L2正則化—tensorflow實現

L2正則化是一種減少過擬合的方法，在損失函數中加入刻畫模型複雜程度的指標。假設損失函數是J(θ) ，則優化的是J(θ)+λR(w) ，R(w)=∑ni=0|w2i| 。

在tensorflow中的具體實現過程如下：

#coding:utf-8

import tensorflow as tf

def get_weight(shape,lambda):
    var = tf.Variable(tf.random_normal(shape),dtype=tf.float32)

    tf.add_to_collection("losses",tf.contrib.layers.l2_regularizer(lambda)(var))#把正則化加入集合losses裏面

    return var

x = tf.placeholder(tf.float32,shape=(None,2))
y_ = tf.placeholder(tf.float32,shape=(none,1))#真值

batcg_size = 8

layer_dimension = [2,10,10,10,1]#神經網絡層節點的個數

n_layers = len(layer_dimension)#神經網絡的層數

cur_layer = x

in_dimension = layer_dimension[0]

for i in range (1,n_layers):
    out_dimension = layer_dimension[i]
    weight = get_weight([in_dimension,out_dimension],0.001)
    bias = tf.Variable(tf.constant(0.1,shape(out_dimension)))

    cur_layer = tf.nn.relu(tf.matmul(x,weight)) + bias)

    in_dimension = layer_dimension[i]


ses_loss = tf.reduce_mean(tf.square(y_ - cur_layer))#計算最終輸出與標準之間的loss

tf.add_to_collenction("losses",ses_loss)#把均方誤差也加入到集合裏

loss = tf.add_n(tf.get_collection("losses"))
#tf.get_collection返回一個列表,內容是這個集合的所有元素
#add_n()把輸入按照元素相加