def prediction(self):
"""
The structure of the network.
"""
h_conv1 = tf.nn.relu(
ne.conv2d(self.data, self.weights['W_conv1']) +
self.biases['b_conv1'])
h_pool1 = ne.max_pool_2x2(h_conv1)
h_conv2 = tf.nn.relu(
ne.conv2d(h_pool1, self.weights['W_conv2']) +
self.biases['b_conv2'])
h_pool2 = ne.max_pool_2x2(h_conv2)
h_pool2_flat = tf.reshape(h_pool2, [-1, 7 * 7 * 64])
h_fc1 = tf.nn.relu(
tf.matmul(h_pool2_flat, self.weights['W_fc1']) +
self.biases['b_fc1'])
h_fc1_drop = tf.nn.dropout(h_fc1, self.p_keep)
h_fc2 = tf.matmul(h_fc1_drop, self.weights['W_fc2']) + \
self.biases['b_fc2']
y_conv = tf.nn.softmax(h_fc2)
return y_conv
def prediction(self):
"""
The structure of the network.
"""
h_conv1 = tf.nn.relu(
ne.conv2d_valid(self.data, self.weights['W_conv1']) +
self.biases['b_conv1'])
h_pool1 = ne.max_pool_2x2(h_conv1)
h_conv2 = tf.nn.relu(
ne.conv2d_valid(h_pool1, self.weights['W_conv2']) +
self.biases['b_conv2'])
h_pool2 = ne.max_pool_2x2(h_conv2)
h_conv3 = tf.nn.relu(
ne.conv2d_valid(h_pool2, self.weights['W_conv3']) +
self.biases['b_conv3'])
h_pool3 = ne.max_pool_2x2(h_conv3)
h_pool3_flat = tf.reshape(h_pool3, [-1, 1 * 6 * 64])
h_fc1 = tf.nn.relu(
tf.matmul(h_pool3_flat, self.weights['W_fc1']) +
self.biases['b_fc1'])
h_fc1_drop = tf.nn.dropout(h_fc1, self.p_keep)
y_conv = tf.nn.softmax(
tf.matmul(h_fc1_drop, self.weights['W_fc2']) +
self.biases['b_fc2']) # softmax
return y_conv, h_pool3_flat