Welcome to my personal website, below you will find some of the things I've made and done:
Most recent news & events
ML in PL Conference 2023, that I co-organized, took place in Warsaw
Our paper titled "Generalized test utilities for long-tail performance in extreme multi-label classification" was accepted at NeurIPS 2023.
MLSS^S 2023 (Machine Learning Summer School on Applications in Science) that I co-organized took place at Jagiellonian University in Kraków.
New version of Vowpal Wabbit (9.7.0) was released with my improved implementation of PLT reduction.
Me and ViZDoom project joined The Farama Foundation.
ML in PL Conference 2022, that I co-organized, took place in Warsaw.
I started an internship at Yahoo in Paris, France.
Our paper titled "On missing labels, long-tails and propensities in extreme multi-label classification" was accepted at KDD'22.
New release of ViZDoom (1.1.12) library with a new official OpenAI Gym wrapper.
I joined the management board of ML in PL Association.
New release of ViZDoom (1.1.11) library with new features, including access to the game's audio buffer.
ML in PL Conference 2021, that I co-organized, took place virtualy.
Our paper "ViZDoom Competitions: Playing Doom From Pixels" received 2022 IEEE ToG Outstanding Paper Award.
Our short paper titled "Propensity-scored Probabilistic Label Trees" was accepted at SIGIR'21.
Our paper titled "Efficient Set-Valued Prediction in Multi-Class Classification" was accepted for publication in Data Mining and Knowledge Discovery.
I gave a talk titled "Probabilistic Label Trees in Vowpal Wabbit" at Vowpal Wabbit Workshop at NeurIPS'20. [ video]
I gave a talk titled "Extreme classification: applications and algorithms" on GHOST Day: AMLC'20. [ video]
Publications
-
Generalized test utilities for long-tail performance in extreme multi-label classification
Erik Schultheis, Marek Wydmuch, Wojciech Kotłowski, Rohit Babbar, Krzysztof Dembczyński
Proceedings of the 37nd International Conference on Neural Information Processing Systems (NeurIPS 2023), 2023
[ paper] [ bibtex] -
On Missing Labels, Long-tails and Propensities in Extreme Multi-label Classification
Erik Schultheis, Marek Wydmuch, Rohit Babbar, Krzysztof Dembczyński
Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD '22), 2022
[ paper] [ poster] [ bibtex] -
Propensity-scored Probabilistic Label Trees
Marek Wydmuch, Kalina Jasinska-Kobus, Rohit Babbar, Krzysztof Dembczyński
Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR '21), 2021
[ code (napkinXC)] [ paper] [ poster] [ bibtex] -
Online probabilistic label trees
Kalina Jasinska-Kobus, Marek Wydmuch, Devanathan Thiruvenkatachari, Krzysztof Dembczyński
Proceedings of the 24th International Conference on Artificial Intelligence and Statistics (AISTATS 2021), 2021
[ code (napkinXC)] [ paper] [ poster] [ AISTATS 2021 talk] [ ICML XC 2020 talk] [ bibtex] -
Efficient set-valued prediction in multi-class classification
Thomas Mortier, Marek Wydmuch, Krzysztof Dembczyński, Eyke Hüllermeier, Willem Waegeman
Data Mining and Knowledge Discovery, 2021
[ paper] [ bibtex] -
Probabilistic Label Trees for Extreme Multi-label Classification
Kalina Jasinska-Kobus, Marek Wydmuch, Krzysztof Dembczyński
Submitted to Journal of Machine Learning Research, 2020
[ code (napkinXC)] [ bibtex] -
A no-regret generalization of hierarchical softmax to extreme multi-label classification
Marek Wydmuch, Kalina Jasinska-Kobus, Mikhail Kuznetsov, Robert Busa-Fekete, Krzysztof Dembczyński
Proceedings of the 32nd International Conference on Neural Information Processing Systems (NeurIPS 2018), 2018
[ code (extremeText)] [ paper] [ poster] [ bibtex] -
ViZDoom Competitions: Playing Doom From Pixels
Marek Wydmuch, Michał Kempka, Wojciech Jaśkowski
IEEE Transactions on Games, 2018
Note: 2022 IEEE ToG Outstanding Paper Award
[ paper] [ movies] [ bibtex] -
ViZDoom: A Doom-based AI research platform for visual reinforcement learning
Michał Kempka, Marek Wydmuch, Grzegorz Runc, Jakub Toczek, Wojciech Jaśkowski
2016 IEEE Conference on Computational Intelligence and Games (CIG 2016), 2016
Note: 2016 IEEE CIG The Best Paper Award
[ www] [ code] [ paper] [ bibtex]
Software
ViZDoom
Doom-based AI Research Platform for Reinforcement Learning from Raw Visual Information.[ www] [ code]
napkinXC
Extremely simple and fast library for extreme multi-class and multi-label classification.[ code]
extremeText
Extension of fastText library for multi-label classification including extreme cases with hundreds of thousands and millions of labels.[ code]
ML in PL
I'm proud to be a member of the awesome ML in PL Association, a non-profit organization devoted to fostering the machine learning community in Poland and around the world.
From 2021 I'm a member of the management board of ML in PL Association and try to support other members of the association in the organization of great machine learning-related events, including taking care of the recruitment of new members of the association and leading the development of new websites for ML in PL Conference 2022, ML in PL Conference 2023 and MLSS^S 2023
Before, I co-organized ML in PL Conference 2019, ML in PL Virtual Event 2020, and ML in PL Conference 2021 serving as Call for Contributions (Talks and Posters) coordinator.
Reviewing scientific papers
I try to review at least a dozen of papers each year. So far I have served as a reviewer for:- Conferences:
- NeurIPS 2020, 2021 (Outstanding Reviewer Award - top 8%), 2022 (Top Reviewer Award - top 8%), and 2023
- ICML 2021, 2022, and 2023
- ICLR 2022, 2023, and 2024
- IJCAI 2020, and 2021, and also AISTATS 2021
-
Journals:
- IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI)
- Machine Learning Journal
- Transactions on Machine Learning Research (TMLR)
Teaching
The one that learns the most from a course is a teacher.As part of my PhD/being an assistant, I teached/teach a few courses at Poznan University of Technology (PUT) related to the topics of Machine Learning, Artificial Intelligence, Data Science and Big Data:Wojciech Jaśkowski (my Eng. thesis supervisor)
- Advanced Methods of Computational Intelligence 2021, 2022 and 2023 (lectures + exercises)
based on Richard S. Sutton and Andrew G. Barto, Reinforcement Learning: An Introduction and selected chapters from Stuart Russell and Peter Norvig, Artificial Intelligence A Modern Approach. - Methods of Artificial and Computational Intelligence 2021 (exercises)
based on selected chapters from Stuart Russell and Peter Norvig, Artificial Intelligence A Modern Approach. - Big Data Processing 2019/2020 and Processing of Massive Datasets 2018/2019 (exercises)
based on selected chapters from Jure Leskovec, Anand Rajaraman, Jeff Ullman, Mining of Massive Datasets. - Information Theory and Lossless Compression Methods 2018 (exercises)
based on the first chapters from David J.C. MacKay, Information Theory, Inference, and Learning Algorithms.
Note: The course was awarded as the best new course in 2018 at the Institute of Computer Science.