Theano_scan_loop_and_ite.ipynb

{
 "metadata": {
  "name": "",
  "signature": "sha256:9c91ec5b7622299c143db260ad3a67d14aa971fe254744898714186ad2bb510d"
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "heading",
     "level": 1,
     "metadata": {},
     "source": [
      "Theano Loop & Iteration(map)"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "- http://deeplearning.net/software/theano/tutorial/loop.html\n",
      "- http://deeplearning.net/software/theano/library/scan.html\n",
      "- http://sinhrks.hatenablog.com/entry/2015/04/25/233025"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "from theano import function, config, shared, sandbox\n",
      "import theano\n",
      "import theano.tensor as T\n",
      "import numpy as np\n",
      "import time\n"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stderr",
       "text": [
        "Using gpu device 0: GeForce GT 620\n"
       ]
      }
     ],
     "prompt_number": 1
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "fx2=lambda prior,n: prior * 2"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 2
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "- Loop\n",
      "\n",
      "-- \u95a2\u6570\u3092\u7573\u307f\u8fbc\u3093\u3067\u3044\u3063\u305f\u7d50\u679c\u3092\u51fa\u529b \n",
      "\n",
      "-- \u5f15\u6570non_sequenses\u3092\u6307\u5b9a\n",
      "\n",
      "- iteraion(map)\n",
      "\n",
      "-- \u95a2\u6570\u3092tensor\u306e\u5404\u8981\u7d20\u306b\u9069\u7528\u3057\u305f\u3082\u306e\u3092\u51fa\u529b \n",
      "\n",
      "-- \u5f15\u6570sequences\u3092\u6307\u5b9a\n",
      "\n",
      "--theano.map\u3067\u540c\u7b49\u306e\u3053\u3068\u304c\u3067\u304d\u308b\u3002\n",
      "\n",
      "\u8a73\u3057\u304f\u306f\n",
      "http://sinhrks.hatenablog.com/entry/2015/04/25/233025"
     ]
    },
    {
     "cell_type": "heading",
     "level": 1,
     "metadata": {},
     "source": [
      "Loop"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "[fn(x), fn(fn(x))...]"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "n = T.iscalar('n')\n",
      "v = T.ivector('v')\n",
      "result, updates = theano.scan(fx2,\n",
      "                              sequences=None,\n",
      "                              outputs_info=v,\n",
      "                              non_sequences=v,\n",
      "                              n_steps=n)\n",
      "                              \n",
      "floop = theano.function(inputs=[v, n], outputs=result[-1], updates=updates)\n",
      "print floop([1, 2, 3], 3)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[ 8 16 24]\n"
       ]
      }
     ],
     "prompt_number": 4
    },
    {
     "cell_type": "heading",
     "level": 1,
     "metadata": {},
     "source": [
      "Iteration(map)"
     ]
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "[fn(x) for x in ...]\n",
      "\n",
      "map\u3092\u3064\u304b\u3046\u3068\u3044\u3044\u304b\u3082"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "a = T.ivector('a')\n",
      "outputs = T.as_tensor_variable(np.asarray(0))\n",
      "result, updates = theano.scan(fx2,\n",
      "                              sequences=[a],\n",
      "                              outputs_info=outputs,\n",
      "                              non_sequences=None)\n",
      "                              \n",
      "fite = theano.function(inputs=[a], outputs=result)#, updates=updates)\n",
      "print fite([5,6,7])"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[10 12 14]\n"
       ]
      }
     ],
     "prompt_number": 5
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "result, updates = theano.scan(fn=lambda seq, prior: seq + prior,\n",
      "                              sequences=[a],\n",
      "                              outputs_info=outputs,\n",
      "                              non_sequences=None)\n",
      "\n",
      "fite2 = theano.function(inputs=[a], outputs=result)#, updates=updates)\n",
      "print fite2([5,6,7])"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[ 5 11 18]\n"
       ]
      }
     ],
     "prompt_number": 8
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "m = T.imatrix('m')\n",
      "result, updates = theano.scan(lambda n:n+2,\n",
      "                              sequences=[m],\n",
      "                              non_sequences=None)\n",
      "\n",
      "fite_m=theano.function([m],result)\n",
      "print fite_m([[1,2,3],[4,5,6]])"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[[3 4 5]\n",
        " [6 7 8]]\n"
       ]
      }
     ],
     "prompt_number": 7
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "X=T.scalar('X')\n",
      "A=T.vector('A')\n",
      "fx=lambda x,y:x*y\n",
      "\n",
      "fm,_=theano.map(fx,\n",
      "                sequences=[A], \n",
      "                non_sequences=X\n",
      "                )\n",
      "fmapx = theano.function(inputs=[X,A], outputs=fm) \n",
      "print fmapx(4,[1,2,3]) "
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[  4.   8.  12.]\n"
       ]
      }
     ],
     "prompt_number": 15
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "fm,_=theano.map(fx,\n",
      "                sequences=[m], \n",
      "                non_sequences=X\n",
      "                )\n",
      "fmapxm = theano.function(inputs=[X,m], outputs=fm) \n",
      "print fmapxm(4,[[1,2,3],[4,5,6]]) "
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[[  4.   8.  12.]\n",
        " [ 16.  20.  24.]]\n"
       ]
      }
     ],
     "prompt_number": 18
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "V=T.vector('V')\n",
      "fm,_=theano.map(lambda x,v:x+v,\n",
      "                sequences=[m], \n",
      "                non_sequences=V\n",
      "                )\n",
      "fmapxmv = theano.function(inputs=[V,m], outputs=fm) \n",
      "print fmapxmv([2,2,2],[[1,2,3],[4,5,6]])"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "[[ 3.  4.  5.]\n",
        " [ 6.  7.  8.]]\n"
       ]
      }
     ],
     "prompt_number": 26
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [],
     "language": "python",
     "metadata": {},
     "outputs": []
    }
   ],
   "metadata": {}
  }
 ]
}