from metanet.networks.artificial_networks.feedforward_network import FeedforwardNetwork
from metanet.datasets import mnist
import argparse
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("-t", "--train", default="true")
args = parser.parse_args()
inp, tgt = mnist.get_mnist(10)
net = FeedforwardNetwork(len(inp[0]), [len(inp[0])], len(tgt[0]))
if args.train == "true":
for i in range(50):
print(" [*] Epoch:", i, "error:", net.train(inp, tgt, 1))
else:
net.load_net("./data/net.pkl")
err = 0
for i in range(len(tgt)):
ans = net.test(inp[i])
if tgt[i].index(max(tgt[i])) != ans.index(max(ans)):
err += 1
if args.train == "true":
net.save_net("./data/net.pkl")
print(" [*]", err, "errors")
