def train(self, xmlfile_list):
""" Calculate the column identification model given the columned xmlfile list
"""
parser = etree.XMLParser(recover=True)
featpile = []
labelpile = []
print("loading")
for xmlfile in xmlfile_list:
root = etree.parse(xmlfile, parser).getroot()
htmlnode = root[0]
feats, labels = self.getAllFeature(htmlnode)
if len(feats) > 0:
featpile.append(self.formatFeature(feats))
labelpile.append(array([LABEL[k] for k in labels]))
allFeats = vstack(featpile)
allLabels = hstack(labelpile)
print(len(allFeats))
print(len(allLabels))
print("load complete")
models = {}
for key in self.label_set:
print("training", key)
keyLabel = numpy.int8(allLabels == LABEL[key])
models[key] = adaboost.train(allFeats, keyLabel, balance_factor=15.0)
return models
from scipy import *
from zhiqiang import adaboost
##tset = array([[1,2,3],[0,5,1]])
##lset = array([1,0])
##param = adaboost.train(tset,lset)
##testset = array([[9,10,-1],[-1,2,3]])
##pls = adaboost.predict(testset,param)
from scipy import io
dataset = io.loadmat('G:/dataset/mnist/mnist_all.mat')
trainset = concatenate((dataset['train0'],dataset['train1']))
trainset = trainset.astype('float')
(n1,m) = dataset['train1'].shape
(n0,m) = dataset['train0'].shape
labels = ones(n1+n0)
labels[:n0] = -1
param = adaboost.train(trainset, labels, 2)
testset = concatenate((dataset['test0'],dataset['test1']))
testset = testset.astype('float')
pls = adaboost.predict(testset,param)
tn0 = len(dataset['test0'])
tn = len(testset)
tls = ones(tn)
tls[:tn0] = -1
print(sum(tls == pls))
