Please enable JavaScript.

Coggle requires JavaScript to display documents.

機器學習與基礎演算法 (通用技術 (資料預處理 (特徵轉換 (Thresholding, Apply Log/ Exp. function,…

- - - - Thresholding
      - Apply Log/ Exp. function
      - Standardize
      - Scaling
    - - Under-Sample
        
        Random Under-Sample
      - Over-Sample
        
        Smote
  - - - 是模型與真實狀況之間的誤差 (與validation error相關)
    - - Variance是模型本身的不穩定度 (與training error相關)
    - - noise代表資料本身的錯誤(有時無解)
  - - - SVD (Singular Value Decomposition)
      - t-SNE (T-Distributed Stochastic Neighbor Embedding)
      - PCA (Principal component analysis)
- - - - 預測的數值型態
        
        Catergorical(類別型資料)
        
        分類問題(Classification)
        
        K-nearest neighbors(KNN)
        
        不需要訓練，但預測極端費時
        
        Logistic Regression
        
        Likelihood and Log-Likelihood
        
        -1 * Log-Likelihood = Cross entropy loss
        
        Cross entropy loss
        
        Support Vector Classification (SVC)
        
        Kernel選擇
        
        "rbf" / "poly" = 用線性分類實現非線性分類
        
        "linear" = 線性分類
        
        Decision Tree (決策樹)
        
        Ensemble Learning (集成學習)
        
        Bagging
        
        1 more item...
        
        Boosting
        
        2 more items...
        
        一定可以Overfitting
        
        Numerical(數值型資料)
        
        回歸問題(Regression)
        
        One-Hot encoding
        
        One-hot Encoding可以用regression
        實現分類問題
        
        Linear Regression (線性回歸)
        
        衍生產物
        
        Gradient Descent (梯度下降) (GD)
        
        3 more items...
        
        計算RMSE
        
        Regularized Regression (正規化回歸)
        
        Penalty (懲罰項)
        
        Ridge (L2)
        
        1 more item...
        
        Lasso (L1)
        
        2 more items...
        
        Support Vector Regression (SVR)
        
        Tube
        
        計算Tube (寬度 = margin的距離)外數據的誤差
  - - - Clutering
        
        K-means
        
        Hierarchical Clustering