Skip to content
This repository

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
tree: eb501fd137
Fetching contributors…

Cannot retrieve contributors at this time

executable file 297 lines (256 sloc) 9.954 kb
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297
#!/usr/bin/python

import random
import data
from numpy import *
from PyML import *
from PyML.containers import *
from maxent import MaxentModel
from scipy.sparse import csr_matrix, lil_matrix, csc_matrix, issparse
import sys
from ngrams import *
import tempfile
import os

"""
A classifier has a addFeatureVector method that takes a feature
vector, which is a dictionary that maps feature names to values,
like {"word1":4, "word2":2, ...} and has a classify method that
takes in a new vector and returns a class
"""

class Classifier:
    def __init__(self):
        self.nfeatures = 0
        self.nvectors = 0
        self.index = {}

    def addToIndex(self, words):
        self.compiled = False
        words = set([i for i in words])
        keys = set(self.index.keys())
        words = words - keys
        for w in words:
            self.index[w] = self.nfeatures
            self.nfeatures += 1
            
    def vectorFromDict (self, words):
        self.addToIndex(words.keys())
        vec = zeros(self.nfeatures)
        for w in words:
            vec[self.index[w]] = words[w]
        return vec
                
            
class OneClassifier:
    def addFeatureVector (self, vec, cls):
        pass
    def classify(self, point):
        return 1


class RandomClassifier:

    def addFeatureVector (self, vec, cls):
        pass
    def classify(self, point):
        return random.randint(0,1)


class BayesClassifier(Classifier):
    def __init__(self, restrictFeatures = False) :
        Classifier.__init__(self)
        print "Bayes: Creating model"
        self.length = 0
        self.compiled = True
        self.classes = {}
        self.restrictFeatures = restrictFeatures
        if restrictFeatures:
            self.addToIndex(self.restrictFeatures)

    def addToIndex(self, words):
        words = set(words) - set(self.index.keys())
        for cls in self.classes:
            self.classes[cls] = hstack((self.classes[cls], ones(len(words))))
        Classifier.addToIndex(self, words)
        
    def addFeatureVector(self, vec, cls, binary=False):
        self.compiled = False
        if cls not in self.classes:
            self.classes[cls] = ones(self.nfeatures)
        if not self.restrictFeatures:
            self.addToIndex(vec)
        for feature in vec:

            if self.restrictFeatures and feature not in self.restrictFeatures:
                continue
            if feature in self.index:

                if binary:
                    self.classes[cls][self.index[feature]] += 1
                else:
                    self.classes[cls][self.index[feature]] += vec[feature]

        self.nvectors += 1
        self.length += 1;

    def compile(self):
        if self.compiled:
            return
        self.compiled = True
        self.normalized = self.classes
        self.lengths = {}
        for i in range(self.nfeatures):
            total = 0
            for cls in self.classes:
                total += self.classes[cls][i]
            for cls in self.classes:
                self.normalized[cls][i] = float(self.classes[cls][i])/total
        for cls in self.classes:
            self.lengths[cls] = 0
            for i in range(self.nfeatures):
                self.lengths[cls] += self.classes[cls][i]
            self.lengths[cls] = sqrt(self.lengths[cls])
            for i in range(self.nfeatures):
                self.normalized[cls][i] /= self.lengths[cls]

    def classify(self, vec, binary=False):
        self.compile()
        mx = -sys.maxint
        mx_cls = 0
        point = ones(self.nfeatures)
        for feature in vec:
            if feature in self.index:
                if binary:
                    point[self.index[feature]] += 1
                else:
                    point[self.index[feature]] += vec[feature]
        for cls in self.classes:
            dotprod = dot(log(self.classes[cls]), log(point)) - log(self.lengths[cls])
            if dotprod > mx:
                mx = dotprod
                mx_cls = cls
        return mx_cls


class BayesPresenceClassifier(BayesClassifier):
    def classify(self, point):
        return BayesClassifier.classify(self, point.clip(max=2))


class LinearSVMClassifier(Classifier):
    def __init__(self, restrictFeatures=False):
        Classifier.__init__(self)
        print "LinearSVM: Creating model"
        self.file = tempfile.NamedTemporaryFile(delete=False)
        self.filename = self.file.name
        print self.filename
        self.data = SparseDataSet(0)
        self.svm = SVM(optimizer='liblinear')
        self.restrictFeatures = restrictFeatures
        self.binary = False
    def vectorToString(self, vec, cls = False, binary=False):
        # granted, this is kind of silly
        # creates a string of the format "[class if point is labeled] feature1:value1 feature2:value2..."
        # where the only allowed features are the ones in restrictFeatures, if we're restricting the features
        if binary:
            self.binary = True
            return ((str(cls) + " ") if cls else "") + \
                    " ".join(["-".join(str(i).split()) + ":1"
                    for i in vec if (not self.restrictFeatures) or
                    (i in self.restrictFeatures)]) + "\n"
        return ((str(cls) + " ") if cls else "") + \
                " ".join(["-".join(str(i).split()) + ":" + str(vec[i])
                for i in vec if (not self.restrictFeatures) or
                (i in self.restrictFeatures)]) + "\n"

    def addFeatureVector(self, point, cls, binary=False):
        self.compiled = False
        vec = self.vectorToString(point, cls, binary=binary)
        self.file.write(vec)
        
    def compile(self):
        if self.compiled == True:
            return
        self.compiled = True
        self.file.close()
        self.data = SparseDataSet(self.filename)
        self.file = open(self.filename)
        self.svm.train(self.data)
        print self.data
# self.validate(3)
    def validate(self, n):
        self.compile()
        print self.data
        outp = self.svm.cv(self.data, numFolds = n)
        print outp
    def classify(self, pt, binary = False):
        self.compile()
        f = tempfile.NamedTemporaryFile(delete=False)
        fname = f.name
        f.write(self.vectorToString(pt, binary = binary))
        f.close()
        data = SparseDataSet(fname)
        os.remove(fname)
        r = self.svm.test(data, verbose=0).getPredictedLabels()[0]
        return int(r)

class MaximumEntropyClassifier(Classifier):
    def __init__(self, restrictFeatures=False):
        Classifier.__init__(self)
        print "MaximumEntropy: Creating model"
        self.model = MaxentModel()
        self.model.verbose = 1
        self.restrictFeatures = restrictFeatures
        self.model.begin_add_event()

    def addToIndex(self, trainingset):
        for (vec,cls) in trainingset:
            self.addFeatureVector(vec,cls)
        
    def addFeatureVector(self, vec, cls, value=1, binary=False):
        for key in vec.keys():
            if key not in self.restrictFeatures:
                del vec[key]
        context = vec.keys()
        label = "%s" % cls
        self.model.add_event(context,label,value)

    def compile(self):
        self.model.end_add_event()
        self.model.train(30, "lbfgs", 2, 1E-03)
        #self.model.train(100, 'gis', 2)
        print "> Models trained"

    def classify(self, point, label='1', binary=False):
        result = self.model.eval(point.keys(), label)
        if result >= 0.5:
            return 1
        return -1

class MajorityVotingClassifier(Classifier):
    def __init__(self):
        self.classifiers = []
        self.reliabilities = []

    def addClassifier(self, classifier, train_files, test_files = [], reliability=1):
        self.classifiers.append(classifier)
        self.reliability.append(reliability)

    def addFeatureVector(self, vec):
        for cls in self.classifiers:
            cls.addFeatureVector(vec)

    def classify(self, vec):
        results = {}
        for cls in self.classifiers:
            r = cls.classify(vec)
            if r not in results:
                results[r] = 1
            else:
                results[r] += 1
        mx = 0
        mxarg = 0

        for r in results:
            if results[r] > mx:
                mxarg = r
                mx = results[r]
        return mxarg
        
def test_bayes():
    trainingset = array([[2, 2, 2, 1],
                         [1, 1, 2, 0],
                         [1, 1, 2, 0],
                         [2, 1, 1, 0]]).T
    bc = BayesClassifier()
    for vec in trainingset:
        bc.addFeatureVector(vec[:-1], vec[-1])
    print bc.classify(array([2, 2, 2]))
    print bc.classify(array([3, 1, 1]))

            
def test_svm():
    trainingset = [ngrams(1, "foo foo bar baz"), ngrams(1, "foo foo bar bar baz baz"), ngrams(1,"foo foo bar baz")]
    labels = [1, -1, -1]
    lsc = LinearSVMClassifier()
    for vec in zip(trainingset, labels):
        lsc.addFeatureVector(vec[0], vec[1])
    print lsc.classify(ngrams(1, "foo foo bar bar baz baz"))
    print lsc.classify(ngrams(1, "foo foo foo bar baz"))

def test_maxent():
    trainingset = [(['good'],'pos',1),
                   (['wonderful'],'pos',1),
                   (['ugly'],'neg',1),
                   (['terrible','ick'],'neg',1)]
    m = MaximumEntropyClassifier(trainingset)
    
    print "other label: %s" % m.classify(['mmm'],'otherlabel') # other label
    print "OOD: %s" % m.classify(['mmm'],'pos') # OOD
    print "ick: %s" % m.classify(['ick'],'pos')
    print "mmm awesome good: %s" % m.classify(['mmm','awesome','good'],'pos')
    print "mmm terrible good: %s" % m.classify(['mmm','terrible','good'],'pos')
    print "wonderful terrible good: %s" % m.classify(['wonderful','terrible','good'],'pos')

if __name__ == "__main__":
    test_maxent()
Something went wrong with that request. Please try again.