神经网络各子模块设计

 

平均池化层

def avgpool(x_tensor, pool_ksize, pool_strides):
    return tf.nn.avg_pool(x_tensor, ksize = [1, pool_ksize[0], pool_ksize[1], 1], strides = [1, pool_strides[0], pool_strides[1], 1], padding = 'VALID')

def conv2d(x_tensor, conv_num_outputs, conv_ksize, conv_strides, name):
    # TODO: Implement Function
    x_shape = x_tensor.get_shape().as_list()
    regularizer = tf.contrib.layers.l2_regularizer(scale = 0.0001)
    n = conv_ksize[0] * conv_ksize[1] * conv_num_outputs
    with tf.variable_scope(name):
        weights = tf.get_variable('conv_weights', shape = [conv_ksize[0], conv_ksize[1], x_shape[3], conv_num_outputs],initializer = tf.random_normal_initializer(stddev = np.sqrt(2.0 / n)),regularizer = regularizer)
        L = batch_norm(x_tensor,'bn')
        L = relu(L)
        L = tf.nn.conv2d(L, weights, strides = [1, conv_strides[0], conv_strides[1], 1], padding = 'SAME')
    return L

def conv2d_nobn(x_tensor, conv_num_outputs, conv_ksize, conv_strides, name):
    # TODO: Implement Function
    x_shape = x_tensor.get_shape().as_list()
    regularizer = tf.contrib.layers.l2_regularizer(scale = 0.0001)
    n = conv_ksize[0] * conv_ksize[1] * conv_num_outputs
    with tf.variable_scope(name):
        weights = tf.get_variable('conv_weights', shape = [conv_ksize[0], conv_ksize[1], x_shape[3], conv_num_outputs],
                              initializer = tf.random_normal_initializer(stddev = np.sqrt(2.0 / n)),
                              regularizer = regularizer)
        L = tf.nn.conv2d(x_tensor, weights, strides = [1, conv_strides[0], conv_strides[1], 1], padding = 'SAME')
    return L

def output(x_tensor, num_outputs):
    #全连接层
    x_shape = x_tensor.get_shape().as_list()
    regularizer = tf.contrib.layers.l2_regularizer(scale = 0.0001)

    weights = tf.get_variable('weight', shape = [x_shape[1], num_outputs],
                              initializer = tf.uniform_unit_scaling_initializer(factor = 1.0),
                              regularizer = regularizer)
    bias = tf.Variable(tf.zeros([num_outputs]))
    out = tf.add(tf.matmul(x_tensor, weights), bias)

    return out