This module focuses on problem solving through classical search algorithms in observable, deterministic, known environments where the solution is a sequence of actions. The module begins with formally defining the problem formulation for the environment and defining the search techniques. Students will be able to understand, apply and evaluate solutions based on classical search techniques like Depth First Search, Breadth First Search and Uniform Cost Search.

This module focuses on exploring a gaming scenario through adversarial search techniques. The unpredictability of the other agents can introduce contingencies into the agent’s problem-solving process. Students will understand and explore the working of a competitive environment through Adversarial search techniques. They will be able to design solutions for such adversarial problems based on the min-max algorithm. Finally, they will also learn to optimise the memory efficiency of solutions through alpha-beta pruning.

This module focuses on exploring local and bio-inspired search techniques to deal with partially observable and unknown environments. Students will know when to use local search techniques and deal with sub-optimal solutions. They will also learn about evolutionary algorithms and apply them to solve real-world problems. Finally, they will be able to understand the working of swarm intelligence algorithms and their Stigmergy principles.

This module explores the fundamental concepts of logical agents and knowledge representation using propositional logic, cornerstones of artificial intelligence (AI). Students can develop logic as a general class of representations to support knowledge-based agents. Such agents can combine and recombine information to suit myriad purposes. Students will learn about logical agents and formal reasoning to make decisions and infer new knowledge. They will also be able to create a simple knowledge base on PL and make inferences from it.

This module introduces the expressive power of First Order Logic (FOL) for representing relationships and reasoning about the world. You will learn its syntax and semantics, and how to perform inference using Propositionalisation and Forward Chaining. The module concludes with a real-world application where FOL is used to support knowledge-based reasoning systems.

In this module, you will learn how to represent and reason with uncertain knowledge. You will be introduced to probability theory and how it is used in AI to model uncertainty. You will understand joint probability distributions, conditional independence, and the structure and semantics of Bayesian Networks.

This module introduces the classical reinforcement learning problem of the multiarmed bandit. You will learn how agents can make sequential decisions in an uncertain environment, balancing the trade-off between exploration and exploitation. The module also introduces the idea of regret minimisation, which quantifies the performance of a learning strategy over time. One real-world application of the multiarmed bandit will also be discussed.

This module introduces the formal framework of Markov Decision Processes (MDPs), which underlies much of reinforcement learning. Students will learn how to model sequential decision-making problems using MDPs by defining states, actions, transition probabilities, and rewards. The concepts of return and value functions will be explored in detail, along with the Bellman equations. The module concludes with an application of MDPs to robot navigation.

This module introduces Temporal-Difference (TD) Learning, a fundamental idea in reinforcement learning where agents update their value estimates based on partial experience. Students will learn how value functions can be incrementally improved without requiring the full outcome of an episode. The module includes the key TD algorithms: TD(0), SARSA, and Q-learning. A final video shows how TD learning can be used to train an agent to play Tic-Tac-Toe or Snake, demonstrating its real-world applicability.