Artur Bekasov

Photo of Artur

I am an Applied Scientist at Amazon, where I develop novel methods in machine learning and statistics, and apply them to practical problems. My research interests lie in probabilistic machine learning, and especially in the intersection of deep learning and Bayesian inference. I have previously worked on normalizing flows for generative modelling/density estimation, and Bayesian inference for robust predictive models.

I received a PhD in Machine Learning from the School of Informatics, University of Edinburgh, where I was supervised by Iain Murray. During my studies I completed an applied machine learning internship with Amazon (Berlin), and a machine learning research internship with Google Research (Zurich), where I worked on generative modelling of human motion with normalizing flows.

For my MSc thesis I worked on generative video modelling via latent space transitions. Before diving deeper into machine learning, I worked on recommendations at Amazon (Edinburgh) and studied Computer Science at the University of Manchester, where my final-year project was on evolutionary computation.

Research

Learning Action Embeddings for Off-Policy Evaluation
Matej Cief, Jacek Golebiowski, Philipp Schmidt, Ziawasch Abedjan, Artur Bekasov
In submission, 2023
[arXiv]

Variational Boosted Soft Trees
Tristan Cinquin, Tammo Rukat, Philipp Schmidt, Martin Wistuba, Artur Bekasov
AISTATS, 2023
[arXiv]

Accurate and Reliable Probabilistic Modeling with High-dimensional Data
Artur Bekasov
PhD thesis
University of Edinburgh, 2022
[Edinburgh Research Archive]

Ordering Dimensions with Nested Dropout Normalizing Flows
Artur Bekasov, Iain Murray
Workshop on Invertible Neural Networks, Normalizing Flows, and Explicit Likelihood Models, ICML, 2020
Selected for a spotlight
[arXiv] [Code] [Virtual poster talk]

Neural Spline Flows
Conor Durkan*, Artur Bekasov*, Iain Murray, George Papamakarios
NeurIPS, 2019
* Equal contribution
[NeurIPS] [arXiv] [Code]

Cubic-Spline Flows
Conor Durkan*, Artur Bekasov*, Iain Murray, George Papamakarios
Workshop on Invertible Neural Nets and Normalizing Flows, ICML, 2019
* Equal contribution. Selected for a contributed talk
[arXiv] [Code]

Bayesian Adversarial Spheres: Bayesian Inference and Adversarial Examples in a Noiseless Setting
Artur Bekasov, Iain Murray
Bayesian Deep Learning Workshop, NeurIPS, 2018
Selected for a spotlight
[arXiv] [Poster]

See more at my Google Scholar page.

Code

See more at my GitHub page.

Teaching

While at the University of Edinburgh, I provided teaching support for:

Reviewing

I reviewed for the following conferences/workshops:

Contact