Michalis K. Titsias

Papers

Total Citations

H-Index

About

Michalis K. Titsias is a leading figure in machine learning, renowned for his foundational contributions to probabilistic modeling and variational inference. His research centers on developing scalable, principled methods for complex data, particularly through Gaussian processes and Bayesian nonparametrics. Titsias is best known for pioneering the sparse variational Gaussian process framework, which revolutionized the field by enabling efficient inference on large datasets—a breakthrough that underpins his highly cited work on variational Gaussian process dynamical systems (2011, 57 citations). This paper introduced a practical, nonlinear probabilistic approach for modeling high-dimensional time series from robotics, biology, and vision, offering a robust alternative to traditional methods. His impact is further cemented by innovations in variational inference for latent variable models, including the doubly stochastic variational inference technique, which has become a cornerstone of modern Bayesian deep learning. With thousands of citations across his body of work, Titsias’s algorithms are widely adopted in both academia and industry, making complex probabilistic models accessible for real-world applications. His achievements include shaping the theoretical foundations of scalable inference, earning him recognition as a key architect of modern variational methods.