def new_acc(nn, data, label, thr = 0.5):
predict = []
for i in range(len(data)):
print nnS.forward(nn,data[i])
if np.argmax(nnS.forward(nn,data[i])) == label[i]:
predict.append(True)
else:
predict.append(False)
return 100 * np.double(len(np.where(np.array(predict) == False)[0])) / np.double(len(predict))
def accuracy(nn, data, label, thr = 0.5):
#print "start"
#print np.int8(nnS.forward(nn,data[1500].T) > .05)
#print label[1500]
acc = 0
for i in range(len(data)):
x = np.int8(nnS.forward(nn,data[i].T) > thr)
if np.array_equal(x, label_clean(label[i])):
acc += 1
return float(acc) / float(len(data))
def accuracy(nn, data, label, thr=0.5):
#print "start"
#print np.int8(nnS.forward(nn,data[1500].T) > .05)
#print label[1500]
acc = 0
for i in range(len(data)):
x = np.int8(nnS.forward(nn, data[i].T) > thr)
if np.array_equal(x, label_clean(label[i])):
acc += 1
return float(acc) / float(len(data))
