RecurrentAutoencoderHexamodal.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Populating the interactive namespace from numpy and matplotlib\n",
      "Couldn't import dot_parser, loading of dot files will not be possible.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using gpu device 0: GeForce GTX880M (CNMeM is disabled, cuDNN 5103)\n",
      "/usr/local/lib/python2.7/dist-packages/theano/sandbox/cuda/__init__.py:599: UserWarning: Your cuDNN version is more recent than the one Theano officially supports. If you see any problems, try updating Theano or downgrading cuDNN to version 5.\n",
      "  warnings.warn(warn)\n"
     ]
    }
   ],
   "source": [
    "%pylab inline\n",
    "\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import theano\n",
    "import theano.tensor as T\n",
    "import lasagne\n",
    "from math import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "\n",
    "def sample(N):\n",
    "    z1 = np.random.randint(3,size=N)\n",
    "    z2 = np.random.randn(N,2)\n",
    "    z3 = np.random.randint(2,size=N)+1\n",
    "    \n",
    "    x = np.array([(z3+0.2*z2[:,1])*(np.sin(2*3.14159*z1/3.0+0.2*z3 + z3*z2[:,0]*0.1)), (z3+0.2*z2[:,1])*(np.cos(2*3.14159*z1/3.0+0.2*z3 + z3*z2[:,0]*0.1))]).T\n",
    "    y = 1.0/(1.0+np.exp(-1.5*x))\n",
    "    return y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAW8AAAD7CAYAAAClvBX1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7985c41b10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "data = sample(100000)\n",
    "plt.hist2d(data[:,0],data[:,1],40)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "invar = T.matrix()\n",
    "\n",
    "TEMP = 0.4\n",
    "TEMP2 = 0.0\n",
    "HIDDEN = 256\n",
    "\n",
    "input = lasagne.layers.InputLayer((None,2),invar)\n",
    "p1_dense1 = lasagne.layers.DenseLayer(input, num_units = HIDDEN)\n",
    "p1_dense2 = lasagne.layers.DenseLayer(p1_dense1, num_units = HIDDEN, nonlinearity = lasagne.nonlinearities.sigmoid)\n",
    "p1_noisify = lasagne.layers.GaussianNoiseLayer(p1_dense2, sigma = TEMP)\n",
    "p1_dense3 = lasagne.layers.DenseLayer(p1_noisify, num_units = HIDDEN)\n",
    "p1_out = lasagne.layers.DenseLayer(p1_dense3, num_units=2, nonlinearity = lasagne.nonlinearities.sigmoid)\n",
    "p1_outnoise = lasagne.layers.GaussianNoiseLayer(p1_out, sigma = TEMP2)\n",
    "\n",
    "# Now repeat...\n",
    "p2_dense1 = lasagne.layers.DenseLayer(p1_outnoise, num_units = HIDDEN, W=p1_dense1.W, b=p1_dense1.b)\n",
    "p2_dense2 = lasagne.layers.DenseLayer(p2_dense1, num_units = HIDDEN, W=p1_dense2.W, b=p1_dense2.b, nonlinearity = lasagne.nonlinearities.sigmoid)\n",
    "p2_noisify = lasagne.layers.GaussianNoiseLayer(p2_dense2, sigma = TEMP)\n",
    "p2_dense3 = lasagne.layers.DenseLayer(p2_noisify, num_units = HIDDEN, W=p1_dense3.W, b=p1_dense3.b)\n",
    "p2_out = lasagne.layers.DenseLayer(p2_dense3, num_units=2, nonlinearity = lasagne.nonlinearities.sigmoid, W=p1_out.W, b=p1_out.b)\n",
    "p2_outnoise = lasagne.layers.GaussianNoiseLayer(p2_out, sigma = TEMP2)\n",
    "\n",
    "# Now repeat...\n",
    "p3_dense1 = lasagne.layers.DenseLayer(p2_outnoise, num_units = HIDDEN, W=p1_dense1.W, b=p1_dense1.b)\n",
    "p3_dense2 = lasagne.layers.DenseLayer(p3_dense1, num_units = HIDDEN, W=p1_dense2.W, b=p1_dense2.b, nonlinearity = lasagne.nonlinearities.sigmoid)\n",
    "p3_noisify = lasagne.layers.GaussianNoiseLayer(p3_dense2, sigma = TEMP)\n",
    "p3_dense3 = lasagne.layers.DenseLayer(p3_noisify, num_units = HIDDEN, W=p1_dense3.W, b=p1_dense3.b)\n",
    "p3_out = lasagne.layers.DenseLayer(p3_dense3, num_units=2, nonlinearity = lasagne.nonlinearities.sigmoid, W=p1_out.W, b=p1_out.b)\n",
    "p3_outnoise = lasagne.layers.GaussianNoiseLayer(p3_out, sigma = TEMP2)\n",
    "\n",
    "# Now repeat...\n",
    "p4_dense1 = lasagne.layers.DenseLayer(p3_outnoise, num_units = HIDDEN, W=p1_dense1.W, b=p1_dense1.b)\n",
    "p4_dense2 = lasagne.layers.DenseLayer(p4_dense1, num_units = HIDDEN, W=p1_dense2.W, b=p1_dense2.b, nonlinearity = lasagne.nonlinearities.sigmoid)\n",
    "p4_noisify = lasagne.layers.GaussianNoiseLayer(p4_dense2, sigma = TEMP)\n",
    "p4_dense3 = lasagne.layers.DenseLayer(p4_noisify, num_units = HIDDEN, W=p1_dense3.W, b=p1_dense3.b)\n",
    "p4_out = lasagne.layers.DenseLayer(p4_dense3, num_units=2, nonlinearity = lasagne.nonlinearities.sigmoid, W=p1_out.W, b=p1_out.b)\n",
    "p4_outnoise = lasagne.layers.GaussianNoiseLayer(p4_out, sigma = TEMP2)\n",
    "\n",
    "# Now repeat...\n",
    "p5_dense1 = lasagne.layers.DenseLayer(p4_outnoise, num_units = HIDDEN, W=p1_dense1.W, b=p1_dense1.b)\n",
    "p5_dense2 = lasagne.layers.DenseLayer(p5_dense1, num_units = HIDDEN, W=p1_dense2.W, b=p1_dense2.b, nonlinearity = lasagne.nonlinearities.sigmoid)\n",
    "p5_noisify = lasagne.layers.GaussianNoiseLayer(p5_dense2, sigma = TEMP)\n",
    "p5_dense3 = lasagne.layers.DenseLayer(p5_noisify, num_units = HIDDEN, W=p1_dense3.W, b=p1_dense3.b)\n",
    "p5_out = lasagne.layers.DenseLayer(p5_dense3, num_units=2, nonlinearity = lasagne.nonlinearities.sigmoid, W=p1_out.W, b=p1_out.b)\n",
    "p5_outnoise = lasagne.layers.GaussianNoiseLayer(p5_out, sigma = TEMP2)\n",
    "\n",
    "# Now repeat...\n",
    "p6_dense1 = lasagne.layers.DenseLayer(p5_outnoise, num_units = HIDDEN, W=p1_dense1.W, b=p1_dense1.b)\n",
    "p6_dense2 = lasagne.layers.DenseLayer(p6_dense1, num_units = HIDDEN, W=p1_dense2.W, b=p1_dense2.b, nonlinearity = lasagne.nonlinearities.sigmoid)\n",
    "p6_noisify = lasagne.layers.GaussianNoiseLayer(p6_dense2, sigma = TEMP)\n",
    "p6_dense3 = lasagne.layers.DenseLayer(p6_noisify, num_units = HIDDEN, W=p1_dense3.W, b=p1_dense3.b)\n",
    "p6_out = lasagne.layers.DenseLayer(p6_dense3, num_units=2, nonlinearity = lasagne.nonlinearities.sigmoid, W=p1_out.W, b=p1_out.b)\n",
    "p6_outnoise = lasagne.layers.GaussianNoiseLayer(p6_out, sigma = TEMP2)\n",
    "\n",
    "# Now repeat...\n",
    "p7_dense1 = lasagne.layers.DenseLayer(p6_outnoise, num_units = HIDDEN, W=p1_dense1.W, b=p1_dense1.b)\n",
    "p7_dense2 = lasagne.layers.DenseLayer(p7_dense1, num_units = HIDDEN, W=p1_dense2.W, b=p1_dense2.b, nonlinearity = lasagne.nonlinearities.sigmoid)\n",
    "p7_noisify = lasagne.layers.GaussianNoiseLayer(p7_dense2, sigma = TEMP)\n",
    "p7_dense3 = lasagne.layers.DenseLayer(p7_noisify, num_units = HIDDEN, W=p1_dense3.W, b=p1_dense3.b)\n",
    "p7_out = lasagne.layers.DenseLayer(p7_dense3, num_units=2, nonlinearity = lasagne.nonlinearities.sigmoid, W=p1_out.W, b=p1_out.b)\n",
    "\n",
    "out1,out2,out3,out4,out5,out6,out7 = lasagne.layers.get_output([p1_out,p2_out,p3_out,p4_out,p5_out,p6_out,p7_out])\n",
    "\n",
    "loss = T.mean( (out1-invar)**2 + (out2-invar)**2 + (out3-invar)**2 + (out4-invar)**2 + (out5-invar)**2 + (out6-invar)**2 + (out7-invar)**2)\n",
    "params = lasagne.layers.get_all_params(p7_out,trainable=True)\n",
    "updates = lasagne.updates.adam(loss,params,learning_rate = 1e-3)\n",
    "\n",
    "train = theano.function([invar], loss, updates=updates, allow_input_downcast = True)\n",
    "project = theano.function([invar], out7, allow_input_downcast = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.797172904015\n",
      "0.376685261726\n",
      "0.330215841532\n",
      "0.0788879096508\n",
      "0.0630481690168\n",
      "0.0597122460604\n",
      "0.0560206323862\n",
      "0.0497223995626\n",
      "0.0474856905639\n",
      "0.047755997628\n",
      "0.0415515452623\n",
      "0.0433428660035\n",
      "0.0338953062892\n",
      "0.0355174131691\n",
      "0.0318645462394\n",
      "0.0319139994681\n",
      "0.0274570751935\n",
      "0.0296505186707\n",
      "0.0262829530984\n",
      "0.0260078329593\n",
      "0.0264870841056\n",
      "0.0237705167383\n",
      "0.0237536150962\n",
      "0.0257007870823\n",
      "0.0232343785465\n",
      "0.0219830330461\n",
      "0.0220802780241\n",
      "0.0233245864511\n",
      "0.0196057856083\n",
      "0.0213776249439\n",
      "0.0200257394463\n",
      "0.0185995455831\n",
      "0.019552256912\n",
      "0.0197945032269\n",
      "0.0174520798028\n",
      "0.0189339984208\n",
      "0.0176494531333\n",
      "0.0173293892294\n",
      "0.0166314132512\n",
      "0.0152208674699\n",
      "0.0161750726402\n",
      "0.0162981338799\n",
      "0.0142134828493\n",
      "0.014818216674\n",
      "0.0138083994389\n",
      "0.0154497595504\n",
      "0.0140280453488\n",
      "0.0159476976842\n",
      "0.0132405646145\n",
      "0.0123723652214\n",
      "0.0126754622906\n",
      "0.0122866230085\n",
      "0.0114721413702\n",
      "0.0120979528874\n",
      "0.0133304968476\n",
      "0.0122421383858\n",
      "0.0105509553105\n",
      "0.0100729474798\n",
      "0.0103964926675\n",
      "0.00956105999649\n",
      "0.0107394456863\n",
      "0.0101858833805\n",
      "0.0112659847364\n",
      "0.0101041719317\n",
      "0.0109075140208\n",
      "0.00955528020859\n",
      "0.00978027191013\n",
      "0.00915901269764\n",
      "0.00912770535797\n",
      "0.00889035500586\n",
      "0.00847449153662\n",
      "0.00929876789451\n",
      "0.00854007340968\n",
      "0.00834699161351\n",
      "0.00861335452646\n",
      "0.0089213848114\n",
      "0.00843678507954\n",
      "0.00803559366614\n",
      "0.00845359824598\n",
      "0.00831267982721\n",
      "0.00803651660681\n",
      "0.00786259584129\n",
      "0.00868070870638\n",
      "0.00775858340785\n",
      "0.00782014429569\n",
      "0.00793773494661\n",
      "0.00809574779123\n",
      "0.00766344787553\n",
      "0.00732969911769\n",
      "0.00775694055483\n",
      "0.00730093801394\n",
      "0.00685981987044\n",
      "0.00687608169392\n",
      "0.00767840677872\n",
      "0.00779128959402\n",
      "0.00679909437895\n",
      "0.0070042964071\n",
      "0.00788304209709\n",
      "0.00661312509328\n",
      "0.00661308784038\n"
     ]
    }
   ],
   "source": [
    "#data = sample(1000)\n",
    "\n",
    "for epoch in range(10000):\n",
    "    data = sample(1000)\n",
    "    err = train(data)\n",
    "    if epoch%100 == 0:\n",
    "        print err"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "fakes = np.random.rand(100000,2)\n",
    "pfakes = project(fakes)\n",
    "for i in range(10):\n",
    "    pfakes = project(pfakes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAW8AAAD7CAYAAAClvBX1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSIjwBoCECG8ASIjwBoCE/h8FLmSFKpvMKwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f796fc09750>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAW8AAAD7CAYAAAClvBX1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f796fbaccd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist2d(pfakes[:,0],pfakes[:,1],40)\n",
    "plt.show()\n",
    "data = sample(1000000)\n",
    "plt.hist2d(data[:,0],data[:,1],40)\n",
    "plt.plot(pfakes[::50,0],pfakes[::50,1],'r.')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}