IE 506: Machine Learning: Principles and Techniques

Prerequisite: Strong foundations in Linear algebra, Probability and Statistics. Adequate exposure to Python Programming.

Course Description:

• Introduction to nature of machine learning tasks using motivating applications

• Supervised Learning

  Regression: Least squares regression, sparse regression

  Binary and Multi-class Classification: MAP, Minimum Misclassification Rate and Bayes Decision, Rules. Logistic regression, Naïve Bayes, k Nearest Neighbor, MLE, and Gaussian models, Linear and Quadratic Discriminant Analysis, Gaussian Mixture Models, Perceptron, Support Vector Machines, Kernel methods, Neural Networks, Classification, and Regression Trees, Decision Trees, Rule sets, Ensemble Methods: Bagging, Boosting, Random Forests. The bias-variance tradeoff, Model Selection, Cross-validation, Universal consistency, Introductory concepts of computational learning theory. Performance metrics for classification.

  Additional topics: Multi-label Classification, Ranking and Ordinal Regression, Structured, Classification, Gaussian Processes      

• Unsupervised Learning

  Dimensionality Reduction, Clustering, Density Estimation, Outlier and anomaly detection, Change Detection

• Other Topics: Semi-supervised Learning, Multi-task learning, Introduction to Reinforcement Learning.

• Implementation of machine learning algorithms and techniques: Programming using sci-kit learn machine learning toolkit and applying ML algorithms to several practical applications arising from healthcare, predictive maintenance, business analytics, decision sciences and other areas.