test.py

import math as math
import numpy as np
from chainer import Function, FunctionSet, gradient_check, Variable, optimizers
import chainer.functions as F

model = FunctionSet(
    l1 = F.Linear(1, 10),
    l3 = F.Linear(10, 1),
)

def forward(x_in,y_in):
    x = Variable(x_in)
    t = Variable(y_in)
    h1 = F.relu(model.l1(x))
    y = model.l3(h1)
    f = x*y
    return F.mean_squared_error(f, t), f

def sol(x):
    return math.exp(-(x/5.0))*math.sin(x)

optimizer = optimizers.SGD()
optimizer.setup(model.collect_parameters())

xx = np.zeros( (21,1), dtype = np.float32 )
yy = np.zeros( (21,1), dtype = np.float32 )

for i in xrange(21):
    x = (i) * 0.1
    y = sol(x)
    xx[i][0] = x
    yy[i][0] = y

for epoch in xrange(25000):
    optimizer.zero_grads()
    loss, yyy = forward(xx, yy)
    loss.backward()
    optimizer.update()