Lectures and Supporting material
Homework and Solution
Schedule
Mon and Wed, 1:30pm - 2:45pm, (Anderson Wing SEC, Room 212)
Course Schedule
Instructor
Prof. Usman Khan
574 Boston Ave, Office 318.B
Email: khan AT ece DOT tufts DOT edu
Office Hours: 12-1pm, Mon and Wed, or by email
TA
Mohammad Panahazari
Office hours: TBA
Course description
(Cross-listed as CS 144) Design and analysis of modern machine learning methods with emphasis on convex and nonconvex problems, and centralized, federated, and distributed computational architectures. Topics include convergence, complexities, contractions, fixed point theorems, and perturbation techniques; gradient descent and stochastic gradient descent in addition to accelerated methods including Polyak and Nesterov momentum, minibatching, and variance reduction. State of the practice methods will be covered including Adagrad, Autogard, Adam, sgdm with applications in image classification and document clustering.
Recommendations: MATH 70 and CS 11
Course structure
HW and Reading assignments (25%)
Midterm exam 1 (25%)
Midterm exam 2 (25%)
Final project (25%)
Helpful references
Lectures on Convex Optimization, 2nd Ed, Yurii Nesterov, Springer, 2018
Introducotry Lectures on Convex Optimziatio, A Basic Course, Yurii Nesterov, Springer, 2003.
Principles of Mathematical Analysis, 3rd Ed, W. Rudin, McGraw Hill, 1976
Matrix Analysis, 2nd Ed, R. A. Horn and C. R. Johnson, Cambridge University Press, New York, NY, 2013
Matrix Computations, G. H. Golub and C. F. Van Loan, Johns Hopkins Press, 1996
Statistical Signal Processing, L. Scahrf, Addison-Wesley, 1991
Fundamentals of Statistical Signal Processing: Estimation Theory, S. M. Kay, Prentice Hall, 1993
Fundamentals of Statistical Signal Processing: Detection Theory, S. M. Kay, Prentice Hall, 1998
Mathematical Statistics, P. J. Bickel and K. A. Docksum, Prentice Hall, 2001