Reinforcement Learning (5cr)

Evaluation scale

H-5

Content scheduling

- Basics of reinforcement learning concepts (including exploration and exploitation)
- Common reinforcement learning methods and processes
- Policies: evaluation, improvement, iteration
- Conventional Reinforcement learning
- Deep Reinforcement Learning

Objective

You understand the core principles behind reinforcement learning
You understand the differences of reinforcement learning regarding classic machine learning and conventional deep learning
You can use conventional reinforcement learning solutions to create an AI that functions in a limited moving space
You can use deep learning methods in order to create situational reinforcement learning solutions
You can share your results and exercises via a version control system.

Content

Basics of reinforcement learning concepts (including exploration and exploitation)
Markov Decision Processes, Monte Carlo Methods, Bellman Equation
Policies: evaluation, improvement, iteration
Conventional Reinforcement learning
Deep Reinforcement Learning

Location and time

The course will be studied on-campus. Lapland University of Applied Sciences, Rantavitikka Campus, 12.1.2026 - 15.5.2026.

Materials

The study material of the course will be in the course's Moodle workspace. Additional material can be searched from the internet as needed. It's also recommended to independently search for more information during the course.

Teaching methods

Lectures, workshops, examples, exercises and self-supervised work.

The Arene's AI recommendations and traffic light model for universities of applied sciences will be applied during the course. Whether AI is allowed during an exercise, will depend on the exercise itself.

Employer connections

The study unit has been connected to the semester project held during the same semester. The technologies and related examples during the course will be partially based on the needs of the involved company in the semester project.

Exam schedules

The course will be graded based on personal work and exercises.

International connections

The study unit will be in English.

Completion alternatives

Alternative ways to study the course can be inquired from the instructor.

Student workload

The study unit is 5 ECTS, which amounts to 135 hours of work in total (approximately). ON average, your work will be distributed the following way:

- Lectures and workshops: 50 h
- Self-study and preparation: 45 h
- Graded tasks: 40h

Assessment criteria, satisfactory (1)

You can create a simple reinforcement application
You are aware of the basic principles behind reinforcement learning
You understand the difference of reinforcement learning when compared to other conventional machine learning technologies
You can share your results and exercises via a version control system.

Assessment criteria, good (3)

You can create various reinforcement applications, using both conventional methods and deep learning methods
You understand the basic principles behind reinforcement learning on the general level
You understand the difference of reinforcement learning when compared to other conventional machine learning technologies
You can share your results and exercises via a version control system.

Assessment criteria, excellent (5)

You can create various reinforcement applications, using both conventional methods and deep learning methods
You understand the basic principles behind reinforcement learning on the general level
You understand the difference of reinforcement learning when compared to other conventional machine learning technologies
You can optimize your reinforcement learning applications to improve performance
You can share your results and exercises via a version control system.

Qualifications

Basics of programming
Basics of Python data analytics modules/libraries
Basics of conventional machine learning methods
Basics of Deep Learning