Postgraduate / Master's · Artificial Intelligence
Reinforcement Learning
What is it
Reinforcement learning studies how an agent learns to make sequences of decisions by interacting with an environment and receiving rewards or penalties — covering Markov decision processes, value functions, policy gradient methods, and the exploration-exploitation trade-off.
Why it matters
Reinforcement learning underlies some of the most striking AI achievements — from game-playing systems that surpass human experts to the reward-based fine-tuning behind modern language models — and it requires a genuinely different mathematical framework from standard supervised learning.
Exam tip
When analysing a reinforcement learning setup, always identify the reward function precisely before evaluating the agent's behaviour — a huge proportion of unexpected or undesirable agent behaviour traces back to a poorly specified reward function rather than a flaw in the learning algorithm itself.
Related topics
Want help mastering Reinforcement Learning?
Tell us about the student's goals and confidence — we'll design a personalised plan.