From f3079789f4e5c605abcf24fc0cdc750bbe2612b8 Mon Sep 17 00:00:00 2001 From: varyshare Date: Fri, 1 Mar 2019 18:02:15 +0800 Subject: [PATCH] =?UTF-8?q?Update=2016.=E5=BC=BA=E5=8C=96=E5=AD=A6?= =?UTF-8?q?=E4=B9=A0.md?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit 173 hidden = fully_connected(X, n_hidden, activation_fn=tf.nn.elu,weights_initializer=initializer) # 隐层激活函数使用指数线性函数 173行应该单独成行，而不应当跟在上一句的后面 --- "docs/16.\345\274\272\345\214\226\345\255\246\344\271\240.md" | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git "a/docs/16.\345\274\272\345\214\226\345\255\246\344\271\240.md" "b/docs/16.\345\274\272\345\214\226\345\255\246\344\271\240.md" index 10bc836..33f5958 100644 --- "a/docs/16.\345\274\272\345\214\226\345\255\246\344\271\240.md" +++ "b/docs/16.\345\274\272\345\214\226\345\255\246\344\271\240.md" @@ -169,7 +169,8 @@ n_hidden = 4 # 这只是个简单的测试，不需要过多的隐藏层 n_outputs = 1 # 只输出向左加速的概率 initializer = tf.contrib.layers.variance_scaling_initializer() # 2. 建立神经网络 -X = tf.placeholder(tf.float32, shape=[None, n_inputs]) hidden = fully_connected(X, n_hidden, activation_fn=tf.nn.elu,weights_initializer=initializer) # 隐层激活函数使用指数线性函数 +X = tf.placeholder(tf.float32, shape=[None, n_inputs]) +hidden = fully_connected(X, n_hidden, activation_fn=tf.nn.elu,weights_initializer=initializer) # 隐层激活函数使用指数线性函数 logits = fully_connected(hidden, n_outputs, activation_fn=None,weights_initializer=initializer) outputs = tf.nn.sigmoid(logits) # 3. 在概率基础上随机选择动作 -- GitLab