def main():
f = "/Users/gabriel/Desktop/instances.txt"
base_code = ord("A")
with open(f) as f:
lines = f.readlines()
s = int(lines[0]) * int(lines[1])
n = int(lines[2])
l = 4
instances = list()
while n > 0:
n -= 1
pattern = np.zeros((1, s))
pattern_class = np.zeros((1, int(lines[2])))
pt_index = 0
inst = Instance()
for i in range(int(lines[1])):
for c in lines[l].strip():
pattern[0][pt_index] = 1 if c == "#" else 0
pt_index += 1
l += 1
pattern_class[0][ord(lines[l].strip()) - base_code] = 1
l += 1
inst.attributes = pattern
inst.output_values = pattern_class
inst.normalize()
instances.append(inst)
ann = NeuralNet([35, 35, 26], True)
ann.learning_rate = 0.5
ann.momentum = 0.4
ann.instances(instances)
ann.train(55)
correct = 0
for i in instances:
ann.instance(i)
ann.feed_forward()
f = chr(65 + np.argmax(i.output_values))
print "F={}".format(f)
f_hat = chr(65 + np.argmax(ann.output))
print "F'={}".format(f_hat)
if f == f_hat:
correct += 1
else:
print "ERROR!"
print
print str(correct / float(len(instances)) * 100) + " %"
for element in input_values:
letter = element[0]
out = np.zeros((1, 26))
out[0, ord(letter) - 65] = 1
element = np.delete(element, 0)
element = element.astype(np.float)
element = element.reshape((1, element.shape[0]))
inst = Instance()
inst.attributes = element
inst.output_values = out
inst.normalize()
instances.append(inst)
print "...Ready!"
print "Instantiate neuralnet..."
ann = NeuralNet([16, 32, 26], True)
ann.learning_rate = 0.09
ann.momentum = 0.44
ann.instances(instances)
print "...Ready!"
print "Generate model..."
ann.train(50, True)
print "...Ready!"
print ann
