Pattern Classification

Weeks 1-3

• Talked about the basics of pattern classification as well as the common types of algorithms, methods, and shortcomings

• Discussed priors and their affect on classification

• Discussed Bayes Formula of likelihood x prior / evidence = posterior

• Discussed Risk and decision making and how to calculate risk.

• Discussed loss functions and how taking an action affects loss

Below are notes from weeks 1-3

Week 4

• Bayesian Decision Theory

  • Maximum likelihood Estimation

  • Bayesian Estimation

    • Gaussian & General

Week 5

• Dimensionality

  • Accuracy, Dimension, Training Sample Size

  • Computation Complexity

  • Overfitting

  • Principal Component Analysis

  • Fisher Linear Discriminant

Week 6

• Missing or Noisy Features

  • Marginal Distribution

  • Independent Noise

• Expectation Maximization

  • Observed vs latent variables

• Hidden Markov

  • For evaluation, decoding, or learning

Week 8

• Density Estimation

  • Parzen Windows

    • Shrink region V = 1 /sqrt(n)

    • Can be represented by a PNN

  • K Nearest Neighbor

    • Increase region

Week 9

• Linear Discriminant Functions

• Decision Surfaces

• Two Category Linear Separable Case

• Minimizing the Perceptron Criterion Function

• Support Vector Machine

Week 10

• Feedforward Operation and Classification

  • Activation Function

• Backpropagation Algorithm - training

  • Learning Rate, sensitivity

  • Net

• Stochastic, Batch, Online Training