本章講解了單層神經網路的學習規則。

This chapter explains the learning rulesfor a single-layer neural network.

第3章討論了多層神經網路的學習規則。

The learning rules for a multi-layer neuralnetwork are addressed in Chapter 3.

神經網路的節點（Nodes of a Neural Network）

每當我們學到東西時，大腦就儲存知識。

Whenever we learn something, our brainstores the knowledge.

計算機使用儲存器來儲存資訊。

The computer uses memory to storeinformation.

雖然大腦和儲存器都儲存資訊，但是二者的機制是非常不同的。

Although they both store information, theirmechanisms are very different.

計算機在指定位置儲存資訊，而大腦則是改變神經元之間的聯絡。

The computer stores information atspecified locations of the memory, while the brain alters the association ofneurons.

神經元本身沒有儲存能力，它只是將訊號從一個神經元傳遞到另一個神經元。

The neuron itself has no storagecapability; it just transmits signals from one neurons to the other.

大腦是由這些神經元組成的巨大網路，神經元之間的聯絡形成了特定的資訊。

The brain is a gigantic network of theseneurons, and the association of the neurons forms specific information.

神經網路模仿大腦的工作機制。

The neural network imitates the mechanismof the brain.

由於大腦是由眾多神經元的連線構成的，神經網路是由眾多節點的連線構成的，這些節點對應於大腦神經元。

As the brain is composed of connections ofnumerous neurons, the neural network is constructed with connections of nodes,which are elements that correspond to the neurons of the brain.

神經網路模仿神經元之間的聯絡使用權值，這是大腦最重要的工作機制。

The neural network mimics the neurons’association, which is the most important mechanism of the brain, using theweight value.

下表總結了大腦和神經網路之間的類比關係。

The following table summarizes the analogybetween the brain and neural network.

使用語言進一步解釋這一點可能會導致更多的混淆。

Explaining this any further using text maycause more confusion.

通過一個簡單的例子來更好地理解神經網路的機制。

Look at a simple example for a betterunderstanding of the neural network’s mechanism.

考慮一個三輸入的神經網路節點，如圖2-2所示。

Consider a node that receives three inputs,as shown in Figure 2-2.

圖2-2 三輸入的神經網路節點A node that receives threeinputs

圖中的圓圈和箭頭分別表示節點和訊號流。

The circle and arrow of the figure denotethe node and signal flow, respectively.

也就是說，神經網路的資訊以權值和偏置的形式儲存。

In other words, the information of theneural net is stored in the form of weights and bias.

來自外部的輸入訊號在到達節點之前乘以權重。

The input signal from the outside ismultiplied by the weight before it reaches the node.

——本文譯自Phil Kim所著的《Matlab Deep Learning》

更多精彩文章請關注微訊號：