Fedor Sergeev, MSc
"With four parameters I can fit an elephant, and with five I can make him wiggle his trunk." - John von Neumann
PhD Student
- fedor.sergeev@inf.ethz.ch
- Address
-
Department of Computer Science
Biomedical Informatics Group
Universitätstrasse 6
8092 Zürich - Room
- CAB F53
Informing machine learning models with human insights for better performance, reliability, and interpretability
I did my BSc in Applied Mathematics and Physics at MIPT. I developed computational methods for simulations in physics under the supervision of Igor Petrov and Nikolay Khoklov. In parallel, I worked on applications of deep learning in high-energy physics at GSI and LAMBDA.
In my MSc I studied Computational Sciences and Engineering EPFL, combining my interest in numerical and data-driven modeling. My thesis with Pascal Fua and Jonathan Donier was on physics-informed neural networks for modeling fluid flow. During my studies, I interned at startup companies Spiden and Neural Concept, working on synthetic data generation for medical spectroscopic data and 3D computer vision for advanced engineering, respectively.
I joined BMI lab in July 2023 to work on multimodal, representation and Bayesian deep learning on intensive care unit (ICU) data. I am also an ELLIS PhD student, co-supervised by Vincent Fortuin.
Latest Publications
Abstract Knowing which features of a multivariate time series to measure and when is a key task in medicine, wearables, and robotics. Better acquisition policies can reduce costs while maintaining or even improving the performance of downstream predictors. Inspired by the maximization of conditional mutual information, we propose an approach to train acquirers end-to-end using only the downstream loss. We show that our method outperforms random acquisition policy, matches a model with an unrestrained budget, but does not yet overtake a static acquisition strategy. We highlight the assumptions and outline avenues for future work.
Authors Fedor Sergeev, Paola Malsot, Gunnar Rätsch, Vincent Fortuin
Submitted SPIGM ICML Workshop