Fit Sine

Sine curve fitting example.

from miml.neural_network import FeedforwardNetwork

i = 1
x = arange(-2, 2.01, 0.1)
hiddensizes = [5,10]
for hs in hiddensizes:
    print 'Hidden layer size: %i' % hs
    net = FeedforwardNetwork(hs)
    net.layers[0].set_activation('tanh')
    for k in range(1, 6):
        t = sin(k * x)
        net.train(x, t)
        y = net.predict(x.reshape(len(x), 1))
        subplot(len(hiddensizes), 5, i)
        i += 1
        plot(x, t, 'r.', x, y, 'g-')
        ylim(-1, 1)
        xticks([])
        yticks([])