ECE586BH: Interplay between Control and Machine Learning (Fall 2023)
Course Information
Course DescriptionAdvanced graduate course focuses on interplay between control and machine learning. One half of the course focuses on tailoring control tools to study algorithms and network models in large-scale machine learning. In the other half of the course, students will study how to combine machine learning and model-based control methods for control design problems. The following topics will be covered: empirical risk minimization; first-order methods for large-scale machine learning; stochastic optimization; dissipation inequality; jump system theory; Lur’e-Postnikov type Lyapunov functions; integral quadratic constraints; KYP Lemma; graphical interpretations for optimization methods; control-theoretic methods for designing certifiably robust neural networks; control-oriented analysis tools for temporal difference learning and Q-learning; reinforcement learning for linear quadratic regulator (LQR) problems; zeroth-order optimization and evolutionary strategies; policy gradient methods for robust control; Goldstein’s subgradient method for H-infinity control; adversarial reinforcement learning; imitation learning for control; regularization of model-free control via prior model-based design; constrained policy optimization.Required MaterialsThere is no required textbook for the class. All course material will be presented in class and/or provided online. Links for relevant papers will be listed in the resourse section of the course website. PrerequisitesECE 515. ECE 534 and ECE 490 are recommended, but not required. Grading60% regular homework sets (4 sets of homework in total, 15% for each); 40% written research report (detailed guidelines for the final projects will be posted in the resource section). Homework: There are roughly 4 homework assignments. Use entry code XVW2DN to add the course on Gradescope where you will be submitting assignments. Discussion on homework problems is permitted, however each student must write and submit independent solutions. Extensions will be granted with instructor approval in advance. Otherwise late homeworks without such prior approval will not be accepted. Gradescope submission: Read about how to submit assignments on Gradescope here. Make sure your written work is legible. If the TA cannot read your work, then there is no way to grade it. Then the TA may ask you to resubmit or dock points off. Do upload high quality PDF scans; most modern smartphones have a document scanning functionality. Make sure to use the tools in Gradescope to mark where in your upload the answer to each question can be found. Start each question on a new page. For a multi-page answer, associate all relevant pages with the question. You are encouraged to typeset the homework. But this is not required. |