Francisco is a PhD student in the Computer Laboratory. He is Interested in the duality between optimisation and sampling with a focus on applications. In particular,  exploring stochastic control based methodologies (e.g. Schrödinger Bridges) in practical contexts such as Bayesian machine learning as well as generative modelling, for example developing better samplers for Bayesian Deep Learning.  Overall, he aims to focus on dynamical formulations of different learning tasks to explore physically motivated efficient algorithms, always keeping the practical/application component as the main focus.

Related Publications

Bayesian Learning via Neural Schrödinger-Föllmer Flows

Francisco Vargas, Andrius Ovsianas, David Fernandes, Mark Girolami, Neil D. Lawrence, Nikolas Nüsken

Fourth Symposium on Advances in Approximate Bayesian Inference, :

Adversarial Concept Erasure in Kernel Space

Shauli Ravfogel, Francisco Vargas, Yoav Goldberg, Ryan Cotterell


Efficient Representations for Privacy-Preserving Inference

Han Xuanyuan, Francisco Vargas, Stephen Cummins