tf.layers.dense用法

先看例3吧

官方文档：
https://www.tensorflow.org/api_docs/python/tf/layers/dense

例1

import tensorflow as tf

input = tf.ones([5, 1])  # [5,1]的矩阵，5组数据，每组数据为1个。tf.layers.dense会根据这个shape，自动调整输入层单元数。
output = tf.layers.dense(input, 10)
print(output.get_shape())  # (5, 10)

input = tf.ones([3, 2])
output = tf.layers.dense(input, 10)
print(output.get_shape())  # (3, 10)

input = tf.ones([1, 7, 20])
output = tf.layers.dense(input, 10)
print(output.get_shape())  # (1, 7, 10)

input = tf.ones([1, 7, 11, 20])
output = tf.layers.dense(input, 10)
print(output.get_shape())  # (1, 7, 11, 10)

参考文献：
https://blog.csdn.net/yangfengling1023/article/details/81774580

(5, 10)
(3, 10)
(1, 7, 10)
(1, 7, 11, 10)

例2

伪代码
def add_layer(inputs, in_size, out_size, activation_function=None):
    Weights = tf.Variable(tf.random_normal([in_size, out_size]))
    biases = tf.Variable(tf.zeros([1, out_size]) + 0.1)
    Wx_plus_b = tf.matmul(inputs, Weights) + biases
    if activation_function is None:
        outputs = Wx_plus_b
    else:
        outputs = activation_function(Wx_plus_b)
    return outputs

# 二者等价
l1 = add_layer(xs, 1, 10, activation_function=tf.nn.relu)
l1=tf.layers.dense(xs,10,activation=tf.nn.relu)
# 二者等价
prediction = add_layer(l1, 10, 1, activation_function=None)
prediction=tf.layers.dense(l1,1,activation=None)

例3

一图胜千言

常用参数

tf.layers.dense(
inputs,
units, 整数或长整数，输出空间的维数
activation=None,
use_bias=True,
kernel_initializer=None,
bias_initializer=tf.zeros_initializer(),
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
trainable=True,
name=None,
reuse=None
)
返回值：一个Tensor，shape与input相同（不含最后一位），最后一位是unit参数（tf.layers.dense的第二个位置参数）的值。