263-5300-00L Guarantees for Machine Learning
| Semester | Spring Semester 2020 |
| Lecturers | F. Yang |
| Periodicity | yearly recurring course |
| Language of instruction | English |
Courses
| Number | Title | Hours | Lecturers | ||||
|---|---|---|---|---|---|---|---|
| 263-5300-00 V | Guarantees for Machine Learning Special selection process. Preference is given to Masters and Doctorate students. If need be other criteria are degree program and previous courses taken. | 2 hrs |
| F. Yang | |||
| 263-5300-00 A | Guarantees for Machine Learning | 2 hrs | F. Yang |
Catalogue data
| Abstract | This course teaches classical and recent methods in statistics and optimization commonly used to prove theoretical guarantees for machine learning algorithms. The knowledge is then applied in project work that focuses on understanding phenomena in modern machine learning. |
| Objective | This course is aimed at advanced master and doctorate students who want to understand and/or conduct independent research on theory for modern machine learning. For this purpose, students will learn common mathematical techniques from statistical learning theory. In independent project work, they then apply their knowledge and go through the process of critically questioning recently published work, finding relevant research questions and learning how to effectively present research ideas to a professional audience. |
| Content | This course teaches some classical and recent methods in statistical learning theory aimed at proving theoretical guarantees for machine learning algorithms, including topics in - concentration bounds, uniform convergence - high-dimensional statistics (e.g. Lasso) - prediction error bounds for non-parametric statistics (e.g. in kernel spaces) - minimax lower bounds - regularization via optimization The project work focuses on active theoretical ML research that aims to understand modern phenomena in machine learning, including but not limited to - how overparameterization could help generalization ( interpolating models, linearized NN ) - how overparameterization could help optimization ( non-convex optimization, loss landscape ) - complexity measures and approximation theoretic properties of randomly initialized and trained NN - generalization of robust learning ( adversarial robustness, standard and robust error tradeoff ) - prediction with calibrated confidence ( conformal prediction, calibration ) |
| Prerequisites / Notice | It’s absolutely necessary for students to have a strong mathematical background (basic real analysis, probability theory, linear algebra) and good knowledge of core concepts in machine learning taught in courses such as “Introduction to Machine Learning”, “Regression”/ “Statistical Modelling”. It's also helpful to have heard an optimization course or approximation theoretic course. In addition to these prerequisites, this class requires a certain degree of mathematical maturity—including abstract thinking and the ability to understand and write proofs. |
Performance assessment
| Performance assessment information (valid until the course unit is held again) | |
Performance assessment as a semester course | |
| ECTS credits | 5 credits |
| Examiners | F. Yang |
| Type | graded semester performance |
| Language of examination | English |
| Repetition | Repetition only possible after re-enrolling for the course unit. |
| Additional information on mode of examination | Last cancellation/deregistration date for this graded semester performance: second Friday in March! Please note that after that date no deregistration will be accepted and a "no show" will appear on your transcript. one Midterm exam (50%) Homework (10%) Course project (40%) |
Learning materials
| Main link | Information |
| Only public learning materials are listed. | |
Groups
| No information on groups available. |
Restrictions
| Places | Limited number of places. Special selection procedure. |
| Waiting list | until 08.03.2020 |
| End of registration period | Registration only possible until 01.03.2020 |
