当需要区分多个隐藏层时,学者们给这些深度神经网络单独命名。
When the need to distinguish multiplehidden layers arose, they gave a separate name to the deep neural network.
如图2-4所示。
See Figure 2-4.
图2-4 基于分层结构的神经网络分支示意图The branches of theneural network depend on the layer architecture
在分层神经网络中,信号进入输入层,通过隐藏层,最后通过输出层给出结果。
In the layered neural network, the signalenters the input layer, passes through the hidden layers, and leaves throughthe output layer.
在以上过程中,信号逐层前进。
During this process, the signal advanceslayer by layer.
换言之,同一层上的节点同时接收信号,并同时将处理后的信号发送到下一层。
In other words, the nodes on one layerreceive the signal simultaneously and send the processed signal to the nextlayer at the same time.
让我们根据一个简单的例子来查看输入数据在通过各层节点时的处理方式。
Let’s follow a simple example to see howthe input data is processed as it passes through the layers.
考虑图2-5所示的具有单个隐藏层的神经网络。
Consider the neural network with a singlehidden layer shown in Figure 2-5.
图2-5 单隐层神经网络A neural network with asingle hidden layer
为了方便起见,假设每个节点的激活函数是图2-6所示的线性函数。
Just for convenience, the activationfunction of each node is assumed to be a linear function shown in Figure 2-6.
图2-6 每个节点的激活函数是一个线性函数The activationfunction of each node is a linear function
现在我们将计算隐藏层的输出(如图2-7)。
Now we will calculate the output from thehidden layer (Figure 2-7).
——本文译自Phil Kim所著的《Matlab Deep Learning》
更多精彩文章请关注微信号: