-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.py
34 lines (27 loc) · 903 Bytes
/
main.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
import network as net
from layers import denseLayer, dropout
import activations as af
import costFunctions as cst
import data_parser
def func():
model = net.Model(cst.MultiClassCrossEntropy)
model.sequential([
denseLayer.DenseLayer(400, 100, af.ReLU),
dropout.Dropout(0.9),
denseLayer.DenseLayer(100, 50, af.ReLU),
dropout.Dropout(0.9),
denseLayer.DenseLayer(50, 20, af.ReLU),
dropout.Dropout(0.9),
denseLayer.DenseLayer(20, 10, af.Softmax),
])
#model.load_network("Weights.pkl")
X, Y, y = data_parser.parse_400()
X, Y = X.T, Y.T
pivot = 4000
X_train, X_test = X[:, :pivot], X[:, pivot:]
Y_train, Y_test = Y[:, :pivot], Y[:, pivot:]
model.train(0.1, 100, X_train, Y_train, X_test, Y_test)
model.evaluate(X_test, Y_test)
model.save_network("Weights.pkl")
if __name__ == '__main__':
func()