Machine Learning and Optimization

What it is about

You will learn advanced concepts and methods of machine learning as well as common optimization procedures for their training. In the end you will be able to apply and further develop modern learning methods (e.g. deep neural networks), design optimization algorithms and analyze their behavior theoretically and empirically.

What you will be able to do

• Theoretical foundations of advanced ML algorithms
• Apply and modify advanced ML methods (e.g. Deep Neural Networks)
• Develop and tune optimization algorithms for ML models
• Rigorous performance analysis: generalization bounds and convergence/complexity
• Practical implementation and empirical evaluation of ML methods
• Familiarity with topics beyond classical supervision (Active Learning, Fairness)

What the module consists of

• LectureIntroduction and discussion of the algorithms and related theory, including current research findings
• Exercises/Discussion SessionInteractive forum for questions on problems and deepening individual topics on the board
• Homework (homeworks)Regular theoretical and programming tasks for practice (fulfillment of prerequisite for taking the exam)

Teaching method

• LectureDelivery of the theory behind ML methods and presentation of current results
• Problems with solution exchangeRepetitive practice: tasks every two weeks; solutions for self-check distributed
• Open tasks/project questionsEncourages adaptation of existing methods and approach to practical/research questions
• Interactive discussion sessionTargeted clarification of exercise questions and detailed explanations on demand