Please enable JavaScript.

Coggle requires JavaScript to display documents.

MACHINE LEARNING - Coggle Diagram

- - - - TS = f: SAMPLE SPACE --> OUTPUT SPACE
        
        Learning problem: trova f(x) ragionevole per un test x
        
        All nearest neighbours (ALLnn)
        
        Proximity weighted algorithm (NWA)
        
        K-nearest neighbors (knn)
        
        Feature, Sample, Observable, Test
  - - - Recognition
      - Multitable classification
      - Association
      - Regression
- - - - A priori Probability (in a Feature)
      - Observed Probability (in a Feature)
      - ...others...
    - - Non negativity
      - Subaddictive
    - - How information grows in respect to uniform probabolity?
      - Joint information: I(X_! ... X_n)
      - Condizional information: I(X|Y)
      - Chain rule
        
        I(X,Y) = I(X) + I(Y|X)
      - Mutual information: I(X;Y)
        
        what tells me x all about y, and viceversa?
        
        Features Dependency
        
        x depends from y <=> I(X;Y)>0
        
        y funzioen di x <=> I(X;Y)=I(Y)
      - I_ts(X=a) / I_ts(X)
- - - - Error Function: in a Sample/ in TS
      - General Learning Rule: w_i = w_i + aEx_i
  - - - Delta Rule (Learning Rule)
        
        (t_j - y_s) g'(h_j)x_i
    - - Error back-Propagation (Learning rule)
- - - - x_i | H(x_i|x_1, ... , x_i)
  - - - Random Forest Classification