def runscore(gen_best):
score = 0.0
for gen in gen_best:
genout = test_generator(gen, cf['test_times'])
gentxt = "".join([pick_letter(cf['alphabet'], out) for out in genout])
score += abbafit(gentxt)
return score / len(gen_best)
def evalstep(classifier, teststring, times, balph): correct = 0.0 testout = '' for out in classifier.run_signal(teststring): ac = out[0][0,0] > 0 correct += 1.0/times if ac else 0.0 testout += '+' if ac else '-' refin = ''.join([helpers.pick_letter(balph, t) for t in teststring]) return (correct, testout, refin)
def teststep(gen, cla, ref_data):
genout = test_generator(gen, cf['test_times'])
# print genout
gentxt = [pick_letter(cf['alphabet'], out) for out in genout]
gcltxt = test_classifier(cla, genout)
score = ""
if cf['datasource'] == 'abba':
score = " score: " + str(abbafit(''.join(gentxt)))
print "".join(gentxt) + score
print "".join(gcltxt)
corin = [c for c in ref_data.next()]
corin = corin[:min(cf['test_times'], len(corin))]
# corin = []
# for c in cortxt:
# corin.append(charvectors[alphabet.index(c)])
# corin = [charvectors[alphabet.index(c)] for c in cortxt]
# print corin
cortxt = [pick_letter(cf['alphabet'], c) for c in corin]
ccltxt = test_classifier(cla, corin)
print "".join(cortxt)
print "".join(ccltxt)
c = c.lower() if c in balph: yield c yield " " def vcs(): for c in charstream(): yield np.mat([[1.0 if d == c else -1.0] for d in balph]) teacher = [v for (v,_) in zip(vcs(), xrange(40000))] RESSIZE = 400 INSIZE = 40 # __init__(self, ressize, insize, outsize, sparseness, scale): gen = resnet.ResGen(RESSIZE, INSIZE, len(balph), 0.3, 0.5) #print gen.outnet gen.teacher_forced(teacher, 80) print ".. after training:" #print gen.outnet testout = "" for i in xrange(100): testout += helpers.pick_letter(balph, gen.update_step()[0]) print testout
# while True: # yield np.mat([[1.0],[-1.0]]) TIMES = 9000 RESSIZE = 600 # ressize, insize, sparseness, scale classifier = resnet.ResClass(RESSIZE, len(balph), 0.3, 0.5) classifier.pretrain(vcs(), resnet.extremeRI(len(balph)), TIMES, 80) teststring = [v for (v,_) in zip(vcs(), xrange(100))] testout = '' for out in classifier.run_signal(teststring): testout += '+' if out[0][0,0] > 0 else '-' print ''.join([helpers.pick_letter(balph, t) for t in teststring]) print testout teststring = [v for (v,_) in zip(resnet.extremeRI(len(balph)), xrange(100))] testout = '' for out in classifier.run_signal(teststring): testout += '+' if out[0][0,0] > 0 else '-' print ''.join([helpers.pick_letter(balph, t) for t in teststring]) print testout
