I work as a postdoctoral researcher in the Computer Vision group supervised by Prof. Dr. Michael Moeller at the University of Siegen since April 2023. Previously, I was a Ph.D. candidate supervised by Prof Dr.-Ing. Margret Keuper in the Computer Vision and Machine Learning group at the Max-Planck-Institute for Informatics and in the focus group of Computer Vision in the Data and Web Science Group at the University of Mannheim.
Together with Aaron Klein, Arber Zela, Giovanni Zappella and Rhea Sukthanker I also organize the AutoML Seminars as part of the ELLIS units Berlin and Freiburg. I was part of the organization team of the second workshop on neural architecture search @ ICLR 2021.
My research mainly focuses on efficient, unsupervised graph representations and embeddings for supervised surrogate models for neural architecture search.
- 12/2023 Our paper Improving Native CNN Robustness with Filter Frequency Regularization was accepted at TMLR!
- 11/2023 I presented my work about Multi-Objective Performance Prediction for Neural Architecture Search at the Doctoral Consortium at BMVC 2023!
- 10/2023 We have one workshop paper (Implicit Representations for Image Segmentation) accepted to Unireps@NeurIPS
- 07/2023 Our paper An Evaluation of Zero-Cost Proxies - from Neural Architecture Performance to Model Robustness was accepted at GCPR 2023
- 07/2023 I successfully defended my Ph.D. thesis with the title Topology Learning for Prediction, Generation, and Robustness in Neural Architecture Search
- 04/2023 I started as a PostDoc at the University of Siegen in the Group of Prof. Dr. Michael Moeller
- 01/2023 Our paper Neural Architecture Design and Robustness: A Dataset was accepted at ICLR 2023
- 09/2022 Our paper Learning Where to Look - Generative NAS is Surprisingly Efficient was accepted at ECCV 2022
Improving Native CNN Robustness with Filter Frequency Regularization
J. Lukasik*, P. Gavrikov*, J.Keuper, M. Keuper
Implicit Representations for Image Segmentation
J. P. Schneider, M. Fatima, J. Lukasik, A. Kolb, M. Keuper, M. Moeller
An Evaluation of Zero-Cost Proxies - from Neural Architecture Performance to Model Robustness
J. Lukasik, M. Moeller, M. Keuper
GCPR 2023 (oral)
Differentiable Architecture Search: a One-Shot Method?
J. Lukasik*, J. Geiping*, M. Moeller, M. Keuper
AutoML Conference 2023 Workshops
Surrogate NAS Benchmarks: Going Beyond the Limited Search Spaces of Tabular NAS Benchmarks
A. Zela*, JN. Siems*, L. Zimmer*, J. Lukasik, M. Keuper, F. Hutter
Neural Architecture Performance Prediction Using Graph Neural Networks
J. Lukasik, D. Friede, H. Stuckenschmidt, M. Keuper