nvidia-AI.summerschool17-no.../teaching-lw/130-kmeans.ipynb

800 lines
934 KiB
Text
Raw Permalink Normal View History

2017-09-22 17:26:46 +00:00
{
"metadata": {
"name": "30-kmeans"
},
"nbformat": 2,
"worksheets": [
{
"cells": [
{
"cell_type": "code",
"collapsed": true,
"input": [
"from pylab import *",
"from scipy.spatial import distance",
"from scipy.ndimage import interpolation",
"import tables",
"import random as pyrandom"
],
"language": "python",
"outputs": [],
"prompt_number": 1
},
{
"cell_type": "code",
"collapsed": true,
"input": [
"with tables.openFile(\"mnist.h5\",\"r\") as mnist:",
" images = array(mnist.root.deskewed,'f')",
" labels = array(mnist.root.labels,'i')",
" test_images = array(mnist.root.test_deskewed,'f')",
" test_labels = array(mnist.root.test_labels,'i')"
],
"language": "python",
"outputs": [],
"prompt_number": 2
},
{
"cell_type": "code",
"collapsed": true,
"input": [],
"language": "python",
"outputs": [],
"prompt_number": 2
},
{
"cell_type": "code",
"collapsed": true,
"input": [
"def showgrid(images,rows=None,cols=10,size=1):",
" if rows is None: rows = (len(images)+cols-1)//cols",
" figsize(cols*size,rows*size)",
" for i in range(len(images)):",
" subplot(rows,cols,i+1)",
" imshow(images[i])"
],
"language": "python",
"outputs": [],
"prompt_number": 3
},
{
"cell_type": "code",
"collapsed": true,
"input": [],
"language": "python",
"outputs": [],
"prompt_number": 3
},
{
"cell_type": "markdown",
"source": [
"The k-Means Algorithm",
"====================="
]
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"protos = array(pyrandom.sample(images,50),'f')",
"showgrid(protos)"
],
"language": "python",
"outputs": [
{
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAAkwAAAEzCAYAAADD+FxKAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n1iKMn0ZiT/KYKD8Q+Y0MXAOs
}
],
"prompt_number": 4
},
{
"cell_type": "code",
"collapsed": true,
"input": [
"from scipy.spatial.distance import cdist",
"def make2d(a): return a.reshape(len(a),prod(a.shape[1:]))",
"closest = argmin(cdist(make2d(protos),make2d(images)),axis=0)",
"for i in range(len(protos)):",
" protos[i,:] = average(images[closest==i,:],axis=0)"
],
"language": "python",
"outputs": [],
"prompt_number": 5
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"showgrid(protos)"
],
"language": "python",
"outputs": [
{
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAAkwAAAEzCAYAAADD+FxKAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nco+PhcY63r3vFsbPhsi+1YM+
}
],
"prompt_number": 6
},
{
"cell_type": "code",
"collapsed": true,
"input": [
"def kmeans(images,k=50,protos=None,maxrounds=50,verbose=1):",
" last_closest = -ones(len(images))",
" if protos is None: protos = array(pyrandom.sample(images,k))",
" for round in range(maxrounds):",
" dists = cdist(make2d(protos),make2d(images))",
" distortion = sum(amin(dists,axis=0))/len(images)",
" closest = argmin(dists,axis=0)",
" delta = sum(closest!=last_closest)",
" last_closest = closest",
" if verbose: print round,delta,distortion",
" if delta==0: break",
" for i in range(len(protos)):",
" protos[i,:] = average(images[closest==i,:],axis=0)",
" return protos"
],
"language": "python",
"outputs": [],
"prompt_number": 7
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"protos = kmeans(images,protos=protos)"
],
"language": "python",
"outputs": [
{
"output_type": "stream",
"stream": "stdout",
"text": [
"0 60000 4.89633096854",
"1"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 7917 4.80152271087",
"2"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 5169 4.76616597228",
"3"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 3965 4.74739624393",
"4"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 3100 4.73606869655",
"5"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 2666 4.72846797302",
"6"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 2296 4.72260930411",
"7"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 2040 4.71798355986",
"8"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 1830 4.713926274",
"9"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 1751 4.71005046462",
"10"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 1616 4.7063091299",
"11"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 1382 4.70317694431",
"12"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 1228 4.70068537836",
"13"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 1034 4.69875396393",
"14"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 954 4.69717610557",
"15"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 855 4.69589637415",
"16"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 776 4.69484689542",
"17"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 649 4.69405445444",
"18"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 567 4.69343020086",
"19"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 520 4.69297141312",
"20"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 508 4.69251434001",
"21"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 472 4.69212180104",
"22"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 443 4.69176203552",
"23"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 433 4.6914174986",
"24"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 416 4.69103357681",
"25"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 397 4.69069165281",
"26"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 401 4.69034944859",
"27"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 380 4.6900712992",
"28"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 346 4.68978428872",
"29"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 293 4.68953482214",
"30"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 280 4.68932054066",
"31"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 264 4.68913012026",
"32"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 240 4.68898832769",
"33"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 212 4.68885735153",
"34"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 217 4.68870451944",
"35"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 225 4.68856641584",
"36"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 205 4.68844243897",
"37"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 228 4.68830257083",
"38"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 205 4.68815295769",
"39"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 219 4.68801849828",
"40"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 197 4.68788090185",
"41"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 162 4.68777117098",
"42"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 155 4.68766759273",
"43"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 152 4.68758096931",
"44"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 150 4.68751857016",
"45"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 150 4.68745809549",
"46"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 150 4.68738490994",
"47"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 173 4.68731986822",
"48"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 161 4.68722955056",
"49"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 185 4.68712567791"
]
}
],
"prompt_number": 8
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"showgrid(protos)"
],
"language": "python",
"outputs": [
{
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAAkwAAAEzCAYAAADD+FxKAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n4DWdJ5Yunev1XX7bV+RheuaZ
}
],
"prompt_number": 9
},
{
"cell_type": "markdown",
"source": [
"k-Means as a Classifier",
"======================="
]
},
{
"cell_type": "code",
"collapsed": true,
"input": [
"dists = cdist(make2d(protos),make2d(images))",
"closest = argmin(dists,axis=0)"
],
"language": "python",
"outputs": [],
"prompt_number": 10
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"figsize(8,4)",
"_=hist(labels[closest==2])"
],
"language": "python",
"outputs": [
{
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAAe0AAAD9CAYAAACRFS/UAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEG0AAAxBtAEAMATRBgDAEBlH
}
],
"prompt_number": 11
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"showgrid(images[closest==2][:50])"
],
"language": "python",
"outputs": [
{
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAAkwAAAEzCAYAAADD+FxKAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nGTQBI2CPQJWVvNJhVmfuJbew
}
],
"prompt_number": 12
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"def chist(classes,n=10):",
" result = zeros(n)",
" for i in range(n):",
" result[i] = sum(classes==i)",
" return array(result,'i')",
"",
"dists = cdist(make2d(protos),make2d(images))",
"closest = argmin(dists,axis=0)",
"proto_labels = [argmax(chist(labels[closest==i])) for i in range(50)]",
"sorted(set(proto_labels))"
],
"language": "python",
"outputs": [
{
"output_type": "pyout",
"prompt_number": 13,
"text": [
"[0, 1, 2, 3, 4, 5, 6, 7, 8, 9]"
]
}
],
"prompt_number": 13
},
{
"cell_type": "code",
"collapsed": true,
"input": [
"def nneval(training,ltraining,testing,ltesting,metric='euclidean',p=2):",
" ds = distance.cdist(make2d(testing),make2d(training),metric=metric,p=p)",
" closest = argmin(ds,axis=1)",
" predictions = [ltraining[i] for i in closest]",
" errors = sum(array(predictions)!=array(ltesting))",
" return errors*1.0/len(testing)"
],
"language": "python",
"outputs": [],
"prompt_number": 14
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"nneval(protos,proto_labels,test_images,test_labels)"
],
"language": "python",
"outputs": [
{
"output_type": "pyout",
"prompt_number": 15,
"text": [
"0.093299999999999994"
]
}
],
"prompt_number": 15
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"for k in [20,20,20,50,100,200,500]:",
" protos = kmeans(images,k=k,verbose=0,maxrounds=10)",
" dists = cdist(make2d(protos),make2d(images))",
" closest = argmin(dists,axis=0)",
" proto_labels = [argmax(chist(labels[closest==i])) for i in range(len(protos))]",
" print \"RESULT\",k,nneval(protos,proto_labels,test_images,test_labels)"
],
"language": "python",
"outputs": [
{
"output_type": "stream",
"stream": "stdout",
"text": [
"RESULT 20 "
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
"0.215",
"RESULT"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 20 "
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
"0.1915",
"RESULT"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 20 "
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
"0.1722",
"RESULT"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 50 "
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
"0.107",
"RESULT"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 100 "
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
"0.0792",
"RESULT"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 200 "
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
"0.0593",
"RESULT"
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
" 500 "
]
},
{
"output_type": "stream",
"stream": "stdout",
"text": [
"0.0407"
]
}
],
"prompt_number": 16
}
]
}
]
}