Optimization for Machine Learning

Course code

MLM6316.DT

old course code

Course title in Estonian

Optimeerimine masinõppes

Course title in English

Optimization for Machine Learning

ECTS credits

6.0

Assessment form

Examination

lecturer of 2023/2024 Spring semester

Not opened for teaching. Click the study programme link below to see the nominal division schedule.

lecturer of 2024/2025 Autumn semester

Not opened for teaching. Click the study programme link below to see the nominal division schedule.

Course aims

This course provides a broad introduction to optimization with a special focus on practical algorithms to solve problems from different areas of knowledge, like mathematics, statistics, computer science, aerospace, electrical engineering, and operations research. A wide variety of optimization topics are covered to that end. In each case, the underlying mathematical problem formulations and the algorithms for solving them are introduced. At the end of this course, the student will possess a solid knowledge on different types of optimization problems, the computational/numerical methods to solve them and the assumptions (both mathematical and computational) under which those techniques provide reliable solutions. The course assumes prior exposure to multivariable calculus, linear algebra, and probability concepts.

Brief description of the course

Derivatives and gradients, bracketing, local descent, first-order methods, second-order methods, direct methods, stochastic methods, population methods, constrained optimization, linear constrained optimization, multi-objective optimization.

Learning outcomes in the course

Upon completing the course the student:

- has a solid understanding of different methodologies to solve a wide number of optimization problems in mathematics, statistics, computer science, aerospace, electrical engineering, and operations research;

- can implement computationally optimization algorithms using free software in order to solve efficiently classical and multi-objective optimization problems;

- is able to determine the best solution technique for different optimization problems depending on the particular features of the particular problem and the constraints on the variables involved;

- interprets the solutions obtained by a computer program in order to propose strategies to optimize processes in different areas of the sciences, engineering, industry, and technology.

- has a solid understanding of different methodologies to solve a wide number of optimization problems in mathematics, statistics, computer science, aerospace, electrical engineering, and operations research;

- can implement computationally optimization algorithms using free software in order to solve efficiently classical and multi-objective optimization problems;

- is able to determine the best solution technique for different optimization problems depending on the particular features of the particular problem and the constraints on the variables involved;

- interprets the solutions obtained by a computer program in order to propose strategies to optimize processes in different areas of the sciences, engineering, industry, and technology.

Teacher

Jorge Eduardo Macias-Diaz

Prerequisite course 1