homework 30%

question types:

  • description
  • proof
  • calc

Bayesian Decision:

  • Min Error / Min Risk
  • when Normal Distribution

non-parameter estimation

  • k-nn / parzen window, relation and difference and calc
  • EM

parameter estimation

MLE / BE, calc

svm

  • linear hard/soft margin: form / derivation
  • non-linear: kernel

cross-validation

feature extraction

  • curse of dimension
  • linear (PCA, LDA)
  • non-linear (KPCA, LLE)

feature selection

  • criterion
  • methods

clustering

  • similarity
  • k-means / spectral / hierarchical

neural networks

  • perceptron
  • multilayer, bp
  • deep: CNN / autoencoder, RNN

discriminant functions

  • linear discriminant function
  • perceptron

model selection

  • boosting