Machine Learning (AING356)
Definition of pattern, feature and pattern classes. Noise in pattern recognition. Decision rules, decision boundaries, discriminant functions and classifiers. Bayes classifiers, Bayes decision rules, Bayes discriminant function, minimum-error classification, minimum-risk classification, Bayes discriminant functions for Normal distributions. Nearest mean and quadratic discriminant classification. Parametric classification, maximum-likelihood and Bayesian parameter estimation. Non-parametric density estimation using Parzen window approach. Dimensionality reduction using principal component analysis and Fisher’s linear discriminant analysis. Generalized linear discriminant functions. Perceptron algorithm, its training using gradient descent. Least square method for training linear classifiers. Nonlinear classification. Solution of AND and OR problems using linear discrimination. XOR problem solution using two-layer perceptron. Training a feedforward network using backpropagation algorithm. Large margin decision boundaries. Support vector machines for large margin classification.