Courses and Schedule

The main focus of this course lies on methodologies for data-driven decision-making achieved by integrating machine learning and optimization and challenges that arise in this context. At a high level, the course covers a broad range of methods designed to address problems with single or two-stage formulations. (Note that the primary focus is not on multistage problems traditionally tackled using reinforcement learning or approximate dynamic programming). We cover applications rooted in sustainable development aligning with the 17 United Nations Sustainable Development Goals. These goals cover a wide spectrum of application domains, for example, humanitarian logistics, wildlife protection, agriculture, power systems, and transportation. The course examines the opportunities and challenges associated with solving decision-making problems aligned with these goals using integrated learning and optimization methods. We highlight gaps in the literature that are important to address such challenges.

This course is designed to prepare you to tackle hands-on machine learning projects. Throughout the semester, you will work through the key phases of an applied machine learning pipeline, including data acquisition, data preprocessing, model training, performance analysis, and the presentation of your findings. Core skills developed in this course include conducting literature reviews on domain-specific problems, applied and theoretical machine learning, Python for data science, PyTorch, version control with Git, essential Linux command-line tools, and data storytelling and presentation.

This course explores the theoretical and practical aspects of how neural networks learn and generalize. The goal of this course is not to introduce deep learning architectures or algorithms. Instead, this course will focus on understanding the training dynamics of neural networks and how they generalize. We will discuss in detail the optimization challenges involved in training overparameterized models.

The course will include 5 chapters, starting from basics (but not spending much time there) and moving toward the modern algorithms : Recall of probability basics, Basics of Bayesian statistics, Variational Inference, Monte Carlo Methods and Advanced Variational Inference.

Selected topics in Network Science, Graph Mining and Graph Learning, including patterns in real world networks, ranking and similarity measures for graphs, graph clustering and community mining techniques, and node classification and link prediction methods.

This course will cover the fundamental tools and models used for analysing and generating natural language data in a computational setting. We will learn about the core principles behind contemporary natural language processing methods and about the structure of natural language data and how it may be used for downstream tasks such as document search, classification, text generation, and translation.

This course presents an introduction to the computational modelling of natural language. Topics covered include: language modelling, syntactic parsing, lexical and compositional semantics, and discourse analysis. We will consider selected applications such as automatic summarization, machine translation, and speech processing. We will also study machine learning paradigms and algorithms that are used in natural language processing, including neural network architectures (primarily recurrent neural networks). 

Structured prediction is the problem of learning a prediction mapping between inputs and structured outputs, i.e. outputs that are made of interrelated parts often subject to constraints. Examples include predicting trees, orderings, alignments, etc., and appear in many applications from computer vision, natural language processing and computational biology among others.

This is an advanced machine learning course that will focus on the fundamental principles and related tools for structured prediction.  It will also present several advanced optimization concepts along the way, as they tend to be useful for structured prediction. The course will review the state of the art, tie older and newer approaches together, as well as identify open questions. It will consist of a mix of faculty lectures, class discussions and paper presentations by students, as well as a research project.

Prerequisite: I will assume that most of the content of IFT 6269 Probabilistic Graphical Models is known by the students.

In this course we will explore both the fundamentals and recent advances in the area of deep learning. Our focus will be on neural network-type models including convolutional neural networks and recurrent neural networks such as LSTMs.

The goal of this course is to present an overview of linear and multilinear algebra techniques for designing/analyzing ML algorithms and models, and to engage students with new research in the area.
- Fundamental notions of linear and multilinear algebra.
- Old and new ML methods leveraging matrix and tensor decomposition: PCA/CCA, collaborative filtering, spectral graph clustering, spectral methods for HMM, K-FAC, spectral normalization, tensor method of moments, NN/MRF compression, tensor regression/completion, etc.
- Introduction to Tensor Networks and their use in ML.
- Open problems.

This seminar-style course will focus on recent advances in the rapidly developing area of “foundation models”, i.e. large-scale neural network models (e.g., GPT-3, CLIP, DALL-e, etc) pretrained on very large, diverse datasets.

This graduate course is an introduction to the treatment of nonlinear differential equations, and more generally to the theory of dynamical systems. The objective is to introduce the student to the theory of dynamical systems and its applications.

This course is at the interface between game theory, optimization, and machine learning. It tries to understand how to learn models to play games. It will start with some quick notions of game theory to eventually delve into machine learning problems with game formulations such as GANs or Multi-agent RL. This course will also cover the optimization (a.k.a training) of such machine learning games.

This course aims to introduce the basic techniques of deep learning including feedforward neural networks, convolutional neural networks, and recurrent neural networks.

Computer Science (Sci) : Bandit algorithms, finite Markov decision processes, dynamic programming, Monte-Carlo Methods, temporal-difference learning, bootstrapping, planning, approximation methods, on versus off policy learning, policy gradient methods temporal abstraction and inverse reinforcement learning.

This is a seminar course on multimodal vision and language research. Some recent examples of highly successful vision-language models include GPT-4(V) and Gemini. This course will teach you the modeling techniques behind such systems, what their shortcomings are, what kind of tasks and datasets they are trained and evaluated on etc.

Basic elements of statistical learning algorithms. Examples of applications in data mining, nonlinear regression, and temporal data, and deep learning.

This course will combine lectures and seminar-style paper discussions to cover both the foundations of causality and its intersection with machine learning, a fast-moving area that connects to out-of-distribution generalization, scientific discovery and even interpretability of large models. 

This class aims to discuss these limitations and study methods to overcome them and enable agents capable of training autonomously, becoming learning and adapting systems that require little supervision.

This course is for people with no experience is NLP and would like to see how it can be used for exciting data science applications.

The course covers representation, inference and learning with graphical models; the topics at high level include directed and undirected graphical models; exact inference; approximate inference using deterministic optimization based methods, as well as stochastic sampling based methods; learning with complete and partial observations.

This course will teach students to recognize where and understand why ethical issues and policy questions can arise when applying data science to real world problems. It will focus on ways to conceptualize, measure, and mitigate bias in data-driven decision-making. This is a graduate course, in which we will cover methods for trustworthy and ethical machine learning and AI, focusing on the technical perspective of methods that allow addressing current ethical issues. 

Recent years have shown that unintended discrimination arises naturally and frequently in the use of machine learning and algorithmic decision making. We will work systematically towards a technical understanding of this problem mindful of its social and legal context. This course will bring analytic and technical precision to normative debates about the role that data science, machine learning, and artificial intelligence play in consequential decision-making in commerce, employment, finance, healthcare, education, policing, and other areas. Students will learn to think critically about how to plan, execute, and evaluate a project with these concerns in mind, and how to cope with novel challenges for which there are often no easy answers or established solutions.

We will cover a selection of topics in pragmatics (e.g., speech act theory, Gricean maxims, presuppositions and implicature), computational approaches to discourse (e.g., the Rational Speech Act framework), and their modern manifestations or lack thereof in large language models. This is a graduate seminar course in which students and attendees are expected to actively participate.