def loadForest():
npic=0
forest=[]
for root, dirs, files in os.walk(rootdir):
for f in files:
if f.endswith('pic'):
npic=npic+1
print f
#reading the tree pickle file
pickleFile = open(rootdir + "/" + f, 'rb')
root = pickle.load(pickleFile)
pickleFile.close()
#init the test tree
t=tree()
t.settree(root)
forest.append(t)
return forest,npic
def recall(dsetname='ss', rfile=''):
ts=time.time()
strtime=datetime.datetime.fromtimestamp(ts).strftime('%Y-%m-%d %H:%M:%S')
print("Starting time: "+strtime)
#reading the tree pickle file
pickleFile = open(rfile, 'rb')
root = pickle.load(pickleFile)
pickleFile.close()
#init the test tree
t=tree()
t.settree(root)
print("----main::recall::loadtree")
ts=timestamp(ts)
#compute recall rate
loader= imp.load_source('dataset', dsetname+'.py')
dset=loader.dataset()
correct=0;
for x in xrange(dset.size):
cL=dset.getL(x)
#print prob
p=t.getP(np.array([x]),dset)
ids = p.argsort()[::-1][:3]
L=ids[0]
#print max likelihood
#L=t.getL(np.array([x]),dset)
if cL== L:
correct=correct+1
if cL!=0:
print("\n%03d: correct L"%cL)
for i in xrange(len(ids)):
print("%03d_%03d"%(ids[i],100*p[ids[i]])),
dset.setL(x,L)
print("recall rate: {}%".format(correct/float(dset.size)*100))
print("----main::recall::evaluate")
ts=timestamp(ts)
return t, dset
def recall(dsetname='dataset_pickle', rfile='',mylog=None):
mylog.current("recall")
#reading the tree pickle file
pickleFile = open(rfile, 'rb')
root = pickle.load(pickleFile)
pickleFile.close()
#init the test tree
t=tree()
t.settree(root)
print(mylog.finished("main::train>> loading tree"))
#compute recall rate
loader= imp.load_source('dataset', dsetname+'.py')
dset=loader.dataset()
print(mylog.finished("main::train>> loading dset\ndataset: %s"%dset))
conf=np.zeros((dset.clmax,dset.clmax))
correct=0;
for x in xrange(dset.size):
cL=dset.getL(x)#cL:correct label
#print prob
p=t.getP(np.array([x]),dset)
ids = p.argsort()[::-1][:3]
L=ids[0]
#print max likelihood
#L=t.getL(np.array([x]),dset)
if cL== L:
correct=correct+1
conf[L,cL]=conf[L,cL]+1
# print("\n%03d: correct L"%cL)
# for i in xrange(len(ids)):
# print("%03d_%03d"%(ids[i],100*p[ids[i]])),
dset.setL(x,L)
print(mylog.finished("main::train>> classifying\nrecall rate: {}%".format(correct/float(dset.size)*100)))
print(mylog.finished("main::train>> classifying\nconfusion matrix: \n{}".format(conf)))
return t, dset
# -*- coding: utf-8 -*-
"""
Created on Thu Sep 24 03:21:34 2015
@author: Wasit
"""
import pickle
from sctree import tree
import numpy as np
import codecs
with open('root.pic', 'rb') as pickleFile:
root = pickle.load(pickleFile)
#init the test tree
t=tree()
t.settree(root)
t.show()
pattern=np.zeros(10,dtype=np.int32)
p=t.classify(pattern)
newtext=unicode("")
mincode=3585
for i in xrange(100):
nextchar=p.argsort()[::-1][0]
newtext+=unichr( nextchar + mincode)
pattern=np.roll(pattern,-1)
pattern[-1]=nextchar
p=t.classify(pattern)
print 'writing text'
