HCS6349 - Statistical Machine Learning

HCS 6349 (ACN 6349) Statistical Machine Learning (3 semester credit hours) Mathematical tools for investigating the asymptotic behavior of both batch and adaptive machine learning algorithms including convergence of gradient descent batch learning algorithms convergence of adaptive stochastic approximation learning algorithms, and convergence of Monte Carlo Markov Chain algorithms. M-estimation and bootstrap asymptotic statistical theory for characterizing asymptotic behavior of parameter estimates as a function of sample size to support model selection, specification analysis, and hypothesis testing. Emphasizes applications of theory to unsupervised, supervised, and reinforcement learning machines and deep learning with MATLAB software implementations. Prerequisites: (ACN 6348 or HCS 6348) and department consent required. (3-0) T