CS201INTERMEDIATECOMPUTING

Introduction to Reinforcement Learning

From animal learning to mathematical models and Python agents.

Course Information

Course Information
ModeInstructor-led or blended · proposed
TermFA1 · Aug–Sep · proposed
Duration8 weeks · 90-minute lecture and 90-minute lab weekly; about 24 further hours of study
Credits1.5 credits
PrerequisitesBasic Python (CS101 or equivalent), school algebra, graphs and elementary probability. MA101 helpful; no prior RL or animal research required.
Who it is forNeuroscience, psychology, medicine, engineering and computer-science learners new to reinforcement learning.
FacultyTo be announced
AssessmentQuizzes 20% · coding assignments 40% · mini project and explanation 40%. 60% to pass.

About the course

Animal learning motivates each mathematical idea, followed by a Python experiment. Learners implement tabular algorithms in NumPy and test them over repeated runs.

Learning outcomes

  • Define states, actions, rewards, policies, returns and value; distinguish reward from expected future reward
  • Explain prediction-error learning and derive simple Bellman and temporal-difference updates
  • Implement bandits, value iteration, TD learning, SARSA and Q-learning, and test their parameters
  • Relate computational predictions to conditioning, instrumental learning and selected dopamine findings without equating every biological signal with a single RL quantity

Modules and Python laboratories

8 modules

  • Thorndike, Skinner, reinforcement and punishment

  • the agent–environment loop

    Lab simulate rewarded action acquisition.

Labs and projects: Weekly Python labs; a mini project on blocking, reversal learning or maze choice.

Books and readings: Sutton & Barto, Reinforcement Learning (2018); Schultz, Dayan & Montague (1997); course notes