圖4-4 改變分類符號的表示方法Change the class symbolsand the data is classified differently

圖4-4所示的訓練資料是我們用來訓練神經網路的。

The training data shown in Figure 4-4 iswhat we use to train the neural network.

二元分類神經網路通常採用前面討論的交叉熵函式進行訓練。

The binary classification neural networkusually adopts the cross entropy function of the previous equation fortraining.

雖然不必總是採用交叉熵函式，但它對訓練效能和實現過程是有益的。

You don’t have to do so all the time, butit is beneficial for the performance and implementation process.

二元分類神經網路的學習過程概括為以下步驟。

The learning process of the binaryclassification neural network is summarized in the following steps.

當然，我們使用交叉熵函式作為代價函式，使用sigmoid函式作為隱藏節點和輸出節點的啟用函式。

Of course, we use the cross entropyfunction as the cost function and the sigmoid function as the activation functionof the hidden and output nodes.

1. 二元分類神經網路的輸出層由一個輸出節點組成。

2. The binary classification neural networkhas one node for the output layer.

啟用函式採用sigmoid函式。

The sigmoid function is used for theactivation function.

2. 使用sigmoid函式的最大值和最小值將訓練資料的類別標誌轉換為數字表示。

3. Switch theclass titles of the training data into numbers using the maximum and minimumvalues of the sigmoid function.

3. 用適當的值初始化神經網路的權值。

4. Initialize the weights of the neuralnetwork with adequate values.

5. 將訓練資料對{輸入，正確輸出}輸入神經網路並獲得訓練輸出。

6. Enter the input from the training data {input, correct output } into the neural network and obtain the output.
   

7. 對所有訓練資料重複步驟4 - 7。

8. Repeat Steps 4 - 7 for all training datapoints.

9. 重複步驟4 - 8直到神經網路被正確訓練。

10. Repeat Steps 4-8 until the neuralnetwork has been trained properly.

雖然以上步驟看起來很複雜，但這個過程基本上與第三章的反向傳播過程相同。

Although it appears complicated because ofits many steps, this process is basically the same as that of the back-propagationof Chapter 3.

詳細的解釋就不再贅述了。

The detailed explanations are omitted.

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

更多精彩文章請關注微訊號：