TensorFlowのMNISTチュートリアルをinput_data.pyを使わずにやってみる。

TensorFlow_MNIST_without_input_data_py.py

# coding: utf-8

# In[26]:

get_ipython().magic(u'matplotlib inline')

import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.cm as cm
#import seaborn as sns
import pandas as pd


# In[27]:

mnist = pd.read_pickle('mnist.pkl')


# In[28]:

def num2vec10(num):
    holder = np.zeros(10)
    holder[num] = 1
    return holder

def nums2vecs10(nums):
    return np.array([num2vec10(num) for num in nums])


# In[29]:

#こんなのがいっぱい入っている。
plt.imshow(mnist[1][0][1].reshape((28,28)), cmap = cm.Greys_r)


# In[30]:

sess = tf.InteractiveSession()
x = tf.placeholder('float', shape=[None, 784])
y_ = tf.placeholder('float', shape=[None, 10])


# In[31]:

W = tf.Variable(tf.zeros([784, 10]))
b = tf.Variable(tf.zeros([10]))

sess.run(tf.initialize_all_variables())


# In[32]:

y = tf.nn.softmax(tf.matmul(x,W) + b)
cross_entropy = -tf.reduce_sum(y_*tf.log(y))


# In[33]:

train_step = tf.train.GradientDescentOptimizer(0.01).minimize(cross_entropy)


# In[34]:

for i in range(1000):
    step = 50
    x_batch = mnist[0][0][i*step:(i+1)*step]
    y_batch = nums2vecs10(mnist[0][1][i*step:(i+1)*step])
    train_step.run(feed_dict={x: x_batch, y_: list(y_batch)})


# In[37]:

#Accuracy（正答率を出す。）
correct_prediction = tf.equal(tf.argmax(y,1), tf.argmax(y_,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
print accuracy.eval(feed_dict={x: mnist[1][0], y_: nums2vecs10(mnist[1][1])})