IN-STK9100 – Reinforcement Learning and Decision Making Under Uncertainty
Schedule, syllabus and examination date
Changes in the course due to coronavirus
Autumn 2020 and Spring 2021 the exams of most courses at the MN Faculty will be conducted as digital home exams or oral exams, using the normal grading scale. The semester page for your course will be updated with any changes in the form of examination.
Please note that there may be changes in the form of examination for some courses taught Spring 2021. We aim to bring both the course description and the semester page of all courses up to date with correct information by 1 February 2021.
This course gives a firm foundation to reinforcement learning and decision theory from mainly a statistical, but also a philosophical perspective. The aim of the course is two-fold. Firstly, to give a thorough understanding of statistical decision making, Markov decision processes, automatic experiment design, and the relation of statistical decision making to human decision making. Secondly, to relate the theory to practical problems in reinforcement learning and artificial intelligence through algorithm design, implementation and a group project in reinforcement learning.
After taking the course, you will:
- Understand the principles of decision theory.
- Understand the basics of Bayesian inference
- Understand Markov Decision Processes
- Understand Dynamic Programming
- Be able to design and implement Reinforcement Learning algorithms
- Be able to critically read research papers in reinforcement learning
- Be able to perform reinforcement learning research
Admission to the course
PhD candidates from the University of Oslo should apply for classes and register for examinations through Studentweb.
If a course has limited intake capacity, priority will be given to PhD candidates who follow an individual education plan where this particular course is included. Some national researchers’ schools may have specific rules for ranking applicants for courses with limited intake capacity.
PhD candidates who have been admitted to another higher education institution must apply for a position as a visiting student within a given deadline.
The course is limited to 30 students (IN-STK5100 and IN-STK9100 together). If the number of enrolled students is higher than the limit, they will be ranked as follows:
- PhD candidates who have the topic approved in their study plan
- Master´s students at the master´s program Data Science who has the course approved in their study plan
- Master´s students at the Faculty of Mathematics and Natural Sciences who have the course approved in their study plan
- Master´s students at the Faculty of Mathematics and Natural Sciences
Recommended previous knowledge
This is a challenging course, so it is highly recommended that you know at least
- Elementary Python programming skills (IN1000, IN1900 or equivalent experience)
- Basic linear algebra and calculus (MAT1100/1120, MAT1110 or equivalent)
- Elementary probability and statistics (STK1000, STK1100)
- A more advanced course like IN-STK5000 is advantageous
- 10 credits overlap with IN-STK5100 – Reinforcement Learning and Decision Making Under Uncertainty.
The course will consist of
- 4 hours of lectures/lab per week, for 8 weeks
- Then 2 hours of lab (project work) per week
Completion of mandatory assignments is required. Read more about requirements for submission of assignments, group work and legal cooperation under guidelines for mandatory assignments.
All mandatory assignments must be approved before you can take the final digital exam.
Candidates will be assessed based on:
Presentation of Paper (20')
All parts must be passed, and they must be passed in the same semester.
It will also be counted as one of your three attempts to sit the exam for this course, if you sit the exam for one of the following courses: IN-STK5100 – Reinforcement Learning and Decision Making Under Uncertainty
Examination support material
Any written material
Grades are awarded on a pass/fail scale. Read more about the grading system.