mirror of
https://github.com/saymrwulf/nvidia-AI.summerschool17-notebooks.git
synced 2026-11-20 09:52:58 +00:00
327 lines
107 KiB
Text
327 lines
107 KiB
Text
|
|
{
|
||
|
|
"metadata": {
|
||
|
|
"name": "39-incremental-pca"
|
||
|
|
},
|
||
|
|
"nbformat": 3,
|
||
|
|
"nbformat_minor": 0,
|
||
|
|
"worksheets": [
|
||
|
|
{
|
||
|
|
"cells": [
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"Incremental PCA\n",
|
||
|
|
"==============="
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": true,
|
||
|
|
"input": [
|
||
|
|
"from pylab import *\n",
|
||
|
|
"from scipy.spatial import distance\n",
|
||
|
|
"from scipy.ndimage import interpolation\n",
|
||
|
|
"import tables\n",
|
||
|
|
"import random as pyrandom"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [],
|
||
|
|
"prompt_number": 38
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": true,
|
||
|
|
"input": [
|
||
|
|
"def showgrid(images,rows=None,cols=10,size=1,cmap=cm.gray):\n",
|
||
|
|
" if rows is None: rows = (len(images)+cols-1)//cols\n",
|
||
|
|
" figsize(cols*size,rows*size)\n",
|
||
|
|
" for i in range(len(images)):\n",
|
||
|
|
" subplot(rows,cols,i+1)\n",
|
||
|
|
" imshow(images[i],cmap=cmap)"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [],
|
||
|
|
"prompt_number": 39
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"with tables.openFile(\"mnist.h5\",\"r\") as mnist:\n",
|
||
|
|
" print mnist\n",
|
||
|
|
" images = array(mnist.root.deskewed,'f')\n",
|
||
|
|
" labels = array(mnist.root.labels,'i')\n",
|
||
|
|
" test_images = array(mnist.root.test_deskewed,'f')\n",
|
||
|
|
" test_labels = array(mnist.root.test_labels,'i')"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
"mnist.h5 (File) ''\n",
|
||
|
|
"Last modif.: 'Tue Jun 5 10:16:56 2012'\n",
|
||
|
|
"Object Tree: \n",
|
||
|
|
"/ (RootGroup) ''\n",
|
||
|
|
"/deskewed (EArray(60000, 28, 28), shuffle, zlib(9)) ''\n",
|
||
|
|
"/images (CArray(60000, 28, 28), shuffle, zlib(9)) ''\n",
|
||
|
|
"/labels (CArray(60000,), shuffle, zlib(9)) ''\n",
|
||
|
|
"/test_deskewed (EArray(10000, 28, 28), shuffle, zlib(9)) ''\n",
|
||
|
|
"/test_images (CArray(10000, 28, 28), shuffle, zlib(9)) ''\n",
|
||
|
|
"/test_labels (CArray(10000,), shuffle, zlib(9)) ''"
|
||
|
|
]
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 40
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"figsize(6,6)\n",
|
||
|
|
"mu = mean(images,axis=0)\n",
|
||
|
|
"gray()\n",
|
||
|
|
"imshow(mu)"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "pyout",
|
||
|
|
"prompt_number": 78,
|
||
|
|
"text": [
|
||
|
|
"<matplotlib.image.AxesImage at 0x8071dd0>"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "display_data",
|
||
|
|
"png": "iVBORw0KGgoAAAANSUhEUgAAAWwAAAFpCAYAAAC4UhIKAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nennaRSuXNVhreZlwguZorfi0
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 78
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"Let's start by centering the images."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": true,
|
||
|
|
"input": [
|
||
|
|
"centered = images - mu[newaxis,:,:]"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [],
|
||
|
|
"prompt_number": 42
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"Next, we start off with some random vectors as our initial guesses for the eigenvectors."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": true,
|
||
|
|
"input": [
|
||
|
|
"d = 28*28\n",
|
||
|
|
"evs = array([randn(d) for i in range(6)])\n",
|
||
|
|
"evs = array([e/norm(e) for e in evs])"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [],
|
||
|
|
"prompt_number": 71
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"The key idea behind the algorithm is the following.\n",
|
||
|
|
"\n",
|
||
|
|
"We the first eigenvector is reached in the limit as $M^k x / ||M^k x||$\n",
|
||
|
|
"for some random starting vector $x$.\n",
|
||
|
|
"Actually, for the computation, it doesn't really matter when we renormalize $x$.\n",
|
||
|
|
"We can perform some matrix multiplications and then normalize if we like.\n",
|
||
|
|
"\n",
|
||
|
|
"$M$ is the covariance matrix $\\frac{1}{N}\\sum x_i^Tx$.\n",
|
||
|
|
"So, we get $Mx = \\frac{1}{N}(\\sum x_ix_i^T)x=\\frac{1}{N}\\sum x_i(x_i\\cdot x)$.\n",
|
||
|
|
"\n",
|
||
|
|
"Now, if the training vectors $x_i$ are actually just i.i.d. samples from some\n",
|
||
|
|
"distribution, then it really doesn't matter that much how large we choose $N$. All\n",
|
||
|
|
"we are really doing is computing an average over a chunk of them and then multiplying\n",
|
||
|
|
"with $x$--or exchaning the multiplications. And it also doesn't really matter when\n",
|
||
|
|
"we renormalize $x$.\n",
|
||
|
|
"\n",
|
||
|
|
"To arrive at a fully incremental algorithm, we replace the average with\n",
|
||
|
|
"a moving average. This amounts to giving the samples different amounts of weight."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": true,
|
||
|
|
"input": [
|
||
|
|
"eta = 0.001\n",
|
||
|
|
"def update(x):\n",
|
||
|
|
" x = x/norm(x)\n",
|
||
|
|
" for i,e in enumerate(evs):\n",
|
||
|
|
" e += eta*x*dot(x,e)\n",
|
||
|
|
" e /= norm(e)\n",
|
||
|
|
" x -= dot(e,x)*e"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [],
|
||
|
|
"prompt_number": 76
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"Now we just perform a lot of updates.\n",
|
||
|
|
"With the moving average and all this isn't all that\n",
|
||
|
|
"efficient, but it requires very little storage\n",
|
||
|
|
"(just the eigenvectors themselves)."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"for iter in range(1000000):\n",
|
||
|
|
" if iter%100000==0: print iter\n",
|
||
|
|
" update(centered[iter%len(centered)].ravel())"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
"0\n",
|
||
|
|
"100000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
"\n",
|
||
|
|
"200000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
"\n",
|
||
|
|
"300000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
"\n",
|
||
|
|
"400000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
"\n",
|
||
|
|
"500000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
"\n",
|
||
|
|
"600000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
"\n",
|
||
|
|
"700000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
"\n",
|
||
|
|
"800000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
"\n",
|
||
|
|
"900000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": []
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 75
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"metadata": {},
|
||
|
|
"source": [
|
||
|
|
"Let's look at the eigenvectors now. When you look at the eigenvectors of the regular PCA,\n",
|
||
|
|
"you see that these are very similar (some of them have their signs switched)."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"figsize(10,10)\n",
|
||
|
|
"for i in range(6):\n",
|
||
|
|
" subplot(1,6,i+1)\n",
|
||
|
|
" imshow(evs[i].reshape(28,28),cmap=cm.jet)"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "display_data",
|
||
|
|
"png": "iVBORw0KGgoAAAANSUhEUgAAAksAAABsCAYAAAB6i/alAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n0mzdZ3+fBjg+EZrnwjrvPaog
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 77
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": true,
|
||
|
|
"input": [],
|
||
|
|
"language": "python",
|
||
|
|
"metadata": {},
|
||
|
|
"outputs": []
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"metadata": {}
|
||
|
|
}
|
||
|
|
]
|
||
|
|
}
|