Course Syllabus

Artificial Intelligence B

Course Description

Artificial Intelligence (AI) is behind your smart phone’s intelligent personal assistant, driverless cars, robots, government fraud detection systems, and image recognition algorithms on social media, just to mention a few examples.  This course introduces you to core techniques and applications of Artificial Intelligence using primarily symbolic search-based methods in an agent-oriented paradigm.

Classes are a mix of discussions of theory/core concepts and hands-on problem solving. The majority of the course work is carried out in groups.

During the course, you will implement simple search-based agents solving navigation and transportation tasks in a virtual environment. The virtual environment is an idealized model of systems of delivery robots in hospitals, the warehouse robots at Amazon, etc. This part of the course is referred to as the programming project.

Course Overview

1. Intelligent agents and problem solving through search
2. Uninformed search
3. Informed search and estimating information through heuristics
4. Introduction to Game Theory and Adversarial Search
5. Introduction to Subsymbolic AI (*)
6. Introduction to Reinforcement Learning
7. Ethics in Artificial Intelligence
8. Lab work on the programming project

(*) Covered in much more detail in the Neural Networks and Deep Learning course

Course Elements

• Python programming
• Search problems formalization and state-space complexity estimation
• Uninformed search algorithms: Breadth-first and Depth-first search
• Informed search algorithms: Best-first search, Greedy best-first search A*
• Designing heuristic functions: admissibility, consistency and impact on search behavior and performance
• Formalising problems as games, equilibria and finding strategies through search: MINMAX, EXPECTIMAX
• Introduction to Subsymbolic AI: solving regression and classification problems using Machine Learning
• Introduction to Reinforcement learning: Monte Carlo estimations, Temporal Difference methods
• Modern discussions around ethics in Artificial Intelligence

Learning Objectives

• Being able to identify and formulate a real-life problem as a search problem
• Understand how search can be used as a paradigm to achieve intelligent behavior in artificial agents 
• Be able to implement and apply different search strategies to a complex domain and critically analyse its performance (in terms of general behavior, strengths and weaknesses)
• Understand strengths and weaknesses of search-based approaches to AI and compare them to the subsymbolic approach to intelligence
• Understand basic principles of Reinforcement Learning
• Discuss contemporary applications of AI from a technical and an ethical perspective and understand their scope, potential as well as their limitations

Faculty

Lorenzo Belgrano 

Mathematical Modelling and Computation, DTU, 2019, Machine Learning Engineer, Corti, 2019 - 2023

Readings

Stuart Russell and Peter Norvig: Artificial Intelligence - A Modern Approach. 

All readings in the Course Summary below refer to this textbook unless otherwise noted. 

Expectations of the Students      

Students are expected to carry out all the exercises following up a lecture in order to keep up with the material and to be able to actively participate in the feedback sessions. Students who have little or no experience coding in Python should either follow a Python tutorial before the course starts, or prepare to invest some hours getting up to speed with the language once we start.

You are expected to attend all DIS classes, field studies and workshops when scheduled. All of the aforementioned events are mandatory. If you miss multiple classes, the Director of Academic Support and the Director of Student Affairs will be notified. Absences will jeopardize your grade and your standing at DIS. 

Evaluation and Grading

More details can be found on the assignment description on Canvas