The text of this course will be revised later.
Upon successful completion of the course, you will:
- Understand the foundations and core ideas of reinforcement learning algorithms and apply them to small-scale problems
- Formalize a learning problem as a Markov Decision Process (MDP) and understand the use and relevance of value functions in decision-making
- Find exact solutions (optimal value functions and optimal policies) to simple RL problems and apply these ideas to real-world scenarios
- Understand the core classes of methods to solve MDPs, their similarities and differences, strengths and weaknesses and understand their application to different decision problems
- Be prepared to take on more advanced topics in the domains of AI and deep learning.
|
|
Cognitive agents learn by interaction with the environments they are embedded in. Exercising the connection between sensors and actuators (sensorimotor loop) allows the agent to modify its inputs, explore the consequences of its actions and decisions in order to understand how to optimize its behavior to achieve a desired goal. Such interactions are foundational to mammalian cognition and behavior and underlie many theories of learning, decision making and natural intelligence.
Beyond its relevance in the domains of psychology and neuroscience, as a core learning mechanism in biological intelligence, Reinforcement Learning (RL) has broad practical applicability in domains ranging from machine learning, operations research or control engineering.
This course will introduce the core ideas, methods and techniques used in RL algorithms.
|
|
|
|
Knowledge of the following fields is strongly advised: Linear Algebra, Probability Theory and Programming.
|
|
Exam: 60% (one retake possible)
Individual Assignment: 40%
Average grade: >5.5
Grading for the different parts will be made public in Brightspace. Only the final grade will be published in Osiris at the end of the course
|
|
Please sign up for any course at (https://portal.ru.nl/home), it is obligatory.
Students who are enrolled for a course are also provisionally registered for the exam.
Resit: Manual register at (https://portal.ru.nl/home) until five working days prior to the date of the exam. No delayed registration is possible.
We urge you to always read the course information on Brightspace.
|
|