Bernhard Klein
Bernhard Klein is a fifth-year PhD student at the Computing Systems Group lead by Prof. Dr. Holger Fröning at Heidelberg University. He completed his Master of Science at the Institute for Environmental Physics in 2018. His research interests include code generation, machine learning, probabilistic, approximate and noisy computing and all means of closing the gap between machine learning and embedded devices.
Research interests
- Probabilistic Machine Learning, especially Bayesian Neural Networks
- Hardware-aware Machine Learning, includes Neural Architecture Search, Automatic Compression, Machine Learning Compiler,…
- Embedded Resource-constrained Systems
- Analog/Optical Energy-efficient Hardware
Recent news (2-year horizon)
- 03/2023: Invited Talk at the 4th Workshop on Embedded Machine Learning (WEML) - “Galen: HW-specific Automatic Compression”
Recent Service (4-year horizon)
Reviewer
- 2023: IET Computers & Digital Techniques
- 2021: ITEM Workshop
Subreviewer
- 2024: International Conference on Computer Design (ICCD)
- 2024: International Conference on Pattern Recognition (ICPR)
- 2024: International Conference on Computer-Aided Design (ICCAD)
- 2024: European Conference on Parallel and Distributed Computing (Euro-PAR)
- 2024: International Conference on Supercomputing (ICS)
- 2023: Conference on Parallel Processing (ICPP)
- 2022: International Conference on Field-Programmable Logic and Applications (FPL)
- 2021: Conference on Parallel Processing (ICPP)
- 2021: International Conference on Field-Programmable Logic and Applications (FPL)
- 2020: International Conference on Field-Programmable Logic and Applications (FPL)
Recent Teaching (4-year horizon)
Summer term 2024 Lecturer - graduate course “Embedded Machine Learning (2+2)”*
Publications
- Resource-Efficient Neural Networks for Embedded SystemsJournal of Machine Learning Research, 25(50), 1–51, 2024
@article{JMLR:v25:18-566, author = {Roth, Wolfgang and Schindler, G{{\"u}}nther and Klein, Bernhard and Peharz, Robert and Tschiatschek, Sebastian and Fr{\"{o}}ning, Holger and Pernkopf, Franz and Ghahramani, Zoubin}, title = {Resource-Efficient Neural Networks for Embedded Systems}, journal = {Journal of Machine Learning Research}, year = {2024}, volume = {25}, number = {50}, pages = {1--51}, url = {http://jmlr.org/papers/v25/18-566.html}, }
- Walking Noise: On Layer-Specific Robustness of Neural Architectures against Noisy Computations and Associated Characteristic Learning DynamicsEuropean Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD), 2024
@inproceedings{borras2024, title = {Walking Noise: On Layer-Specific Robustness of Neural Architectures against Noisy Computations and Associated Characteristic Learning Dynamics}, author = {Borras, Hendrik and Klein, Bernhard and Fr{\"{o}}ning, Holger}, booktitle = {European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases}, year = {2024}, series = {ECML-PKDD}, url = {https://doi.org/10.1007/978-3-031-70359-1_3}, }
- Function Space Diversity for Uncertainty Prediction via Repulsive Last-Layer EnsemblesICML 2024 Workshop on Structured Probabilistic Inference & Generative Modeling, 2024
@inproceedings{steger2024function, title = {Function Space Diversity for Uncertainty Prediction via Repulsive Last-Layer Ensembles}, author = {Steger, Sophie and Knoll, Christian and Klein, Bernhard and Fr{\"o}ning, Holger and Pernkopf, Franz}, booktitle = {ICML 2024 Workshop on Structured Probabilistic Inference {\&} Generative Modeling}, year = {2024}, url = {https://openreview.net/forum?id=FbMN9HjgHI}, }
- Probabilistic Photonic Computing with Chaotic LightCoRR, abs/2401.17915, 2024
@article{brckerhoffplckelmann2024probabilistic, title = {Probabilistic Photonic Computing with Chaotic Light}, author = {Brückerhoff-Plückelmann, Frank and Borras, Hendrik and Klein, Bernhard and Varri, Akhil and Becker, Marlon and Dijkstra, Jelle and Brückerhoff, Martin and Wright, C. David and Salinga, Martin and Bhaskaran, Harish and Risse, Benjamin and Fr{\"o}ning, Holger and Pernice, Wolfram}, year = {2024}, volume = {abs/2401.17915}, journal = {CoRR}, url = {https://arxiv.org/abs/2401.17915}, }
- On the Non-Associativity of Analog ComputationsCoRR, abs/2309.14292, 2023
@article{DBLP:journals/corr/abs-2309-14292, author = {Kuhn, Lisa and Klein, Bernhard and Fr{\"{o}}ning, Holger}, title = {On the Non-Associativity of Analog Computations}, journal = {CoRR}, volume = {abs/2309.14292}, year = {2023}, url = {https://arxiv.org/abs/2309.14292}, doi = {10.48550/ARXIV.2309.14292}, eprinttype = {arXiv}, eprint = {2309.14292}, timestamp = {Wed, 27 Sep 2023 01:00:00 +0200}, }
- Towards Hardware-Specific Automatic Compression of Neural NetworksCoRR, abs/2212.07818, 2022
@article{DBLP:journals/corr/abs-2212-07818, author = {Krieger, Torben and Klein, Bernhard and Fr{\"{o}}ning, Holger}, title = {Towards Hardware-Specific Automatic Compression of Neural Networks}, journal = {CoRR}, volume = {abs/2212.07818}, year = {2022}, url = {https://arxiv.org/abs/2212.07818}, doi = {10.48550/ARXIV.2212.07818}, eprinttype = {arXiv}, eprint = {2212.07818}, timestamp = {Mon, 02 Jan 2023 00:00:00 +0100}, }
- Towards Addressing Noise and Static Variations of Analog Computations Using Efficient RetrainingMachine Learning and Principles and Practice of Knowledge Discovery in Databases - International Workshops of ECML PKDD 2021, Proceedings Part I (Communications in Computer and Information Science), 1524, 409–420, Springer, 2021
@inproceedings{DBLP:conf/pkdd/KleinKWESSF21, author = {Klein, Bernhard and Kuhn, Lisa and Weis, Johannes and Emmel, Arne and Stradmann, Yannik and Schemmel, Johannes and Fr{\"{o}}ning, Holger}, editor = {Kamp, Michael and Koprinska, Irena and Bibal, Adrien and Bouadi, Tassadit and Fr{\'{e}}nay, Beno{\^{\i}}t and Gal{\'{a}}rraga, Luis and Oramas, Jos{\'{e}} and Adilova, Linara and Krishnamurthy, Yamuna and Kang, Bo and Largeron, Christine and Lijffijt, Jefrey and Viard, Tiphaine and Welke, Pascal and Ruocco, Massimiliano and Aune, Erlend and Gallicchio, Claudio and Schiele, Gregor and Pernkopf, Franz and Blott, Michaela and Fr{\"{o}}ning, Holger and Schindler, G{\"{u}}nther and Guidotti, Riccardo and Monreale, Anna and Rinzivillo, Salvatore and Biecek, Przemyslaw and Ntoutsi, Eirini and Pechenizkiy, Mykola and Rosenhahn, Bodo and Buckley, Christopher L. and Cialfi, Daniela and Lanillos, Pablo and Ramstead, Maxwell and Verbelen, Tim and Ferreira, Pedro M. and Andresini, Giuseppina and Malerba, Donato and Medeiros, Ib{\'{e}}ria and Fournier{-}Viger, Philippe and Nawaz, M. Saqib and Ventura, Sebasti{\'{a}}n and Sun, Meng and Zhou, Min and Bitetta, Valerio and Bordino, Ilaria and Ferretti, Andrea and Gullo, Francesco and Ponti, Giovanni and Severini, Lorenzo and Ribeiro, Rita P. and Gama, Jo{\~{a}}o and Gavald{\`{a}}, Ricard and Cooper, Lee A. D. and Ghazaleh, Naghmeh and Richiardi, Jonas and Roqueiro, Damian and Miranda, Diego Saldana and Sechidis, Konstantinos and Gra{\c{c}}a, Guilherme}, title = {Towards Addressing Noise and Static Variations of Analog Computations Using Efficient Retraining}, booktitle = {Machine Learning and Principles and Practice of Knowledge Discovery in Databases - International Workshops of {ECML} {PKDD} 2021, Proceedings Part {I}}, series = {Communications in Computer and Information Science}, volume = {1524}, pages = {409--420}, publisher = {Springer}, year = {2021}, url = {https://doi.org/10.1007/978-3-030-93736-2_32}, doi = {10.1007/978-3-030-93736-2\_32}, }
- Understanding Cache Boundness of ML Operators on ARM ProcessorsCoRR, abs/2102.00932, 2021
@article{DBLP:journals/corr/abs-2102-00932, author = {Klein, Bernhard and Gratl, Christoph and M{\"{u}}cke, Manfred and Fr{\"{o}}ning, Holger}, title = {Understanding Cache Boundness of {ML} Operators on {ARM} Processors}, journal = {CoRR}, volume = {abs/2102.00932}, year = {2021}, url = {https://arxiv.org/abs/2102.00932}, eprinttype = {arXiv}, eprint = {2102.00932}, timestamp = {Thu, 14 Oct 2021 01:00:00 +0200}, }