Matthias Feurer
@matthiasfeurer.bsky.social
Assistant Professor "AutoML and Optimization" @ Lamarr institute and @tu-dortmund.bsky.social. Working on #AutoML, #HPO, #TabularData #OpenML #Benchmarking | https://matthiasfeurer.de
This helps newer contributors to OpenML who were not part of the author list of the 2014 paper with attribution and allows them to continue to contribute to #OpenML and open source software. Thank you :) 2/2
October 6, 2025 at 1:51 PM
This helps newer contributors to OpenML who were not part of the author list of the 2014 paper with attribution and allows them to continue to contribute to #OpenML and open source software. Thank you :) 2/2
Curious about my research & past role? Check out this recent interview: bsky.app/profile/muni...
Stay tuned for more updates on AutoML and my new group soon!
Stay tuned for more updates on AutoML and my new group soon!
Matthias Feurer, Thomas Bayes Fellow at #MCML & interim prof at #LMU, builds tools to make ML easier for scientists & more efficient for experts. Next semester he starts as junior prof at TU Dortmund—sharing in our interview how the fellowship helped him get there: mcml.ai/news/2025-09...
October 1, 2025 at 8:25 AM
Curious about my research & past role? Check out this recent interview: bsky.app/profile/muni...
Stay tuned for more updates on AutoML and my new group soon!
Stay tuned for more updates on AutoML and my new group soon!
A better understanding of overtuning will help the #AutoML community design more reliable optimization and selection strategies, especially in small-data regimes where robustness is critical. 🚀
September 30, 2025 at 7:08 AM
A better understanding of overtuning will help the #AutoML community design more reliable optimization and selection strategies, especially in small-data regimes where robustness is critical. 🚀
Our large-scale analysis shows overtuning is more common than we expected, occasionally severe, and influenced by factors like resampling strategy, dataset size, and model class.
September 30, 2025 at 7:08 AM
Our large-scale analysis shows overtuning is more common than we expected, occasionally severe, and influenced by factors like resampling strategy, dataset size, and model class.
We formally define and study overtuning, a phenomenon where hyperparameter optimization improves validation scores but hurts true generalization. In addition, we relate it to (meta-)overfitting and put recent literature in perspective.
September 30, 2025 at 7:08 AM
We formally define and study overtuning, a phenomenon where hyperparameter optimization improves validation scores but hurts true generalization. In addition, we relate it to (meta-)overfitting and put recent literature in perspective.
If you’re looking for a #PhDPositions, check out automl4science.de/files/2025_JobPosting_TUDortmund.pdf
I’m grateful for @slds-lmu.bsky.social @munichcenterml.bsky.social @lmumuenchen.bsky.social l for 2.5 inspiring years with great mentorship and collaborations.
I’m grateful for @slds-lmu.bsky.social @munichcenterml.bsky.social @lmumuenchen.bsky.social l for 2.5 inspiring years with great mentorship and collaborations.
automl4science.de
August 20, 2025 at 11:45 AM
If you’re looking for a #PhDPositions, check out automl4science.de/files/2025_JobPosting_TUDortmund.pdf
I’m grateful for @slds-lmu.bsky.social @munichcenterml.bsky.social @lmumuenchen.bsky.social l for 2.5 inspiring years with great mentorship and collaborations.
I’m grateful for @slds-lmu.bsky.social @munichcenterml.bsky.social @lmumuenchen.bsky.social l for 2.5 inspiring years with great mentorship and collaborations.
Reposted by Matthias Feurer
Also, a big shoutout to the TRL organizers @trl-research.bsky.social 🙏
July 31, 2025 at 3:05 PM
Also, a big shoutout to the TRL organizers @trl-research.bsky.social 🙏
Reposted by Matthias Feurer
openreview.net/forum?id=vLt...
TL;DR: A SoftKNN-based training task makes your ICL model for predictive tabular tasks more intelligible while empirically not decreasing performance.
TL;DR: A SoftKNN-based training task makes your ICL model for predictive tabular tasks more intelligible while empirically not decreasing performance.
In-Context Learning of Soft Nearest Neighbor Classifiers for...
With in-context learning foundation models like TabPFN excelling on small supervised tabular learning tasks, it has been argued that "boosted trees are not the best default choice when working with...
openreview.net
July 31, 2025 at 3:05 PM
openreview.net/forum?id=vLt...
TL;DR: A SoftKNN-based training task makes your ICL model for predictive tabular tasks more intelligible while empirically not decreasing performance.
TL;DR: A SoftKNN-based training task makes your ICL model for predictive tabular tasks more intelligible while empirically not decreasing performance.
Check out openml.org and also join the OpenML Slack to learn about latest developments and upcoming workshops: openml.slack.com/join/shared_... 4/4
OpenML
OpenML is an open platform for sharing datasets, algorithms, and experiments - to learn how to learn better, together.
openml.org
July 7, 2025 at 12:15 PM
Check out openml.org and also join the OpenML Slack to learn about latest developments and upcoming workshops: openml.slack.com/join/shared_... 4/4
Joaquin Vanschoren, Jos van der Welde and Marcel Wever. 3/4
July 7, 2025 at 12:15 PM
Joaquin Vanschoren, Jos van der Welde and Marcel Wever. 3/4
with @slds-lmu.bsky.social, @giuseppe88.bsky.social, @tan-iya.bsky.social, myself, Sebastian Fischer, Pieter Gijsbers, Subhaditya Mukherjee, Andreas C. Müller, László Németh, @luisoala.bsky.social, @lennartpurucker.bsky.social, @sahithya.bsky.social, Jan N. van Rijn, @prabhant.bsky.social 2/4
July 7, 2025 at 12:15 PM
with @slds-lmu.bsky.social, @giuseppe88.bsky.social, @tan-iya.bsky.social, myself, Sebastian Fischer, Pieter Gijsbers, Subhaditya Mukherjee, Andreas C. Müller, László Németh, @luisoala.bsky.social, @lennartpurucker.bsky.social, @sahithya.bsky.social, Jan N. van Rijn, @prabhant.bsky.social 2/4
Very proud our paper "OpenML: Insights from 10 years and more than a thousand papers" that describes the current state of OpenML and also analyzes the impact OpenML had over the last years (yes, we manually looked at every paper citing OpenML). #openml #automl #openscience 1/4
July 7, 2025 at 12:15 PM
Very proud our paper "OpenML: Insights from 10 years and more than a thousand papers" that describes the current state of OpenML and also analyzes the impact OpenML had over the last years (yes, we manually looked at every paper citing OpenML). #openml #automl #openscience 1/4