Foundations of Optimal Transport with JAX and OTT

Optimal transport offers a powerful mathematical framework for comparing probability distributions, but scaling these methods to large datasets has historically been a challenge. Modern libraries now allow you to leverage hardware acceleration to solve complex transport problems efficiently. This text-based course guides you from the fundamental mathematical theory of optimal transport to writing clean, accelerated code. You will understand how to formulate transport problems, compute distances, and implement scalable solutions using JAX and the Optimal Transport Tools (OTT) library. What you'll learn: Understand the core mathematical theory of Monge and Kantorovich formulations of optimal transport; Compute Wasserstein distances and define custom cost functions for diverse data types; Implement entropic regularization and the Sinkhorn algorithm for fast, differentiable transport; Leverage JAX features like automatic differentiation and vectorization to scale your transport solvers; Apply the OTT library to solve practical problems in domain adaptation and generative modeling. You will begin with essential mathematical terminology and foundational concepts before moving on to hands-on code implementations. Through clear written explanations and structured code snippets, you will progress from basic distance calculations to advanced regularized transport algorithms. This course is designed for developers, data scientists, and researchers new to optimal transport, with no prior experience in JAX or OTT required. Start mastering optimal transport and accelerate your machine learning workflows today.