Machine Learning and Data Science Seminar

Associate Professor Nicolas Boumal will talk about: 'Smooth, globally Polyak-Łojasiewicz functions are nonlinear least-squares'

Polyak-Łojasiewicz (PŁ) functions abound in the literature, especially in nonconvex optimization. When they are also smooth, they become surprisingly simple---with an exotic twist. The plan is for us to discover the structure of those functions and of their sets of minimizers via gradient flow and fiber bundles.

Joint work with Christopher Criscitiello and Quentin Rebjock.

The dynamic nature of first order methods can be interpreted by means of continuous time models. In this survey talk, we explain how physical concepts like acceleration, inertia or momentum have been used to improve the performance of convex optimization algorithms. 

We give special attention to the historical evolution of complexity results, especially in the form of convergence rates, under the light of this connection. We also discuss different ways in which acceleration schemes can be applied when the smoothness or strong convexity parameters are unknown, and how these ideas extend to saddle point and constrained problems. 

Extragradient methods are a fundamental class of algorithms for solving min-max optimization problems and variational inequalities. While the classical theory is largely developed under smoothness and other relatively restrictive assumptions, many problems arising in modern machine learning call for analysis in weaker regularity regimes and in stochastic large-scale settings. In this talk, we present new convergence results for deterministic and stochastic extragradient methods beyond the classical framework. In particular, we establish convergence guarantees under the (L0, L1)-Lipschitz condition and derive new step-size rules that expand the range of provably convergent regimes. We also introduce Polyak-type step sizes for deterministic and stochastic extragradient methods, leading to adaptive variants with favourable theoretical properties and practical performance. Our results focus primarily on monotone problems, with extensions to selected structured non-monotone settings. We conclude with numerical experiments that illustrate the theory and the empirical behaviour of the proposed methods.

TBA

Professor Pereyra will talk about; 'Physics-informed deep generative models: Applications to computational sensing'

This talk introduces a novel mathematical and computational framework for constructing high-dimensional Bayesian inversion methods that leverage state-of-the-art generative denoising diffusion models as highly informative priors. A central innovation is the construction of physics-informed generative models using Langevin diffusion processes and Markov chain Monte Carlo (MCMC) sampling techniques to develop stochastic neural network architectures capable of near-exact sampling. The obtained networks are modular and composed of interpretable layers that are directly related to statistical image priors and data likelihoods derived from forward observation models. The layers encoding the data likelihood function are designed for flexibility, enabling scene and instrument model parameters to be specified at inference time and seamlessly integrated with pre-trained foundational generative priors. To achieve high computational efficiency, we employ adversarial model distillation, which yields excellent sampling performance with as few as four Markov chain Monte Carlo steps, even in problems exceeding one million dimensions. Our approach is validated through non-asymptotic convergence analysis and extensive numerical experiments in computational image and video restoration. We conclude by discussing unsupervised training strategies that allow the models to be fine-tuned directly from measurement data, thereby bypassing the need for clean reference data.

The talk is based on recent work in physics-informed generative AI for Bayesian imaging: https://arxiv.org/abs/2503.12615 (ICCV 2025), which uses a distilled latent Stable Diffusion XL model trained on five billion clean images as a zero-shot prior, and  https://arxiv.org/pdf/2507.02686, which integrates pixel-based diffusion models with deep unfolding and diffusion distillation (TMLR 2025). The extension to video restoration is presented in https://arxiv.org/abs/2510.01339 (ICLR 2025). Our approach to unsupervised training of diffusion models is introduced in https://arxiv.org/abs/2510.11964.