Theano 初探（二）随机数与梯度下降logistic模型

theano随机数生成：
由RandomStreams 定义生成的随机数流，在相应的函数中如设定no_dafault_updates = True
会相应的使得随机数generator不更新randomState 即产生相同的随机数生成结果。

RandomStreams所定义的随机分布生成对象在调用rng.get_value或set_value方法时
有参数borrow其与return_internal_type 设定返回的RandomState类型是否为
int类型的别名，不要在意这些细节。
与numpy RandomState相似，不同含有随机变量的函数的调用，一个的调用会更新
另一个函数的RandomState，这里也给出了拷贝RandomState并在两个函数间
设定的例子。

from theano.tensor.shared_randomstreams import RandomStreams 
from theano import function 
srng = RandomStreams(seed = 234)
rv_u = srng.uniform((2, 2))
rv_n = srng.normal((2, 2))
f = function([], rv_u)
g = function([], rv_n, no_default_updates = True) 
nearly_zeros = function([], rv_u + rv_u - 2 * rv_u)

f_val0 = f() 
f_val1 = f() 

print f_val0, f_val1

g_val0 = g() 
g_val1 = g()

print g_val0, g_val1

rng_val = rv_u.rng.get_value(borrow = True)

#print rng_val

rng_val.seed(89234)
rv_u.rng.set_value(rng_val, borrow = True)

state_after_v0 = rv_u.rng.get_value().get_state() 
nearly_zeros() 
v1 = f() 
rng = rv_u.rng.get_value() 
rng.set_state(state_after_v0)
rv_u.rng.set_value(rng, borrow = True)
v2 = f() 
v3 = f()

print v1 
print v2 
print v3

import theano
from theano.sandbox.rng_mrg import MRG_RandomStreams 

class Graph():
 def __init__(self, seed = 123):
  self.rng = RandomStreams(seed)
  self.y = self.rng.uniform(size = (1,))

g1 = Graph(seed = 123)
f1 = theano.function([], g1.y)

g2 = Graph(seed = 987)
f2 = theano.function([], g2.y)

print f1() 
print f2()

def copy_random_state(g1, g2):
 if isinstance(g1.rng, MRG_RandomStreams):
  g2.rng.rstate = g1.rng.rstate
 for (su1, su2) in zip(g1.rng.state_updates, g2.rng.state_updates):
  su2[0].set_value(su1[0].get_value())

copy_random_state(g1, g2)
print f1() 
print f2()

下面有一个利用梯度下降法logistic回归的例子，这里数据是无相依关系的，仅仅用来进行
过拟合示例：

import numpy 
import theano 
import theano.tensor as T 
rng = numpy.random 

N = 400 
feats = 784 

D = (rng.randn(N, feats), rng.randint(size = N, low = 0, high = 2))
training_steps = 10000 

x = T.dmatrix('x')
y = T.dvector('y')

w = theano.shared(rng.randn(feats), name = 'w')
b = theano.shared(0., name = 'b')

print "Initial model:"
print w.get_value()
print b.get_value()

p_1 = 1 / (1 + T.exp(-T.dot(x, w) - b))
prediction = p_1 > 0.5 
xext = -y * T.log(p_1) - (1 - y) * T.log(1 - p_1) 
cost = xext.mean() + 0.01 * (w ** 2).sum() 
gw, gb = T.grad(cost, [w, b])

train = theano.function(inputs = [x, y], outputs = [prediction, xext], updates = ((w, w - 0.1 * gw), (b, b - 0.1 * gb)))
predict = theano.function(inputs = [x], outputs = prediction)

for i in range(training_steps):
 pred, err = train(D[0], D[1])

print "Final model:"
print w.get_value()
print b.get_value()
print "target values for D:"
print D[1]
print "prediction on D:"
print predict(D[0])