Courses and Schedule

Statistical learning is a discipline that involves extracting information from observed data in order to solve a specific problem. Examples include detecting patterns in signals (images, time series), identifying associations between different variables, or highlighting group structures in a dataset.

Large Language Models (LLMs) have demonstrated remarkable capabilities in natural language generation, but controlling their outputs to meet specific constraints and requirements remains a fundamental challenge. This seminar explores the cutting-edge research and methodologies for controlled generation in LLMs, focusing on computational methods and frameworks for constraining and guiding language model outputs. The course examines approaches to enforce logical constraints, type safety, and semantic coherence while maintaining generation quality. Students will engage with state-of-the-art research through presentations and discussions, covering probabilistic methods, architectural innovations, and practical applications of controlled generation techniques.

Self-driving vehicles are poised to become one of the most pervasive and impactful applications of autonomy, and have received a great deal of attention recently.

This course considers problems in perception, navigation, planning and control, and their systems-level integration in the context of self-driving vehicles through an open-source curriculum for autonomy education that emphasizes hands-on experience. Integral to the course, students will collaborate to implement concepts covered in lecture on a low-cost autonomous vehicle with the goal of navigating a model town complete with roads, signage, traffic lights, obstacles, and citizens.

This seminar will explore how machine learning can be applied in fighting climate change. We will look at ways that machine learning can be used to help mitigate greenhouse gas emissions and adapt to the effects of climate change – via applications in electricity systems, buildings, transportation, agriculture, disaster response, and many other areas. Particular emphasis will be given to understanding exactly when machine learning is relevant and helpful, and how to go about scoping, developing, and deploying a project so that it has the intended impact.

In this course, we will study machine learning models including models for decision making. Massive datasets are now common and require scalable analysis tools. Machine learning provides such tools and is widely used for modelling problems across many fields including artificial intelligence, bioinformatics, finance, marketing, education, transportation, and health. In this context, we study how standard machine learning models for supervised (classification, regression) and unsupervised learning (for example, clustering and topic modelling) can be scaled to massive datasets using modern computation techniques (for example, computer clusters). In addition, we will discuss recent models for recommender systems as well as for decision making (including multi-arm bandits and reinforcement learning).

This course covers a selected set of topics in machine learning and data mining, with an emphasis on understanding the inner workings of the common algorithms. The majority of sections are related to commonly used supervised learning techniques, and to a lesser degree unsupervised methods. This includes fundamentals of algorithms on linear and logistic regression, decision trees, support vector machines, clustering, neural networks, as well as key techniques for feature selection and dimensionality reduction, error estimation and empirical validation.

This is a seminar-style course, where the class as a whole will work together in running the course. In the first few lectures, I will provide an overview of LLMs and highlight the challenges. By the end of the course, you should be able to meaningfully contribute to cutting-edge research in natural language understanding.

Generative flow networks, also known as GFlowNets or simply GFN, are a class of generative machine learning models that perform amortised probabilistic inference. This course will cover the fundamental aspects of GFlowNets, starting from a motivation and introduction to the method, and progressing towards more advanced concepts and applications, as well as the connection to other generative models and probabilistic inference methods. The course will combine theory with project work, with a special emphasis on the application of GFlowNets for scientific discovery.

Prediction (The Statistical Trap, Probability, Generalization, Repeatability). The Holdout Method (Stability, Internal Validity). Machine Learning Benchmarks (History of Benchmark Design, Evaluation Failures, Dynamic Benchmarks, Multi-Task Benchmarking, Sensitivity). External Validity, Reproducibility, Robustness (Distribution Shift). Adaptivity and Overfitting (Test Set Reuse). Construct Validity (Context, Scope, Lakatosian Defense). Evaluating Uncertainty (Calibration Error, Prediction Sets, Conformal Inference). Forecast Evaluation (Scoring Rules, Accounting for Uncertainty, Confidence Sequences). Games as Benchmarks (History of Benchmark Games). Evaluating Interventions (Temporal Validity, Compliance, Adjustment). Regulation and Deployment (Program Evaluation, Policy Evaluation, Auditing via Sequential Hypothesis Testing).

This course covers data science (DS), a science that combines statistics, data processing, machine learning, scientific inquiry, visualization, business analytics, big data, and big models. The purpose of data science is to address or gain insight into a problem in the real world by the application of computational and statistical techniques.

The course will cover the following subjects: Advanced Data Visualization, Feature engineering, Hypothesis Testing, Feature Selection, outlier detection, Unsupervised Learning, Algorithmic Bias, Virtualization and Distributed learning, Deep Learning (MPLs, CNNs), Computer Vision, Natural language processing, tokenization and transformers, Large Language Model Fine-tuning (SFT), Reinforcement learning (RL, RLHF, RLVR), Vision Language Models (VLMs).

Basic elements of statistical and symbolic learning algorithms. Examples of applications in data mining, pattern recognition, nonlinear regression, and time data.

This is an introductory course on reinforcement learning (RL) and sequential decision-making under uncertainty with an emphasis on understanding the theoretical foundation. We study how dynamic programming methods such as value and policy iteration can be used to solve sequential decision-making problems with known models, and how those approaches can be extended in order to solve reinforcement learning problems, where the model is unknown. Other topics include, but not limited to, function approximation in RL, policy gradient methods, model-based RL, and balancing the exploration-exploitation trade-off. The course will be delivered as a mix of lectures and reading of classical and recent papers assigned to students. As the emphasis is on understanding the foundation, you should expect to go through mathematical detail and proofs. Required background for this course includes being comfortable with probability theory and statistics, calculus, linear algebra, optimization, and (supervised) machine learning.

Computer Science (Sci) : An introduction to the computational modelling of natural language, including algorithms, formalisms, and applications. Computational morphology, language modelling, syntactic parsing, lexical and compositional semantics, and discourse analysis. Selected applications such as automatic summarization, machine translation, and speech processing. Machine learning techniques for natural language processing.

Modern data analysis methods are expected to handle massive amounts of high dimensional data that are being collected in a variety of domains. The high dimensionality of such data introduces numerous challenges, typically referred to as the "curse of dimensionality", which render traditional statistical learning approaches impractical or ineffective for their analysis. To cope with these challenges, significant effort has been focused on developing geometric data analysis approaches that model and capture the intrinsic geometry of processed data, rather than directly modeling their distribution. In this course we will explore such approaches and provide an analytical study of the models and algorithms they use. We will start by considering supervised learning and distinguish classifiers that are based on geometric principles from posterior and likelihood estimation approaches. Next, we will consider the unsupervised learning task of clustering data and contrast approaches based on density estimation from ones that rely on metric spaces or graph constructions. Finally, we will consider more fundamental tasks in intrinsic representation learning, with particular focus on dimensionality reduction and manifold learning, e.g., with diffusion maps, tSNE, and PHATE. Time permitting, we will include guest talks on research areas related to the course, and discuss recent developments in graph signal processing and geometric deep learning.

This course provides a rigorous introduction to the field of machine learning (ML). The aim of the course is not just to teach how to use ML algorithms but also to explain why, how, and when these algorithms work. The course introduces fundamental algorithms in supervised learning and unsupervised learning from the first principles. The course, while covering several problems in machine learning like regression, classification, representation learning, dimensionality reduction, will introduce the core theory, which unifies all the algorithms.

The goal of this course is to introduce the concepts (theory and practice) needed to approach and solve data science problems. The first part of the course will cover the principles of analyzing data, the basics about different kinds of models and statistical inference. The second part expands into the statistical methods and practical techniques to deal with common modalities of data - image, text and graphs. Specific programming frameworks required for data science will be covered in the lab sessions.

Advanced course in reinforcement learning. Topics: Policy gradient methods, gradient estimation, analysis of valued-based function approximation methods, optimal control and automatic differentiation, bilevel optimization in meta-learning and inverse reinforcement learning.

This is a course on representation learning in general and deep learning in particular. Deep learning has recently been responsible for a large number of impressive empirical gains across a wide array of applications including computer vision, natural language processing and speech recognition.

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.