Courses

Here are links to course material for courses that have been/will be offered by me at IIT Kanpur.

[New Course] Statistical Learning Theory

Note: this course will be offered in the Winter 2017-18 semester.

Description: This is intended to be a course on advanced techniques used in the design and analysis of machine learning and statistical estimation algorithms. The course is divided into two broad parts, one that primarily looks at the statistical analysis of learning and estimation algorithms, and one that explores a variety of algorithm design techniques currently popular in machine learning and statistics communities. The course will involve an intense application of probabilistic models and techniques and will benefit from prior familiarity with machine learning/signal processing as a source of basic learning theoretic concepts, as well as motivation for large-scale learning and optimization.

Prerequisites: Fluency in basic results in probability and statistics would be essential. Prior exposure to machine learning or signal processing techniques would be desirable.

Topics: Topics in statistical learning theory (PAC learning, uniform convergence, stability, consistency), algorithmic learning (boosting, non-convex and online optimization, sampling) and additional topics based on interest. Please see the course website for details.

Course Website

  • SLT (Mini-course offered in Winter 2016-17) [link]

[New Course] Optimization Techniques

Note: this course will be offered in the Autumn 2016-17 semester.

Description: The area of optimization plays a critical role in several contemporary areas such as decision and control, signal processing, and machine learning. This course intends to present a thorough treatment of optimization techniques with specific emphasis on modern applications. This will provide students with a sound background in the area and benefit those who wish to pursue doctoral or master level theses in this subject, or apply these techniques to their own areas.

Prerequisites: Familiarity with basics of real analysis, linear algebra, and probability and statistics. Prior exposure to fields such as machine learning, signal processing, operations research, or decision and control, would be valuable as source of, and motivation for, optimization problems. Please see the course website for details.

Topics: Preliminaries, first order methods, second order methods, stochastic optimization, non convex optimization. Please see the course website for details.

Course Website

  • CS774 (Autumn 2016-17) [link]

[New Course] Online Learning and Optimization

Description: This is intended to be a course on advanced techniques used in optimization and learning. The course would cover topics in sequential prediction, and optimization on large-scale streaming data. These are the methods of choice for massive learning tasks and form the basis for a large family of optimization and learning routines.

Prerequisites: Familiarity with basics of probability and statistics, and linear algebra would be essential. Prior exposure to machine learning would be desirable. Please see the course website for details.

Topics: Online prediction with full and partial feedback, online convex optimization. Please see the course website for details.

Course Website

  • CS773 (Winter 2015-16) [link]