Spring 2026.
As stated on the department website:
To provide a comprehensive introduction to current research methods in artificial intelligence. The course is appropriate both for nonspecialists who wish to acquire a strong grounding in the engineering aspects of computing with real-world data, and as a prerequisite to more advanced courses in artificial intelligence.
Modeling structure and uncertainty in real-world data and reasoning with these models, including perception, categorization and learning. Applications to text and web processing, computer vision, speech recognition and language processing, and user interfaces. Planning for risk and reward. Applications to expert systems, medical decision-making, robotics and design.
This course is scheduled between an introductory course in AI, such as CS 440 or CS520, and more advanced courses, such as CS 533 (natural language processing), 535 (pattern recognition), or 536 (machine learning). As such, we will leave in-depth coverage of specific techniques to those courses. This course will focus less on the implementation of specifical techniques in artificial intelligence and more on the application of those techniques. Special attention will be payed to data, noise, and the difficulties of real-world applications.
The recommended book for this course is “Artificial Intelligence: A Modern Approach,” fourth edition, by Russel and Norvig. It is likely not the best book on any particular topic as it lacks in either depth or approachability, (compared to Sutton and Barto’s “Reinforcement Learning”, for example) but it does have a breadth not found in most other texts.
I also recommend students read “The Mind of a Bee,” by Lars Chittka, purely for their own personal growth. We would be pleased with ourselves if we could build agents with a fraction of the capabilities of these tiny creatures at even a small fraction of their efficiency. The research presented in the book makes it clear that robust, real-world systems are a mixture of simple and complicated solutions, and reminds us that we do not truly understand biological learning, knowledge, or instinctive behavior. Observations on how insects solve problems is also a lesson and a warning that our human tendency to assume similarity in other agents can lead us to overestimating the intelligence involved in a task. The book also corrects some mistatements in Russel and Norvig’s book, where they unfairly malign the sphex wasp (or at least unfairly lump all sphex into the same category).
Course material will be posted on the course website.
The course will roughly follow later chapters in the text, and will supplement with examples from published research literature.
Topic 1: Introduction
Chapters 1-2, the Chinese Room example, pg 1036, section 28.2.1.
What are agents? How do we describe an environment? What is “knowledge” and how can we trust AI when using it in never before seen scenarios?
Discrete vs continuous environments. Deterministic vs stochastic environments. Uncertainty and its sources: data fidelity, label limitations, sampling shortfalls, and hidden states. Testing and evaluation in different environments.
Topic 2: Planning in Uncertain Environments
Chapters 11, 12, 13, 14.
Planning under different sources of uncertainty. Probabilistic reasoning. Probabalistic reasoning over time. Markov models. Filters.
Topic 3: Evaluating Decisions
Ch 15, 16.
Decisions must be evaluated over long time periods. When driving, when do we decided it is a good time to change lanes? Just before the exit? A mile before? Or should we always remain in the exit lane?
Topics: Utility theory. Decision networks. Action utility functions. Bellman equation. Q-function. Value iteration, policy iteration, linear programming, and Monte Carlo planning. Online algorithms, such as reinforcement learning.
Topic 4: Learning Behavior
Ch 19-23
A (shallow) survey of different learning approaches and occasions when they are well-suited to a task. Decision trees, ensembles, clustering, the EM algorithm, deep learning, and reinforcement learning. Selected case-studies of these approaches from published research literature.
Evaluation is slightly skewed towards homework and projects, as this class is meant to evaluate mastery of reesearch methods. I am open to slight changes in evaluation, if the class as a whole feels that a different learning experience would be of greater benefit to their education.