tf.layers.dense(
    inputs,					#層的輸入
    units,					#該層的輸出維度
    activation=None,		#激活函數(shù)
    use_bias=True,			
    kernel_initializer=None,  	# 卷積核的初始化器
    bias_initializer=tf.zeros_initializer(),  # 偏置項的初始化器
    kernel_regularizer=None,    # 卷積核的正則化
    bias_regularizer=None,    	# 偏置項的正則化
    activity_regularizer=None, 
    kernel_constraint=None,
    bias_constraint=None,
    trainable=True,
    name=None,  # 層的名字
    reuse=None  # 是否重復使用參數(shù)
)

部分參數(shù)解釋：

inputs：輸入該層的數(shù)據(jù)。

units：該層的輸出維度。

activation：激活函數(shù)。

use_bias：是否使用偏置項。

trainable=True : 表明該層的參數(shù)是否參與訓練。

示例

手寫體例子，利用兩個dense可以構成一個單層網(wǎng)絡，在下面例子中，網(wǎng)絡的神經(jīng)元個數(shù)為200。

import numpy as np
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
def compute_accuracy(x_data,y_data):
    global dense2
    y_pre = sess.run(dense2,feed_dict={xs:x_data})
    correct_prediction = tf.equal(tf.arg_max(y_data,1),tf.arg_max(y_pre,1))     #判斷是否相等
    accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))   #賦予float32數(shù)據(jù)類型，求平均。
    result = sess.run(accuracy,feed_dict = {xs:batch_xs,ys:batch_ys})   #執(zhí)行
    return result
mnist = input_data.read_data_sets("MNIST_data",one_hot = "true")
xs = tf.placeholder(tf.float32,[None,784])
ys = tf.placeholder(tf.float32,[None,10])
dense1 = tf.layers.dense(
    xs,
    200,
    activation = tf.nn.tanh,            
    kernel_initializer=tf.random_normal_initializer(mean=0, stddev=0.3),
    bias_initializer=tf.constant_initializer(0.1),
    name='fc1'
)
dense2 = tf.layers.dense(
    dense1,
    10,
    activation = tf.nn.softmax,            
    kernel_initializer=tf.random_normal_initializer(mean=0, stddev=0.3),
    bias_initializer=tf.constant_initializer(0.1),
    name='fc2'
)
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits = dense2, labels = ys),name = 'loss')
#label是標簽，logits是預測值，交叉熵。
train = tf.train.GradientDescentOptimizer(0.5).minimize(loss)
init = tf.initialize_all_variables()
with tf.Session() as sess:
    sess.run(init)
    for i in range(5001):
        batch_xs,batch_ys = mnist.train.next_batch(100)
        sess.run(train,feed_dict = {xs:batch_xs,ys:batch_ys})
        if i % 1000 == 0:
            print("訓練%d次的識別率為：%f。"%((i+1),compute_accuracy(mnist.test.images,mnist.test.labels)))

實驗結果為：