圖2-9 該示例神經網路等價於單層神經網路This example neuralnetwork is equivalent to a single layer neural network

請記住，當隱藏節點具有線性啟用函式時，隱藏層實際上是無效的。

Keep in mind that the hidden layer becomesineffective when the hidden nodes have linear activation functions.

然而，輸出節點可以、有時也必須採用線性啟用函式。

However, the output nodes may, andsometimes must, employ linear activation functions.

神經網路的有監督學習（Supervised Learning of a Neural Network）

本節介紹神經網路監督學習的概念和過程。

This section introduces the concepts andprocess of supervised learning of the neural network.

在第1章的“機器學習”我們對此概念有所提及。

It is addressed in the “Types of MachineLearning” section in Chapter 1.

在許多訓練方法中，本書只包括有監督學習的內容。

Of the many training methods, this bookcovers only supervised learning.

因此，針對神經網路也只討論有監督學習。

Therefore, only supervised learning isdiscussed for the neural network as well.

在圖中的描述裡，神經網路的監督學習按以下步驟進行：

In the big picture, supervised learning ofthe neural network proceeds in the following steps:

1. 用適當的值初始化權重。

2. Initialize the weights with adequatevalues.

3. 從訓練資料中獲取“輸入”，對應的格式為{輸入，正確輸出}，並將其輸入神經網路。

4. Take the“input” from the training data, which is formatted as { input, correct output}, and enter it into the neural network.

從神經網路獲得輸出，並與正確的輸出比較計算誤差。

Obtain the output from the neural networkand calculate the error from the correct output.

3. 調整權重以減小誤差。

4. Adjust the weights to reduce the error.

5. 對所有訓練資料重複步驟2-3。

6. Repeat Steps 2-3 for all training data.

以上步驟基本上與“機器學習”的監督學習過程相同。

These steps are basically identical to thesupervised learning process of the “Types of Machine Learning” section.

因為監督學習的訓練是一個修改模型以減小正確輸出與模型輸出之間的差異的過程。

This makes sense because the training ofsupervised learning is a process that modifies the model to reduce thedifference between the correct output and model’s output.

唯一的區別是，模型的修改變成了神經網路的權值變化。

The only difference is that themodification of the model becomes the changes in weights for the neuralnetwork.

圖2-10說明了迄今為止已經解釋過的有監督學習的概念。

Figure 2-10 illustrates the concept ofsupervised learning that has been explained so far.

圖2-10 有監督學習的概念模型圖The concept of supervisedlearning

這將有助於你清楚地理解前面所描述的步驟。

This will help you clearly understand thesteps described previously.

單層神經網路的訓練：增量規則（Training of a Single-Layer Neural Network: Delta Rule）

如前所述，神經網路以加權的形式儲存資訊。

As previously addressed, the neural networkstores information in the form of weights.

因此，為了訓練具有新資訊的神經網路，應該相應地改變權重。

Therefore, in order to train the neuralnetwork with new information, the weights should be changed accordingly.

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

更多精彩文章請關注微訊號：