mirror of
https://github.com/saymrwulf/nvidia-AI.summerschool17-notebooks.git
synced 2026-11-21 09:54:43 +00:00
741 lines
321 KiB
Text
741 lines
321 KiB
Text
|
|
{
|
||
|
|
"metadata": {
|
||
|
|
"name": "12-logistic-regression-grad"
|
||
|
|
},
|
||
|
|
"nbformat": 2,
|
||
|
|
"worksheets": [
|
||
|
|
{
|
||
|
|
"cells": [
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": true,
|
||
|
|
"input": [
|
||
|
|
"import scipy,pylab,numpy",
|
||
|
|
"from numpy import *",
|
||
|
|
"from scipy import *",
|
||
|
|
"from pylab import *",
|
||
|
|
"from urllib import urlopen",
|
||
|
|
"from gzip import GzipFile",
|
||
|
|
"from scipy.spatial import distance",
|
||
|
|
"from scipy.spatial.distance import cdist"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [],
|
||
|
|
"prompt_number": 7
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"source": [
|
||
|
|
"Linear Classification Problem",
|
||
|
|
"================",
|
||
|
|
"",
|
||
|
|
"Let's generate a linear problem, as before."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"data = random_sample((1000,2))",
|
||
|
|
"labels = (data[:,0]*0.7+data[:,1]*0.4>0.5)",
|
||
|
|
"sum(labels)"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "pyout",
|
||
|
|
"prompt_number": 8,
|
||
|
|
"text": [
|
||
|
|
"608"
|
||
|
|
]
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 8
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"d0 = data[labels==False]",
|
||
|
|
"d1 = data[labels]",
|
||
|
|
"figure(figsize=(6,6))",
|
||
|
|
"plot(d0[:,0],d0[:,1],\"bo\")",
|
||
|
|
"plot(d1[:,0],d1[:,1],\"ro\")",
|
||
|
|
"print len(d0),len(d1)"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
"392 608"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "display_data",
|
||
|
|
"png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAFtCAYAAAD8jSoDAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nz54+baSekknUHz4MAEgpbMCN
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 9
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"augmented = concatenate([ones((len(data),1)),data],axis=1)",
|
||
|
|
"augmented[:3,:]"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "pyout",
|
||
|
|
"prompt_number": 10,
|
||
|
|
"text": [
|
||
|
|
"array([[ 1. , 0.62170828, 0.20003034],",
|
||
|
|
" [ 1. , 0.03303012, 0.36609942],",
|
||
|
|
" [ 1. , 0.34112627, 0.96111128]])"
|
||
|
|
]
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 10
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"source": [
|
||
|
|
"Least Square Solution",
|
||
|
|
"====================",
|
||
|
|
"",
|
||
|
|
"We now want to find a vector $a$ such that $a \\cdot x_i \\approx c_i$ for all $i$.",
|
||
|
|
"",
|
||
|
|
"We can write this as a matrix and vector equation if we put the measurements into the",
|
||
|
|
"rows of a matrix $X = {x_1\\choose x_N}$ and $c = {c_1\\choose c_N}$",
|
||
|
|
"",
|
||
|
|
"$$X \\cdot a^T = c$$",
|
||
|
|
"",
|
||
|
|
"In order to solve this, we take the pseudo-inverse of $X$:",
|
||
|
|
"",
|
||
|
|
"$$a^T = X^\\dagger \\cdot c$$",
|
||
|
|
"",
|
||
|
|
"The pseudo-inverse is a generalization of the inverse matrix for non-rectangular and/or non-full-rank matrices."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"a = dot(linalg.pinv(augmented),labels)",
|
||
|
|
"d,a0,a1 = a",
|
||
|
|
"print d,a0,a1"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
"-0.345563974086 1.27003201537 0.600651460697"
|
||
|
|
]
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 21
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"source": [
|
||
|
|
"Note that the decision boundary is at $a\\cdot x = 0.5$, not at $a\\cdot x = 0$, since",
|
||
|
|
"we want the decision boundary halfway between the two numerical values $0$ and $1$,",
|
||
|
|
"corresponding to the labels."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"d0 = data[labels<=0]",
|
||
|
|
"d1 = data[labels>0]",
|
||
|
|
"figure(figsize=(6,6)); xlim((0,1)); ylim((0,1))",
|
||
|
|
"plot(d0[:,0],d0[:,1],\"bo\")",
|
||
|
|
"plot(d1[:,0],d1[:,1],\"ro\")",
|
||
|
|
"plot([0,-d/a0+0.5],[-d/a1+0.5,0],\"g\")"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "pyout",
|
||
|
|
"prompt_number": 24,
|
||
|
|
"text": [
|
||
|
|
"[<matplotlib.lines.Line2D at 0x36dcdd0>]"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "display_data",
|
||
|
|
"png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAFtCAYAAAD8jSoDAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\naKIveGrFclt9omU3YxPtkQWW
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 24
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"source": [
|
||
|
|
"The problem with these kinds of least square solutions is that they are not very good: ",
|
||
|
|
"although a linear function can represent the decision function just fine,",
|
||
|
|
"it is not a good approximation to the function represented by the labels.",
|
||
|
|
"",
|
||
|
|
"In fact, the function represented by the class labels is, effectively,",
|
||
|
|
"the posterior probability distribution, a smooth function, that is in general",
|
||
|
|
"better approximate by sigmoids (see below)."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"source": [
|
||
|
|
"Sigmoid Functions",
|
||
|
|
"=================",
|
||
|
|
"",
|
||
|
|
"A problem with computing linear least square approximations to",
|
||
|
|
"the class label is that we can already see that a linear approximation",
|
||
|
|
"can never be very good: the class label is essentially a step function,",
|
||
|
|
"but the linear approximant to grows without bounds.",
|
||
|
|
"",
|
||
|
|
"We've already seen that we can interpret the linear functions $g_c(x) = w_c \\cdot x$ ",
|
||
|
|
"as discriminant functions, but there is no reason to limit ourselves to linear functions.",
|
||
|
|
"",
|
||
|
|
"What we would really like to optimize (in the two-class case) is:",
|
||
|
|
"",
|
||
|
|
"$$\\hbox{err} = \\sum_i (H(w \\cdot x_i) - c_i)^2$$",
|
||
|
|
"",
|
||
|
|
"Here, $H(x) = \\left\\lfloor x > 0 \\right\\rfloor$, the Heaviside function. Unfortunately, we can't optimize that via gradient descent (the only non-linear optimization method that we know) because the Heaviside function does not have a derivative at 0.",
|
||
|
|
"",
|
||
|
|
"We can, however, compute the derivative of a continuous approximation of the Heaviside function, namely a sigmoid function.",
|
||
|
|
"",
|
||
|
|
"$$\\sigma_q(x) = \\frac{1}{1+e^{-q\\,x}}$$",
|
||
|
|
"",
|
||
|
|
"This function looks non-symetrical, but it actually is. Let's plot it:"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"import matplotlib.pyplot as plt",
|
||
|
|
"xs = linspace(-6,6,100)",
|
||
|
|
"def sig(x,q): return 1.0/(1.0+exp(-q*x))",
|
||
|
|
"plt.cla()",
|
||
|
|
"plot(xs,sig(xs,1.0))",
|
||
|
|
"plot(xs,sig(xs,2.0))",
|
||
|
|
"plot(xs,sig(xs,4.0))",
|
||
|
|
"plot(xs,sig(xs,8.0))",
|
||
|
|
"savefig(\"temp.png\")"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "display_data",
|
||
|
|
"png": "iVBORw0KGgoAAAANSUhEUgAAAXIAAAD9CAYAAAChtfywAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nC8dYeWQlx549ZsMelvTzz2q7
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 9
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"source": [
|
||
|
|
"Using the sigmoid function, the discriminant functions become:",
|
||
|
|
"",
|
||
|
|
"$$g_c(x) = \\sigma(w_c \\cdot x)$$",
|
||
|
|
"",
|
||
|
|
"We don't bother introducing the scale factor $q$ because that is already contained in the weight vector $w_c$; that is, since the norm of that vector is not controlled and only the direction of the vector matters for the discriminant boundary, the gradient descent algorithm can itself let $q$ go to infinity in order to obtain better and better approximations to the Heaviside function.",
|
||
|
|
"",
|
||
|
|
"Using the sigmoid function, the discriminant functions become:",
|
||
|
|
"",
|
||
|
|
"$$g_c(x) = \\sigma(w_c \\cdot x)$$",
|
||
|
|
"",
|
||
|
|
"As before, we compute the gradient of the error:",
|
||
|
|
"",
|
||
|
|
"$$\\hbox{err} = \\sum_i (\\sigma(w\\cdot x_i)-c_i)^2$$",
|
||
|
|
"",
|
||
|
|
"Using the same sloppy notation for partial derivatives as before, we can write:",
|
||
|
|
"",
|
||
|
|
"$$\\frac{\\partial\\hbox{err}}{\\partial w} = 2 \\sum_i (\\sigma(w\\cdot x_i)-c_i) \\sigma'(w\\cdot x_i) x_i$$",
|
||
|
|
"",
|
||
|
|
"Note that",
|
||
|
|
"",
|
||
|
|
"$$\\sigma'(x) = \\sigma(x) (1-\\sigma(x))$$",
|
||
|
|
"",
|
||
|
|
"",
|
||
|
|
"",
|
||
|
|
"Let's implement this. Note that the class labels are zero and one again, $c\\in\\{0,1\\}$, matching the output of the sigmoid function $\\sigma(x) \\in [0,1]$."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": true,
|
||
|
|
"input": [
|
||
|
|
"def sigmoid(x): return 1.0/(1.0+exp(-x))"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [],
|
||
|
|
"prompt_number": 11
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"w = random_sample((3))",
|
||
|
|
"w"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "pyout",
|
||
|
|
"prompt_number": 12,
|
||
|
|
"text": [
|
||
|
|
"array([ 0.41829084, 0.31289947, 0.58718571])"
|
||
|
|
]
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 12
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"eta = 0.01",
|
||
|
|
"for iter in range(500000):",
|
||
|
|
" i = iter%len(data)",
|
||
|
|
" if iter%100000==0: print iter,eta,w",
|
||
|
|
" x = augmented[i]",
|
||
|
|
" c = labels[i]",
|
||
|
|
" s = sigmoid(dot(w,x))",
|
||
|
|
" sprime = s*(1-s)",
|
||
|
|
" delta = s-c",
|
||
|
|
" w -= eta*(s-c)*sprime*x"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
"0 0.01 [ 0.41829084 0.31289947 0.58718571]",
|
||
|
|
"100000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
" 0.01 [-4.06342381 6.21008656 2.92048786]",
|
||
|
|
"200000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
" 0.01 [-5.49849138 8.13621486 4.09214148]",
|
||
|
|
"300000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
" 0.01 [-6.46527904 9.45443711 4.87551951]",
|
||
|
|
"400000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
" 0.01 [ -7.21890156 10.49123519 5.48224139]"
|
||
|
|
]
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 13
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"d,a0,a1 = w",
|
||
|
|
"d0 = data[labels<=0]",
|
||
|
|
"d1 = data[labels>0]",
|
||
|
|
"figure(figsize=(6,6))",
|
||
|
|
"ylim((0,1))",
|
||
|
|
"xlim((0,1))",
|
||
|
|
"plot(d0[:,0],d0[:,1],\"bo\")",
|
||
|
|
"plot(d1[:,0],d1[:,1],\"ro\")",
|
||
|
|
"plot([0,-d/a0],[-d/a1,0],\"g\")",
|
||
|
|
"print len(d0),len(d1)"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
"433 567"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "display_data",
|
||
|
|
"png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAFtCAYAAAD8jSoDAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ncth2OZB5HL5q96z1HlmlcqL1
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 14
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"source": [
|
||
|
|
"That looks quite good, but not perfect. Let's look at the value of the discriminant function."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"xs = linspace(0,1,100)[:,newaxis]",
|
||
|
|
"ys = linspace(0,1,100)[newaxis,:]",
|
||
|
|
"cla(); imshow(sigmoid(a0*xs+a1*ys+d).T,origin='lower'); savefig(\"temp.png\")"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "display_data",
|
||
|
|
"png": "iVBORw0KGgoAAAANSUhEUgAAAP0AAAD5CAYAAADspDPqAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nGJz49gn2MdQTkDS5xxZLrbsO
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 15
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"source": [
|
||
|
|
"Blue represents zero and red represents one. From this picture, we can see that the transition region between the two values is quite fuzzy; the decision boundary isn't at all sharp."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"xs = linspace(0,1,100)",
|
||
|
|
"ys = xs",
|
||
|
|
"cla(); plot(xs,sigmoid(a0*xs+a1*ys+d)); savefig(\"temp.png\")"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "display_data",
|
||
|
|
"png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD9CAYAAABHnDf0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nLgl3EX8zDLPP/oc/WF2JiP+5
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 16
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"source": [
|
||
|
|
"Let's update a little longer and with a higher training rate."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"eta = 0.1",
|
||
|
|
"for iter in range(500000):",
|
||
|
|
" i = iter%len(data)",
|
||
|
|
" if iter%100000==0: print iter,eta,w",
|
||
|
|
" x = augmented[i]",
|
||
|
|
" c = labels[i]",
|
||
|
|
" s = sigmoid(dot(w,x))",
|
||
|
|
" sprime = s*(1-s)",
|
||
|
|
" delta = s-c",
|
||
|
|
" w -= eta*(s-c)*sprime*x"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
"0 0.1 [ -7.84683128 11.35962138 5.98575717]",
|
||
|
|
"100000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
" 0.1 [-11.60182816 16.61438497 9.00802051]",
|
||
|
|
"200000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
" 0.1 [-13.81577983 19.70742599 10.76316596]",
|
||
|
|
"300000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
" 0.1 [-15.46759654 22.01714172 12.07428822]",
|
||
|
|
"400000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
" 0.1 [-16.81496314 23.90180674 13.14491147]"
|
||
|
|
]
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 17
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"source": [
|
||
|
|
"Now the decision boundary is a lot \"sharper\":"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"d,a0,a1 = w",
|
||
|
|
"xs = linspace(0,1,100)[:,newaxis]",
|
||
|
|
"ys = linspace(0,1,100)[newaxis,:]",
|
||
|
|
"cla(); imshow(sigmoid(a0*xs+a1*ys+d))"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "pyout",
|
||
|
|
"prompt_number": 20,
|
||
|
|
"text": [
|
||
|
|
"<matplotlib.image.AxesImage at 0x4324390>"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "display_data",
|
||
|
|
"png": "iVBORw0KGgoAAAANSUhEUgAAAP0AAAD8CAYAAAC8aaJZAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nFy9e1LzqSeX3k9VlcO2Ci8Wb
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 20
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"xs = linspace(0,1,100)",
|
||
|
|
"ys = xs",
|
||
|
|
"cla(); plot(xs,sigmoid(a0*xs+a1*ys+d))"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "pyout",
|
||
|
|
"prompt_number": 19,
|
||
|
|
"text": [
|
||
|
|
"[<matplotlib.lines.Line2D at 0x43edc50>]"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "display_data",
|
||
|
|
"png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD9CAYAAABHnDf0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nALGFcIetcWAYMDTCHbbGlAww
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 19
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"source": [
|
||
|
|
"Let's do more of that still."
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"eta = 1.0",
|
||
|
|
"for iter in range(500000):",
|
||
|
|
" i = iter%len(data)",
|
||
|
|
" if iter%100000==0: print iter,eta,w",
|
||
|
|
" x = augmented[i]",
|
||
|
|
" c = labels[i]",
|
||
|
|
" s = sigmoid(dot(w,x))",
|
||
|
|
" sprime = s*(1-s)",
|
||
|
|
" delta = s-c",
|
||
|
|
" w -= eta*(s-c)*sprime*x"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
"0 1.0 [-17.9677928 25.51467577 14.06171267]",
|
||
|
|
"100000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
" 1.0 [-25.35304656 35.90923298 20.11625363]",
|
||
|
|
"200000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
" 1.0 [-30.01543776 42.40278885 23.80803724]",
|
||
|
|
"300000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
" 1.0 [-33.64150081 47.45434356 26.67869104]",
|
||
|
|
"400000"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
" 1.0 [-36.69066627 51.70397516 29.09337889]"
|
||
|
|
]
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 21
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"d,a0,a1 = w",
|
||
|
|
"xs = linspace(0,1,100)[:,newaxis]",
|
||
|
|
"ys = linspace(0,1,100)[newaxis,:]",
|
||
|
|
"cla(); imshow(sigmoid(a0*xs+a1*ys+d))"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "pyout",
|
||
|
|
"prompt_number": 22,
|
||
|
|
"text": [
|
||
|
|
"<matplotlib.image.AxesImage at 0x43ddcd0>"
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"output_type": "display_data",
|
||
|
|
"png": "iVBORw0KGgoAAAANSUhEUgAAAP0AAAD8CAYAAAC8aaJZAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n4S8NcRXN2X27dQU3bMCwCf0B
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 22
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"xs = linspace(0,1,100)",
|
||
|
|
"ys = xs",
|
||
|
|
"cla(); plot(xs,sigmoid(a0*xs+a1*ys+d)); savefig(\"temp.png\")"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "display_data",
|
||
|
|
"png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD9CAYAAABHnDf0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n/TF5rohxHyvUN3riiSe0fv36
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 23
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "markdown",
|
||
|
|
"source": [
|
||
|
|
"Using scikit-learn",
|
||
|
|
"=================="
|
||
|
|
]
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": true,
|
||
|
|
"input": [
|
||
|
|
"from sklearn import linear_model,datasets"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [],
|
||
|
|
"prompt_number": 24
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": true,
|
||
|
|
"input": [
|
||
|
|
"logreg = linear_model.LogisticRegression(C=1e5)"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [],
|
||
|
|
"prompt_number": 25
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"logreg.fit(data,labels)"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "pyout",
|
||
|
|
"prompt_number": 26,
|
||
|
|
"text": [
|
||
|
|
"LogisticRegression(C=100000.0, dual=False, fit_intercept=True,",
|
||
|
|
" intercept_scaling=1, penalty='l2', scale_C=False, tol=0.0001)"
|
||
|
|
]
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 26
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": false,
|
||
|
|
"input": [
|
||
|
|
"z = logreg.predict(data)",
|
||
|
|
"print 1*labels[:10]",
|
||
|
|
"print z[:10]",
|
||
|
|
"print \"errors\",sum(z!=labels)"
|
||
|
|
],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": [
|
||
|
|
{
|
||
|
|
"output_type": "stream",
|
||
|
|
"stream": "stdout",
|
||
|
|
"text": [
|
||
|
|
"[1 1 1 1 0 0 1 1 0 1]",
|
||
|
|
"[1 1 1 1 0 0 1 1 0 1]",
|
||
|
|
"errors 0"
|
||
|
|
]
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"prompt_number": 32
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"cell_type": "code",
|
||
|
|
"collapsed": true,
|
||
|
|
"input": [],
|
||
|
|
"language": "python",
|
||
|
|
"outputs": []
|
||
|
|
}
|
||
|
|
]
|
||
|
|
}
|
||
|
|
]
|
||
|
|
}
|