In [1]:

    

import theano
from theano import tensor as T
import numpy as np
from load import mnist


    

In [2]:

    

def floatX(X):
    return np.asarray(X, dtype=theano.config.floatX)

def init_weights(shape):
    return theano.shared(floatX(np.random.randn(*shape) * 0.01))

def model(X, w):
    return T.nnet.softmax(T.dot(X, w))


    

In [8]:

    

trX, teX, trY, teY = mnist(onehot=True)


    

In [16]:

    

X = T.fmatrix()
Y = T.fmatrix()

w = init_weights((784, 10))


    

In [21]:

    

w.get_value().shape


    

    
Out[21]:





(784, 10)

In [26]:

    

py_x = model(X, w)
y_pred = T.argmax(py_x, axis=1)


    

In [27]:

    

cost = T.mean(T.nnet.categorical_crossentropy(py_x, Y))
gradient = T.grad(cost=cost, wrt=w)
update = [[w, w - gradient * 0.05]]


    

In [28]:

    

train = theano.function(inputs=[X, Y], outputs=cost, updates=update, allow_input_downcast=True)
predict = theano.function(inputs=[X], outputs=y_pred, allow_input_downcast=True)


    

In [30]:

    

for i in range(10):
    # batch : 128
    for start, end in zip(range(0, len(trX), 128), range(128, len(trX), 128)):
        cost = train(trX[start:end], trY[start:end])
    print i, np.mean(np.argmax(teY, axis=1) == predict(teX))


    

    




0 0.9159
1 0.9167
2 0.9167
3 0.9168
4 0.9174
5 0.918
6 0.9187
7 0.9187
8 0.9192
9 0.9194

In [41]:

    

print np.argmax(teY, axis=1)
print predict(teX)
print np.argmax(teY, axis=1) == predict(teX)
print np.mean(np.argmax(teY, axis=1) == predict(teX))


    

    




[7 2 1 ..., 4 5 6]
[7 2 1 ..., 4 5 6]
[ True  True  True ...,  True  True  True]
0.9194