxmNqxB71kneqdGQpfvWeFNMDr5hbw+oKfPHSyK+Rn7SRtPGaZ29qwy/BLNMhZGS1yFWl3ddldhDIK+VyTKzEqhXwulKxyqx6evgafiH5IRD91+XeHfTuu+oZk5/Da2hLj1wA8TkSLPY73QO3hcPj06dNBPksonBeA26zd7WK3V4sCl3VGRNUogIbMifk+OidPv/yuhk/TPqhkRM3r1ZcxJWRrOw1xHbrp6FRrCq6vK1yqkiXCTYsXaO0cFzcqWTlrli7Etj3HQhf1yXpdlDyC3rmaWJ/b37UCX5cOM/+q7D0i+ksiuoqZ3yeiqwD8XLKPYu3/N4noRwCGALwh2fYJAE8AVR++7yeIiOxCyRJhltnTryvLRXbLnLCyBCtDUunlJDRk0oS9l22xVG5aFuesLKxsBoCeitsZ5pYrFcaFuA69sp5UOomlJVc8DCouku9Pvh+popZx2X1jz77xunMfu/N6AOmJE0X14e8B8OsAdtT+/55zg1rmzhQzXySiKwHcBOD3Ih5XG7ILZZY5UIDKfrNlXIpZKjPsqUnuXHa3QyVjVOwPTLfPu0lztW2nFVU58SqG8mu7l3aXjR8qsQwd8gnC2PudTwJw3+qBpjTjVhPV4O8AsJuIfgPAOwDuAQAiWgXgN5n5nwL4JQB/SESzqLqQdjDzzyIeVxuyCyWIj805o5fN1ktT1QCtWzHM6o9fzmjtRL1xP9yMVZhqW6GJE0Wp0PnwWbN0IQ4eP93WD19ZxyzgcsZSu30mO6p9ZHUgkz4WJCWhEIZIQVtm/gUz/wozf4KZf5WZz9ReP1wz9mDmHzPzcmZeUfv/j3QMXBc6es2qVvVdnc/h0eHluGnxgqb3Xjl5ru5jNsqaVdy+G6/wYyGfw1s71nn2fw0jY/3koZOpyaOOwsHjp13dD/Pm9KTSOAXBGVOLk6tr6dZuEICXRm5O7fnsemkFmVQvAOVmJSqzRvtD5O1fNG9vN+idqDfuh5s8gPMm9qtUnro0XZ+du21HgO+DXOXh3a4P306/roaHCnhp5Gap0dehniDu43aViDbSCmh2JwR1qfj5D51LPL8bT4ebqZ3wO99efXDtnJ2qeC7r3fKwnS4aVePXjkZSdl0xqpObtLohgiKPy0Xft+iWtnntEmx+ZrIhCaMd9Ia6fobvRhCXyth4EVOX5N3uhbHfeeBEffYqy9sVBl2Hm6mdUDnfqgbWa3aeraVmenUeU32otuPD1ysVtl1cVTKhOJVamqDYV5YCMak4/M6ZZv9iG+gNGYPvgurSVxgPWeZNzspizdKFTQbmgwvTsLJyDRevjlCdiMr51mFgZ5h9Hy4qD1Urm/6ZnBteOftA+l1Vsof1lrGjvrU0Tvr7LE/73N9n1d1Dzr2VKzN4+uV3m4QU20FvyLh0XFB1qahE6t22qcwy8jkL8+b0SDM/0q43rhPZ+bYXq4k0zSir8oJHsK1YKuO6kX24Op9Dzsp4Nwxp45IJ0WtAlqKYZleV7GH99Mvvuhp5L63/s1MVz9oE8Wey8yHbb5rPH2Bm+K6oulRUIvWybc6VK3hp5Ga8tWNdqqP6SbB57RJXPfLztSAsUDVUUeysCNh6rRTErNGvO1RlNv0zOS+8jFKaXVVBja/fTN9rlVYqVzzdrzKvUZrPH2AMviuqLhWVSL3sggmjv9GpDA8VcMXc5sWmfYnsJXmrAqMqeCeqeqMiVgR+GVwCme85DvyOJbtuVbKY4iTsuGVaQV7fc5+V8X1oy9yvwOUVgB0TtG1jRIqX1wxcZSUgm2R0mZKCLzI9EjGr8ythV0HM+HSdetW8fNFI3O573vzsZCxG3y8oDcjrG+yVoUkTdtw5K4t7b7zG9fN4pvBWZpWK+iqzjHm9Pa4BXCdXzE1/PYMx+BFQWQnIug6pdCPqJvxWS2n2jfoFO7fvPeYa4Nu+95j2sahkPLldt7s2DLZUojvsuB+7czkeHV7e9LrO+ZQ93uG1Xy8RtbRggrYR8QuudltOfVg2r12Czc9ONhhGezZMkrrvYfB6IMmyuMI2uw8zDufrupICdGk+RR2383WdjU0Iai0T2+GeNgY/Ztw0Pjo5pz4SzumT7fcktVLCkJabXecEw8+Y69R8Uh23fUz5PgvMcO0YF+Z6sTJATzbb9Dcqq4V2uaeNSydmui2nPiw7D5xoko62Z8M4z2N/n+Wa2dMKCFWtdVnAUZb3Lns9CrqK9lR86jo1n1TG7RzT2akKSuVKfXwbRyfqjYSc14tKE5jKLHDDwPzA30t/n9U293SknrZx04qetobWIOvN6tVH1T7bA1qbHm9lqcEdJYKgjw4vrwZtXcrwd97t3nIxKjrcLH69nYFw31mUcQdx09jlTMbGi/jKd4/i/CX/2X6WCB+dP1fpOGlVxYyzp63BoIUwrgi733bRyL7YxqaCMyjLAJ46dBKrrl3Q0OwlCYllHf551epnnfEpv3EHCdyLFcnhd87guSNFZdfODLPvcawsYedd8Tys48YYfEMqaJdYRzZDmFFU4WKg3nBFZ+V0Es1xVIx50t9Z0MB9uTLj2nvCC5UZfk+G2tLYA8aHb0gJUWMd/QkVsgW9YYqlstaCKxXfepR9i7Gev9hccCSyVdzkq8V3tn5loUEoUGetgVefA13ce+M1vv2W/Sqx00wkHz4R3Q1gG6pdrT7FzK4OdyL6LID/CCAL4JvMvENl/8aHb1BFFDc5XStpo9pYPHyAT8W3HgZnxg1QjTNcMbcHZ6cqrj2HnZ/DbR9RP6+TuF13j28YVOq09naIGEVSePnwo87wfwrgTgB/6nHwLICvA/gcgE8CuJeIPhnxuAZDA8NDBey8a4VSNoYbhGq5fdxEVaSMq4mJTOSvr1Zl6qYY6fwccXdqiyqv4Uc+Z9VXTypjaUci+fCZ+TUAIO+b7FMAXmfmN2vbfgfAHQBS09fW0BkMDxVCNT4XPV1FDnnc2A1KUH98XIV8YR4kzvd0P4yc52bq0nSsmVjnLlSUJU9EbKbdSMKHXwDwru3392qvuUJEDxDRYSI6fPr06dgHZ+gswhi+v7k4jW17joUq6gqzohB/E8YfH1dzHC9pC9V2fjrb/rmdmzgqk+0E8W6nWerDC98ZPhH9EMBHXd76CjN/T/eAmPkJAE8AVR++7v0b0omuzBO3zBG7L9qNmVmW6sP7odJsQ/Y3Xi4Q2WePK8XTL+NGJRsnTNaO+N6LpXJdv95Lxz4tpKWyOii+Bp+ZfzXiMYoArrH9/rHaawYDgOAl+l4PBy+DKCsUShqhvBjWBRJHcxyVB4nfQybow8j5vQsjn3Zjn8Z0YVWSyMP/CYBPENF1qBr6LwL4tQSOa2gTgsx0VR4OMoPolcfd32fhQmU2Ea0eP0G4Vs0evR4kcTxkwrjRiForLZ7W6lpVIhl8IvoCgD8AsBDAPiKaYOa1RHQ1qumXtzLzNBF9GcABVNMyv8XM+nVhDW1LkJluGDeIYPPaJU0SB4J111+FVdcuwIO7J7XMMGV67P19lqfAVytmj6qCZCr7UV2pjY0XQ7nRWm3so6S+poFIQVtm/i4zf4yZ5zDzR5h5be31U8x8q227/cz8d5l5MTP/TtRBGzqLIMG+KJkgw0MF7Lx7hWv65ZOHTmLj6ARmXSyKVxGOHRG+LeRzuG/1QNPfWVkCM+pFSQCaBOHm9GSwaXQicNFS2I5aQudHJkgWpKgrSFpmq1tEBg21t7Mbx46ptDW0nCCZJ1EzQYaHCujtkRtwtwnk+pWFut9dRj5nYdeGQbxd65DmbMrR32cBjCZjClQ/f77PCm1so1TfbttzzHXFIyhXZrBR4QE0Nl6Uusvs1bl+2yZBlsgzllPI53D/6oGOVLg1apmGVKCapaOjmjNotabIGvFqmyd8u7LPELYhRz5nYd6cHs/z4ld963Vug5wL2Xl2+05kf79+ZSGQmFkUrExV8tjJ/asHcPD46VgqltOAV6WtMfiGtiNqCmdc5fk5K+uaDlqaqmjLDnIzul4yxbs2DEofkACwMWChmptBDPIwS0PKpXg4xy0D0SqMPLKho4iaMdJfc5/oxk2aQPdx3ALUecnnyfdZrpkw5coMtu89hgshRMBOlcpND9wgK5dWG3ug+hmSlqxOC2aGb+g6xsaL+K3dE1BUOU4lBNSN1LY9x1wzXnJWRruyY87K4EJltmFF4eXqcpKWGX67u228MDN8Q1cic/2IWVxQd0aasAdnZf7wOGR83fbJaDb6VpYARkNAOGdlccPAfLz0xhnt41JFtfK3U2f9JkvH0JH4Za608iZ26szbCZouWK7MhNLz0d1P127s+/ss7LxrBXbevaIp0+XYqb/RetwgEKoZV36Vv3H0GkgLxuAbOhKVnPCwUsoqyPbcW2uPZ2/Yks9ZeHzDoKsMsQozzK5prUk1hXEiYgPDQwW8NHIzdm0YxPmL09g4OhFas8gLQlXH3q9eggEcPC4XZIxb3jkNGJeOoSNRKdCKy5dsZQmfWtTv6rq4NMM4/M4ZjD98S9N7YaSdAXlKKNAseiaIw/AK7IFltwbuumFAuUI6iNyzyt+0G2aGb+hI/Aq0vJbpBPXqWjcqM+zpp3765ctq4fYK2UyIFYeXT9regtB7H+5mwMpQaNePMJI7D5wIZOzDnnfVh7dXgZ5Oeee0Ygy+oSPxq971WqYzGiUPdPu7hXHaMnYUm0Yn6j5jFaM1rzfb5Bc//M6Zhv3Yfc/DQwWsWbrQc58XKrN4fMNg3a0kPvMVc3tQKlek7qkskfRhIYxkkNnxvN4s5vTEZ5KsDHkGbOPqNZAmjEvH0JH45Vl7GaJCPteQzTM2XsSm3RPahLuyRBgbL+KpQycD+eytLOF3vrC8YVyylEy7W8W+onCDUT1PwrAxquJpbHvfLfVyhhnlirv2kJ8iqJNshnBpehbnL8VYgeuzgOqG3Hxj8A0di1eBlswQEdAwoxOZGzrd/ffeeA12HjgROEB7xZzLt6uKnIF4qKmsHIqlckMTeOdf+O1BPBCc8sGb1y7xTX8t5HM4f3FaW1yBAGRc8v0rM+yrqhqHDHSaMC4dQ1fitnwnAPetHmhq+qFT96UnQ1h17YJQgcCzU5W6q0ZlXMKtopKNlCHUjX0YhLF/aeRmHH7nDBY/tB+LRvbhwd2Tnn/3+IZBvDRys6+x780S7ndRIHVj14ZBV9VToLMCsGEwBt/QlYiApj11ca6Vwb5X32+QGNZtIKZnGZtGJwLn2wuEq8ZvXPaVyr03XuO5LQAtVcfFUhlbxo7iyUMnlbtXbd9bbY3h91C6VHsYqQShH3r+KOZL4i6dFIANgzH4hq7GridTrszi7FSjPHE+hlx2BhClBrZY8y/LcK5UHh1ejvtXD0Q4ohpZIt94gZOzUxWMjReV3E5PHToJAHhp5GbPvPtyZQZEzRk/nRaADUMkg09EdxPRMSKaJSJX7Ybadm8T0VEimiAiI45jSAV+bpFyZUbJd1/I53xnnbpZ9GH58T69eAEOHj/dsFJ5dHh56GyjjOJyZIY5VG3DzgMnlM6fCC4Dl1doMkpTlYZMq07StI9C1KDtTwHcCeAPFbZdw8x/FfF4BoM2VNw158oVT3VN+6xx0+hEYk3SvfL87e+Jlcrhd87g/KXpUMdSdfeEFUY7VSpj14ZBJW0j+3c2PFTAzgMnpH2BOz0AG4aoLQ5fY+bOqTs2dBUq/tyr8zlsvW2Zq/sgn7Pqs8bhoQLuS8BtEoZyZQZPv/xupKCsCmErl4VxVpGCcH5n3ZA7r5OkfPgM4AUiOkJED3htSEQPENFhIjp8+rRc98JgiIqbsbAjDIe9YlW4Bx7fMIiJrbc0zCCFr9zpAVF1icRJnJLEUTSJ7MVQsgerwM2Qu303xnUjx1cPn4h+COCjLm99hZm/V9vmRwD+FTO7+ueJqMDMRSL62wB+AOD/ZuY/9Ruc0cM3xI1dDnd+zgJR1f8bpejGKbG76MO5lkoCx0khn8OpWoVvUPI5C9tuX9ZwjuP4PrqN2Fsc+hl8x7bbAHzAzP+P37bG4BvanbHxYiTffs7K4NIMYyaF3VqyGcJX714h9aN70elNSFqJl8GP3aVDRPOI6EPiZwC3oBrsNRg6njAVtf19Vbnkt3esw2v/7nP4qk1XPk5J56CIh5Cfa8yJ8bG3jkhZOkT0BQB/AGAhgH1ENMHMa4noagDfZOZbAXwEwHepeqH2APhvzPzfI47bYGgLghRuOWUJBCIo7KWd0yo2jk4gn7OUO1n191nYetsyLa4y4+IJTiSDz8zfBfBdl9dPAbi19vObAFZEOY7B0K4EafLt5eJQ0c5pFaVyxdfYyx5mqjg/v0g3BVrbvazdMOJpBkOMbF67RMlQE1CXMwaaZ7PnL04nYuyDNCRXIWdltWTNeHWjMgZfHWPwDYYYcUru5iVFXIyqe2TngRNYs3QhnjtSbJjNxo1wtYTtuiVDV4pkN3SjSgJj8A2GmHFWfC4a2SfdtlgqB9bJj0p/n1VvuRgm40aGzgCzzDXW7WJoQTHiaQZDwvjpxsRh7GWdqYBqnrsgaMaNFzPMdTnnqJiKWj0Yg28waMbep1aIl9nRaVRVuVCZlT5o7LNkUbkqm51nieotEPv7rHp16/2rB1z/RvjZo2IqavVgXDoGg0bcskk2jU5g4+hEU6aKl/tEJXiaszK4NK2mUDk/Z+H8xWbxNJlcAYCmYLNfAFbIFzvR5Wc3YmjRMTN8g0Ejbtkkwhw7m4vLdN1zVhb3rR6oz2b7+yxYDkGeqvG9Hl+9Z4VSM5Xzl5pbCPb3WVIDHmZGLfOnGz97ejAzfINBI36zWWcqoWrjbLeiIwB4cPek70qgz8pgqtLccqWvt0dqwMMUObmloBo/e7owBt9g0IhKoZXzoaDiqnBuI1xHXu4cIU4mS7WUPZzCFjmpPrwMrcMYfINBIyqFVjpcHH7duuziZF5NQtzYtudY6CIn42dPN8aHbzBoxO77BtDkX9fl4vByHTmPESSlcWy8KNXqCRp89ctWMiSPmeEbDJqxz3LjEvySuY6yRE3B1SBxggd3T3oeUxWjfZNOtOjhx4XRwzcY3HETU4uiW6Mizvb4hkHlfd+040XXB5LRwY8fLz18M8M3GNoQ3QFSv5iA2MZ+bC+M9k06MQbfYGhTdAZIVQxxELeM0b5JJyZoazAYlA2xqlSC0b5JJ5EMPhHtJKLjRPQqEX2XiPKS7T5LRCeI6HUiGolyTIPBoJ8g+j4qqwGjfZNOorp0fgDgIWaeJqLfBfAQgN+2b0BEWQBfB/CPALwH4CdEtIeZfxbx2AaDQRNuMYGpS9Ou2v2qqwGTk58+orY4fMH26yEAd7ls9ikAr9daHYKIvgPgDgDG4BsMKUJWzWukEjoHnT78/xPAn7i8XgDwru3392qvuUJEDxDRYSI6fPr0aY3DMxgMQTBumc7Dd4ZPRD8E8FGXt77CzN+rbfMVANMAnoo6IGZ+AsATQDUPP+r+DAZDeIxbprPwNfjM/Kte7xPRlwB8HsCvsHsVVxHANbbfP1Z7zWAwGAwJEjVL57MA/jWA25l5SrLZTwB8goiuI6JeAF8EsCfKcQ0Gg8EQnKg+/K8B+BCAHxDRBBF9AwCI6Goi2g8AzDwN4MsADgB4DcBuZj4W8bgGg8FgCEjULJ2/I3n9FIBbbb/vB7A/yrEMBoPBEA1TaWswGAxdQqrVMonoNIB3NO/2SgB/pXmfOkjjuNI4JsCMKwhpHBNgxhWEoGO6lpkXur2RaoMfB0R0WCYd2krSOK40jgkw4wpCGscEmHEFQeeYjEvHYDAYugRj8A0Gg6FL6EaD/0SrByAhjeNK45gAM64gpHFMgBlXELSNqet8+AaDwdCtdOMM32AwGLoSY/ANBoOhS+h4g5/WrlxEdDcRHSOiWSKSplwR0dtEdLQmXXE4JWNK+lwtIKIfENGf1/7vl2w3UztPE0QUi16T32cnojlENFp7/2UiWhTHOEKM60tEdNp2fv5pAmP6FhH9nIh+KnmfiOj3a2N+lYhuiHtMiuP6DBGds52rhxMY0zVEdJCIfla7B/+lyzbRzxczd/Q/ALcA6Kn9/LsAftdlmyyANwB8HEAvgEkAn4x5XL8EYAmAHwFY5bHd2wCuTOhc+Y6pRefq9wCM1H4ecfsOa+99EPM4fD87gP8LwDdqP38RwGgC35vKuL4E4GtJXEe2Y/5DADcA+Knk/VtR7aFBAFYDeDkl4/oMgO8nfK6uAnBD7ecPAfhfLt9h5PPV8TN8Zn6BqwJuQLUr18dcNqt35WLmSwBEV644x/UaM/t3g04QxTElfq5q+/927edvAxiO+XgyVD67fazPAvgVIqIUjCtxmPlPAZzx2OQOAH/MVQ4ByBPRVSkYV+Iw8/vM/Ert579BVWjS2Ygg8vnqeIPvQEtXroRhAC8Q0REieqDVg0FrztVHmPn92s9/AeAjku3m1rqlHSKi4RjGofLZ69vUJhrnAHw4hrEEHRcArK+5Ap4lomtc3k+aNN93f5+IJonoT4hoWZIHrrkBhwC87Hgr8vmK2sQ8FSTdlUvnuBT4B8xcJKK/jaoM9fHaDKWVY9KO17jsvzAzE5Esl/ja2rn6OIAXiegoM7+he6xtyl4ATzPzRSL6Z6iuQm5u8ZjSyiuoXksfENGtAMYAfCKJAxPRFQCeA7CRmf9a9/47wuBzSrty+Y1LcR/F2v8/J6Lvorp8D23wNYwp8XNFRH9JRFcx8/u1JezPJfsQ5+pNIvoRqrMknQZf5bOLbd4joh4A8wH8QuMYQo2Lme1j+CaqcZFWk8pueHZDy8z7iej/JaIrmTlWUTUislA19k8x8/Mum0Q+Xx3v0qE27spFRPOI6EPiZ1QD0K6ZBQnSinO1B8Cv137+dQBNKxEi6ieiObWfrwRwE4CfaR6Hyme3j/UuAC9KJhmJjsvh670dVR9xq9kD4B/Xsk9WAzhnc921DCL6qIi7ENGnULWTsT60a8f7IwCvMfN/kGwW/XwlGYluxT8Ar6Pq95qo/RMZFFcD2G/b7lZUI+NvoOreiHtcX0DVB3cRwF8COOAcF6pZF5O1f8fiHpfKmFp0rj4M4H8A+HMAPwSwoPb6KgDfrP38aQBHa+fqKIDfiGksTZ8dwCOoTigAYC6AZ2rX3Z8B+Hjc50dxXI/VrqFJAAcBLE1gTE8DeB9ApXZd/QaA3wTwm7X3CcDXa2M+Co9stYTH9WXbuToE4NMJjOkfoBqve9Vmq27Vfb6MtILBYDB0CR3v0jEYDAZDFWPwDQaDoUswBt9gMBi6BGPwDQaDoUswBt9gMBi6BGPwDQaDoUswBt9gMBi6hP8N0zJ5I05IlM8AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "X = create_twomoon_dataset()\n",
    "# You can load your own dataset as below in numpy format\n",
    "# path = Path(args.data_dir, args.filename)\n",
    "# X = np.load(path)\n",
    "# print('Data shape: ', X.shape)\n",
    "\n",
    "dataset = data.TensorDataset(torch.Tensor(X))\n",
    "loader = torch.utils.data.DataLoader(dataset, batch_size=cfg.batch_size, shuffle=True, drop_last=True)\n",
    "cfg.input_dim = X.shape[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "accepted-diversity",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1\\300, loss: -0.000, ls loss: -0.00069, recon loss: 0.961\n",
      "Selection probs: \n",
      "  [0.08895258 0.09500632 0.08862478 0.08452889 0.09643977 0.08996537\n",
      " 0.09198611 0.08099049 0.09697609 0.09512109 0.09342799 0.08812779] \n",
      "\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 6\\300, loss: 0.000, ls loss: -0.00085, recon loss: 0.834\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 11\\300, loss: -0.000, ls loss: -0.00107, recon loss: 0.829\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 16\\300, loss: -0.000, ls loss: -0.00141, recon loss: 0.830\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 21\\300, loss: -0.000, ls loss: -0.00193, recon loss: 0.828\n",
      "Selection probs: \n",
      "  [0.5231066  0.49032244 0.05844297 0.06868261 0.12676029 0.05727994\n",
      " 0.06390956 0.08747183 0.11194766 0.15766361 0.10915606 0.14120847] \n",
      "\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 26\\300, loss: 0.000, ls loss: -0.00280, recon loss: 0.822\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 31\\300, loss: -0.000, ls loss: -0.00674, recon loss: 0.822\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 36\\300, loss: -0.000, ls loss: -0.02289, recon loss: 0.811\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 41\\300, loss: -0.000, ls loss: -0.05022, recon loss: 0.814\n",
      "Selection probs: \n",
      "  [9.9994719e-01 9.9994564e-01 2.8626307e-06 3.5599148e-06 1.8559253e-05\n",
      " 2.0758216e-06 2.7501924e-06 5.8499304e-06 1.8195131e-05 2.1477214e-05\n",
      " 9.5566738e-06 2.2326019e-05] \n",
      "\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 46\\300, loss: -0.000, ls loss: -0.09505, recon loss: 0.817\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 51\\300, loss: -0.000, ls loss: -0.14985, recon loss: 0.821\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 56\\300, loss: -0.000, ls loss: -0.21413, recon loss: 0.824\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 61\\300, loss: -0.000, ls loss: -0.29676, recon loss: 0.826\n",
      "Selection probs: \n",
      "  [1.0000000e+00 1.0000000e+00 1.2572693e-09 1.5553744e-09 1.0513823e-08\n",
      " 6.1067196e-10 9.6102237e-10 1.8712045e-09 1.0610825e-08 8.9010719e-09\n",
      " 4.6905440e-09 1.0042603e-08] \n",
      "\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 66\\300, loss: 0.000, ls loss: -0.36053, recon loss: 0.830\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 71\\300, loss: 0.000, ls loss: -0.43887, recon loss: 0.830\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 76\\300, loss: -0.000, ls loss: -0.52814, recon loss: 0.833\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 81\\300, loss: -0.000, ls loss: -0.59740, recon loss: 0.832\n",
      "Selection probs: \n",
      "  [1.0000000e+00 1.0000000e+00 1.6753903e-12 2.0609616e-12 1.4637262e-11\n",
      " 1.0140303e-12 1.7439189e-12 3.1174399e-12 1.5125221e-11 1.6348867e-11\n",
      " 6.5527093e-12 1.8924222e-11] \n",
      "\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 86\\300, loss: -0.000, ls loss: -0.65538, recon loss: 0.837\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 91\\300, loss: 0.000, ls loss: -0.71276, recon loss: 0.833\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "Epoch 96\\300, loss: 0.000, ls loss: -0.74690, recon loss: 0.834\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0010000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 101\\300, loss: 0.000, ls loss: -0.78312, recon loss: 0.827\n",
      "Selection probs: \n",
      "  [1.00000000e+00 1.00000000e+00 7.51085883e-15 9.36008168e-15\n",
      " 7.02029513e-14 5.07903910e-15 9.29057161e-15 1.58797570e-14\n",
      " 7.33698516e-14 9.13632234e-14 3.12543231e-14 1.06493014e-13] \n",
      "\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 106\\300, loss: 0.000, ls loss: -0.82260, recon loss: 0.826\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 111\\300, loss: 0.000, ls loss: -0.84325, recon loss: 0.828\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 116\\300, loss: -0.000, ls loss: -0.86038, recon loss: 0.828\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 121\\300, loss: 0.000, ls loss: -0.86533, recon loss: 0.831\n",
      "Selection probs: \n",
      "  [1.0000000e+00 1.0000000e+00 1.7292420e-16 2.2049444e-16 2.0405953e-15\n",
      " 1.1309326e-16 2.2098788e-16 3.9805636e-16 2.1436710e-15 2.7553751e-15\n",
      " 8.3586598e-16 3.2625791e-15] \n",
      "\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 126\\300, loss: 0.000, ls loss: -0.88300, recon loss: 0.829\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 131\\300, loss: -0.000, ls loss: -0.89127, recon loss: 0.831\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 136\\300, loss: -0.000, ls loss: -0.90565, recon loss: 0.828\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 141\\300, loss: 0.000, ls loss: -0.89796, recon loss: 0.831\n",
      "Selection probs: \n",
      "  [1.0000000e+00 1.0000000e+00 1.7344925e-18 2.2772789e-18 2.7555615e-17\n",
      " 1.0802542e-18 2.2921299e-18 4.4247574e-18 2.9133079e-17 3.8705343e-17\n",
      " 1.0144920e-17 4.6766519e-17] \n",
      "\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 146\\300, loss: -0.000, ls loss: -0.91300, recon loss: 0.829\n",
      "LS-CAE learning rate = 0.0001000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 151\\300, loss: 0.000, ls loss: -0.90490, recon loss: 0.827\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 156\\300, loss: 0.000, ls loss: -0.91608, recon loss: 0.827\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 161\\300, loss: 0.000, ls loss: -0.90988, recon loss: 0.828\n",
      "Selection probs: \n",
      "  [1.0000000e+00 1.0000000e+00 4.9947520e-21 6.8147873e-21 1.1716476e-19\n",
      " 2.9117462e-21 6.8731633e-21 1.4550488e-20 1.2486166e-19 1.7282305e-19\n",
      " 3.7484833e-20 2.1443864e-19] \n",
      "\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 166\\300, loss: 0.000, ls loss: -0.90140, recon loss: 0.828\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 171\\300, loss: 0.000, ls loss: -0.91433, recon loss: 0.827\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 176\\300, loss: 0.000, ls loss: -0.92185, recon loss: 0.830\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 181\\300, loss: -0.000, ls loss: -0.89798, recon loss: 0.827\n",
      "Selection probs: \n",
      "  [1.0000000e+00 1.0000000e+00 2.1003073e-24 3.0173555e-24 8.3177576e-23\n",
      " 1.1195659e-24 3.0479659e-24 7.3075081e-24 8.9581334e-23 1.3087865e-22\n",
      " 2.2026751e-23 1.6830814e-22] \n",
      "\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 186\\300, loss: -0.000, ls loss: -0.90027, recon loss: 0.828\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 191\\300, loss: 0.000, ls loss: -0.90166, recon loss: 0.829\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "Epoch 196\\300, loss: -0.000, ls loss: -0.91298, recon loss: 0.825\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0001000\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 201\\300, loss: 0.000, ls loss: -0.89998, recon loss: 0.829\n",
      "Selection probs: \n",
      "  [1.0000000e+00 1.0000000e+00 3.9894911e-29 6.1607803e-29 3.2900295e-27\n",
      " 1.8758667e-29 6.2358277e-29 1.7798135e-28 3.5961195e-27 5.6664969e-27\n",
      " 6.6848913e-28 7.6617846e-27] \n",
      "\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 206\\300, loss: -0.000, ls loss: -0.89347, recon loss: 0.825\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 211\\300, loss: 0.000, ls loss: -0.91189, recon loss: 0.826\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 216\\300, loss: 0.000, ls loss: -0.90877, recon loss: 0.827\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 221\\300, loss: -0.000, ls loss: -0.90734, recon loss: 0.824\n",
      "Selection probs: \n",
      "  [1.0000000e+00 1.0000000e+00 3.3705026e-36 5.7998514e-36 8.3417653e-34\n",
      " 1.3132784e-36 5.8882201e-36 2.1822546e-35 9.3221128e-34 1.6450435e-33\n",
      " 1.1396485e-34 2.3978744e-33] \n",
      "\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 226\\300, loss: -0.000, ls loss: -0.88589, recon loss: 0.829\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 231\\300, loss: 0.000, ls loss: -0.90065, recon loss: 0.828\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 236\\300, loss: -0.000, ls loss: -0.91296, recon loss: 0.827\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 241\\300, loss: 0.000, ls loss: -0.90646, recon loss: 0.825\n",
      "Selection probs: \n",
      "  [1.0e+00 1.0e+00 0.0e+00 0.0e+00 8.4e-45 0.0e+00 0.0e+00 0.0e+00 9.8e-45\n",
      " 2.2e-44 0.0e+00 3.6e-44] \n",
      "\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 246\\300, loss: -0.000, ls loss: -0.90234, recon loss: 0.826\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 251\\300, loss: 0.000, ls loss: -0.91426, recon loss: 0.825\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 256\\300, loss: -0.000, ls loss: -0.91326, recon loss: 0.827\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 261\\300, loss: -0.000, ls loss: -0.91019, recon loss: 0.824\n",
      "Selection probs: \n",
      "  [1. 1. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.] \n",
      "\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 266\\300, loss: 0.000, ls loss: -0.89116, recon loss: 0.826\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 271\\300, loss: -0.000, ls loss: -0.91123, recon loss: 0.825\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 276\\300, loss: 0.000, ls loss: -0.90861, recon loss: 0.825\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 281\\300, loss: 0.000, ls loss: -0.90086, recon loss: 0.829\n",
      "Selection probs: \n",
      "  [1. 1. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.] \n",
      "\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 286\\300, loss: 0.000, ls loss: -0.90354, recon loss: 0.825\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 291\\300, loss: 0.000, ls loss: -0.90261, recon loss: 0.827\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Epoch 296\\300, loss: 0.000, ls loss: -0.91290, recon loss: 0.828\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "LS-CAE learning rate = 0.0000100\n",
      "Finished training LS-CAE\n",
      "Selected features: [0 1]\n"
     ]
    }
   ],
   "source": [
    "lscae_model = lscae.Lscae(kwargs=cfg)\n",
    "selected_features = lscae_model.select_features(loader)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "adverse-audience",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "passing-system",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}