Please enable JavaScript.

Coggle requires JavaScript to display documents.

Machine Learning (ML) (Frameworks (sci-kit learn (Python based, Builds on…

- - - - Metrics
        
        F1 score
        
        also works for unbalanced data sets
        
        Accuracy alternative
        
        False predictions (F)
        
        False Negatives (FN)
        
        Type II error
        
        False Positives (FP)
        
        Type I error
        
        True predictions (T)
        
        True Positives (TP)
        
        True Negatives (TN)
        
        Accuracy = T/(T+F)
        
        only reliable for balanced data sets
        
        Precision
        
        TP / (TP + FP)
        
        Recall
        
        TP / (TP + FN)
      - Diagrams
        
        Confusion matrix
        
        Receiver Operating Characteristic (ROC)
        
        Precision-Recall
    - - Legal issues?
  - - - Output = continuous (number)
    - - Output = discrete (class)
      - Binary (+ / -)
      - Multi
- - - - Anaconda
      - Enthought Canopy
- - - - Boosting
      - Voting
      - Bagging
        
        Random Forests
      - Stacking
    - - Maximum a posteriori (MAP) & Maximum Likelihood (ML)
      - Naive Bayes
      - Markov Networks
      - Belief Networks
    - - ProLog
    - - k-Nearest Neighbours (kNN)
    - - k-means
      - Expectation Maximization (EM)
      - Dimensionality reduction
        
        PCA
        
        t-SNE
  - - - Some expected output given
    - - Expected output given
    - - No expected output given
    - - State
      - Environment
      - Agent
      - Reward
      - Example: play 2D Mario game