chainer で XOR の 出力 の 推移 を ffmpeg で mp4 にする

XOR2.py

#! /usr/bin/env python
#encoding: utf-8
import numpy as np
import chainer.functions as F
from chainer import FunctionSet, Variable, optimizers

import matplotlib.pyplot as plt
import matplotlib.animation as animation
fig = plt.figure()
ax = fig.add_subplot(111, aspect='equal')
ims = []

model= FunctionSet(
	l1 = F.Linear(2, 2),
	l2 = F.Linear(2, 1)
)

def forward(x):
	return F.sigmoid(model.l2(F.sigmoid(model.l1(x))))

def calc(x_data):
	x = Variable(x_data.reshape(1,2).astype(np.float32), volatile=False)
	h = forward(x)
	return h

def train(x_data, y_data):
	h = calc(x_data)
	y = Variable(y_data.reshape(1,1).astype(np.float32), volatile=False)
	optimizer.zero_grads()
	error = F.mean_squared_error(h, y)
	error.backward()
	optimizer.update()
	return error.data

def save():
	Writer = animation.writers['ffmpeg']
	writer = Writer(fps=30, metadata=dict(artist='Me'), bitrate=1800)
	ani = animation.ArtistAnimation(fig, ims, interval=50, repeat_delay=1000)
	#動画として保存
	ani.save('im.mp4', writer=writer)
	#plt.show()
	
def draw():
	x = np.arange(0, 1.05, 0.05)
	y = np.arange(0, 1.05, 0.05)
	z= []
	for ix in x:
		yz= []
		for iy in y:
			h = calc(np.array([ix, iy]))
			yz.append(h.data[0][0])
		z.append(yz)
	X, Y = np.meshgrid(x, y)
	Z= np.array(z)
	im = plt.pcolor(X, Y, Z)
	#im = plt.colorbar()
	return im

#optimizer = optimizers.AdaDelta(rho=0.95, eps=1e-06)
#optimizer = optimizers.AdaGrad(lr=0.001, eps=1e-08)
#optimizer = optimizers.Adam(alpha=0.001, beta1=0.9, beta2=0.999, eps=1e-08)
#optimizer = optimizers.MomentumSGD(lr=0.01, momentum=0.9)
#optimizer = optimizers.NesterovAG(lr=0.01, momentum=0.9)
optimizer = optimizers.RMSprop(lr=0.01, alpha=0.99, eps=1e-08)
#optimizer = optimizers.SGD(lr=0.01)

optimizer.setup(model)

data_xor = [
	[np.array([0.25, 0.25]), np.array([0])],
	[np.array([0.25, 0.75]), np.array([1])],
	[np.array([0.75, 0.25]), np.array([1])],
	[np.array([0.75, 0.75]), np.array([0])],
]
N = len(data_xor)

print "###学習前###"
for j in range(0, N):
	x, t= data_xor[j]
	h = calc(x)
	print "{} -> {} : {}".format(x, h.data, t)

#学習
err= []
for k in range(0, 500):
	print "frame {}".format(k)
	for i in range(0, 10):
		perm = np.random.permutation(N)
		s= 0;
		for j in range(0, N):
			x, t= data_xor[perm[j]]
			s+= train(x, t)
		err.append(s)
	im= draw()
	ims.append([im])

print "###学習後###"
for j in range(0, N):
	x, t= data_xor[j]
	h = calc(x)
	print "{} -> {} : {}".format(x, h.data, t)

print "###テスト###"
test_xor = [
	[np.array([0, 0]), np.array([0])],
	[np.array([0, 1]), np.array([1])],
	[np.array([1, 0]), np.array([1])],
	[np.array([1, 1]), np.array([0])],
]
for j in range(0, len(test_xor)):
	x, t= test_xor[j]
	h = calc(x)
	print "{} -> {} : {}".format(x, h.data, t)

save()

#誤差の推移
if (False):
	plt.plot(err)
	plt.show()