Why does adding another layer to a simple Tensorflow neural network example break it?

Question

Here is a basic example of a Tensorflow network (based on MNIST), a complete code that gives approximately 0.92 accuracy:

import numpy as np
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)

x = tf.placeholder(tf.float32, [None, 784])

W = tf.Variable(tf.zeros([784, 10]))
b = tf.Variable(tf.zeros([10]))
y = tf.nn.softmax(tf.matmul(x, W) + b)

y_ = tf.placeholder(tf.float32, [None, 10])

cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(y), reduction_indices=[1]))
train_step = tf.train.GradientDescentOptimizer(0.5).minimize(cross_entropy)
sess = tf.InteractiveSession()
tf.global_variables_initializer().run() # or 
tf.initialize_all_variables().run()

for _ in range(1000):
  batch_xs, batch_ys = mnist.train.next_batch(100)
  sess.run(train_step, feed_dict={x: batch_xs, y_: batch_ys})

correct_prediction = tf.equal(tf.argmax(y,1), tf.argmax(y_,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
print(sess.run(accuracy, feed_dict={x: mnist.test.images, y_: mnist.test.labels}))

Question: Why does adding an extra layer, as in the code below, make it much worse than its accuracy to 0.11?

W = tf.Variable(tf.zeros([784, 100]))
b = tf.Variable(tf.zeros([100]))
h0 = tf.nn.relu(tf.matmul(x, W) + b)

W2 = tf.Variable(tf.zeros([100, 10]))
b2 = tf.Variable(tf.zeros([10]))
y = tf.nn.softmax(tf.matmul(h0, W2) + b2)

Answer 1

In the example, weights are incorrectly initialized, but without a hidden layer, an effective linear softmax regression is obtained, which the demo does not affect this choice. Setting them all to zero is safe, but only for a single layer network.

, . , .

:

W = tf.Variable(tf.random_uniform([784, 100], -0.01, 0.01))
b = tf.Variable(tf.zeros([100]))
h0 = tf.nn.relu(tf.matmul(x, W) + b)

W2 = tf.Variable(tf.random_uniform([100, 10], -0.01, 0.01))
b2 = tf.Variable(tf.zeros([10]))
y = tf.nn.softmax(tf.matmul(h0, W2) + b2)

, , , - , . , . , , - lockstep, , ( ), .

Answer 2

, , , .

, , , 0. , relu(Wx + b) = 0 W = 0 b = 0. - .

, , 1 1 . , 0.11 ( , 0.10).