Martin Jankowiak
Martin Jankowiak is a senior research scientist at Uber whose research focuses on probabilistic machine learning. He is a co-creator of the Pyro probabilistic programming language.
Engineering Blog Articles
Announcing a New Framework for Designing Optimal Experiments with Pyro
Uber AI released a new framework on top of Pyro that lets experimenters seamlessly automate optimal experimental design (OED) for quicker model iteration.
Research Papers
Pyro: Deep Universal Probabilistic Programming
E. Bingham, J. Chen, M. Jankowiak, F. Obermeyer, N. Pradhan, T. Karaletsos, R. Singh, P. Szerlip, P. Horsfall, N. Goodman
Pyro is a probabilistic programming language built on Python as a platform for developing advanced probabilistic models in AI research. [...] [PDF]
Journal of Machine Learning Research (JMLR), 2018
Pyro is a probabilistic programming language built on Python as a platform for developing advanced probabilistic models in AI research. [...] [PDF]
Journal of Machine Learning Research (JMLR), 2018
Pathwise Derivatives for Multivariate Distributions
M. Jankowiak, T. Karaletsos
We exploit the link between the transport equation and derivatives of expectations to construct efficient pathwise gradient estimators for multivariate distributions. We focus on two main threads. [...] [PDF]
International Conference on Artificial Intelligence and Statistics (AI STATS) (in submission), 2019
We exploit the link between the transport equation and derivatives of expectations to construct efficient pathwise gradient estimators for multivariate distributions. We focus on two main threads. [...] [PDF]
International Conference on Artificial Intelligence and Statistics (AI STATS) (in submission), 2019
Pathwise Derivatives Beyond the Reparameterization Trick
M. Jankowiak, F. Obermeyer
We observe that gradients computed via the reparameterization trick are in direct correspondence with solutions of the transport equation in the formalism of optimal transport. We use this perspective to compute (approximate) pathwise gradients for probability distributions not directly amenable to the reparameterization trick: Gamma, Beta, and Dirichlet. [...] [PDF]
International Conference on Machine Learning (ICML), 2018
We observe that gradients computed via the reparameterization trick are in direct correspondence with solutions of the transport equation in the formalism of optimal transport. We use this perspective to compute (approximate) pathwise gradients for probability distributions not directly amenable to the reparameterization trick: Gamma, Beta, and Dirichlet. [...] [PDF]
International Conference on Machine Learning (ICML), 2018
