1. optimization objection (優化目標)
1.1 SVM hypothesis:
- cost1 與 cost0 均爲代價函數;
- 左圖用於正樣本,右圖適用於負樣本;
- SVM / large margin classifier:用較大間距將樣本區分開;
- 大間距分類器;
- 若 number of features >> number of training examples,則使用 logistic regression, or SVM without a kernel(linear kernel);
若 number of features < number of training examples,則使用 SVM with Gaussian kernel;
若 number of features < number of training examples,則 create/add more features, then use logistic regression or SVM without a kernel;
refenence: 《machine learning》Andrew Ng