André Biedenkapp

Researcher at the Machine Learning Lab in Freiburg.

andre_defense_headshot.png

Room 00-011

Georges-Köhler-Allee 74

79110 Freiburg, Germany

My 3 word address: ///forecast.gamer.showcase

I am a researcher at the University of Freiburg, Germany. My primary research interest is in the field of artificial intelligence, with a focus on automated machine learning and algorithm configuration, i.e., the problem of automatically tuning (machine learning) algorithms to maximize their performance. In particular, I focus on using reinforcement learning to tackle the problem of dynamically configuring algorithms.

I completed my bachelor’s degree in 2015 and my master’s degree in 2017 in computer science at the University of Freiburg. From February 2018 to October 2022, I did my Ph.D. at the University of Freiburg, at the Machine Learning Chair under the supervision of Prof. Dr. Frank Hutter and Prof. Dr. Marius Lindauer (Leibniz University Hannover). In October 2022 I successfully defended my PhD (Dr. rer. nat.) with the topic Dynamic Algorithm Configuration by Reinforcement Learning.


news

Sep 12, 2024 Our AutoML’24 paper HPO-RL-Bench: A Zero-Cost Benchmark for HPO in Reinforcement Learning was awarded runner up for the best paper award!
Aug 1, 2024 We have two new RL papers accepted at the European Workshop on Reinforcement Learning (EWRL’24). I hope to see you there!
Jul 12, 2024 I’m happy to contribute to this years AutoML conference with a workshop paper and a tutorial. The paper discusses how to exploit structure in the configuration space for DAC by RL. The tutorial will be held jointly with Theresa Eimer on AutoRL. Looking forward to seeing you there!
Jun 7, 2024 Theresa Eimer and I will be giving a tutorial on AutoRL with a focus on applications to sustainability this September at the AutoML School 2024. I’m looking forward to seeing you there! You can find more information here.
May 15, 2024 Our research on improving zero-shot generalization of world models for Contextual RL has been accepted at the inaugural Reinforcement Learning Conference (RLC). You can find the preprint of the paper here. You should also check out Sai’s twitter thread on the paper here.
Apr 3, 2024 I’m co-orgainzing the AutoRL Workshop at ICML 2024. The workshop will be held on July 26th or 27th (to be confirmed). The workshop will feature invited talks, contributed talks, and a poster session. The call for papers will be out soon. I’m looking forward to seeing you there!
Mar 15, 2024 Jointly with Theresa Eimer I will be giving a tutorial on AutoRL this October at ECAI 2024. I’m looking forward to seeing you there! You can find more information here.
Feb 12, 2024 In our latest paper we propose that attention-based Meta-RL can achieve improved generalization capabilities by learning transition dynamics as well as learning dynamics by paying attention to intra- as well as inter-episode experiences.
Jan 11, 2024 Raghu Rajan, Theresa Eimer, Aditya Mohan and I have written a blog post on the AutoRL blog about the year 2023 in AutoRL research. You can read it here.
Aug 17, 2023 September will be busy. Our (Auto)RL works will be presented in two upcoming venues: Make sure you come by and say hi!
Jul 4, 2023 Our latest paper presenting the MDP Playground has been accepted published by JAIR. With the MDPP, you can create and control your own cheap but challenging RL environments to better analyze and understand the behavior of RL algorithms.
Jun 5, 2023 Our latest TMLR paper on contextual RL discusses how cRL allows principled study on generalization in RL and proposes a flexible benchmark suite for cRL.
Check out the paper here and the benchmark here.
Mar 10, 2023 I’ll be chairing the online experience of this year’s AutoML Conference jointly with Hayeon Lee, Mohamed Abdelfattah and Richard Song. Checkout the blog post here for more details.
Jan 30, 2023 Our paper proposing a novel gray-box BO technique for AutoRL was accepted at ICLR’23.
Dec 1, 2022 Our recent paper in which we discuss the past, present and future of dynamic algorithm configuration (DAC) has been accepted by the Journal for Artificial Intelligence Research.
Nov 11, 2022 We have two new workshop papers tackling AutoRL accepted at MetaLearn@NeurIPS’22:
Oct 14, 2022 🎉 I successfully defended my PhD thesis with the title Dynamic Algorithm Configuration by Reinforcement Learning with summa cum laude (the best possible grade) 🥳
Sep 20, 2022 On the 10th of November I’ll be giving a talk about my research in the Seminar on Advances in Probabilistic Machine Learning of the Aalto University and ELLIS unit Helsinki.
Aug 11, 2022 I am Chair of COSEAL jointly with Alexander Tornede and Lennart Shäpermeier. You can find the announcement here.
Jul 13, 2022 Our GECCO’22 paper Theory-Inspired Parameter Control Benchmarks for Dynamic Algorithm Configuration won the best paper award on the GECH track.
Jun 24, 2022 We just uploaded the talk for ou GECCO’22 paper Theory-Inspired Parameter Control Benchmarks for Dynamic Algorithm Configuration which is nominated for the best paper award.
Jun 19, 2022 Checkout our new paper on DeepCAVE, a tool to analyze and explain AutoML meta-data, which was just accepted at the ReALML@ICML workshop.
Jun 3, 2022 I am co-organizing the 2nd AutoML Fall School. The fall school will be held from 10th - 13th of October. Checkout the exciting list of invited speakers and hands-on sessions.
Jun 3, 2022 We just released a new paper on DAC for AI Planning. I’ll present the paper on the 13th of June at the PRL Workshop. You can also checkout the recorded talk.
Jun 1, 2022 Our survey on AutoRL has been published in the Journal of Artificial Intelligence Research.
May 30, 2022 We just released a new paper on Dynamic Algorithm Configuration (DAC) in which we discuss the DAC journey so far and give a glimpse of the future of DAC.
May 18, 2022 Our GECCO’22 paper Theory-Inspired Parameter Control Benchmarks for Dynamic Algorithm Configuration has been nominated for the best paper award.
Apr 16, 2022 The paper Theory-Inspired Parameter Control Benchmarks for Dynamic Algorithm Configuration is accepted for publication in the proceedings of the Genetic and Evolutionary Computation Conference (GECCO’22).
Apr 16, 2022 I created this personal webpage :grin:

selected publications

  1. Dreaming of Many Worlds: Learning Contextual World Models Aids Zero-Shot Generalization
    Sai Prasanna*, Karim Farid*, Raghu Rajan, and André Biedenkapp
    Reinforcement Learning Journal, 3, pp. 1317–1350, 2024 *Joint first authorship
    Note: To also be presented at the Seventeenth European Workshop on Reinforcement Learning (EWRL 2024)
  2. Theory-inspired Parameter Control Benchmarks for Dynamic Algorithm Configuration
    André Biedenkapp*, Nguyen Dang*, Martin S. Krejca*, Frank Hutter, and Carola Doerr
    In Proceedings of the Genetic and Evolutionary Computation Conference (GECCO’22), pp. 766–775, 2022 *Joint first authorship 🏅Won the best paper award on the GECH track.
  3. Automated Reinforcement Learning (AutoRL): A Survey and Open Problems
    Jack Parker-Holder, Raghu Rajan, Xingyou Song, André Biedenkapp, Yingjie Miao, Theresa Eimer, Baohe Zhang, Vu Nguyen, Roberto Calandra, Aleksandra Faust, Frank Hutter, and Marius Lindauer
    Journal of Artificial Intelligence Research (JAIR), 74, pp. 517-568, 2022
  4. TempoRL: Learning When to Act
    André BiedenkappRaghu RajanFrank Hutter, and Marius Lindauer
    In Proceedings of the 38th International Conference on Machine Learning (ICML 2021), 139, pp. 914–924, 2021
  5. Dynamic Algorithm Configuration: Foundation of a New Meta-Algorithmic Framework
    André Biedenkapp, Furkan H Bozkurt, Theresa EimerFrank Hutter, and Marius Lindauer
    In Proceedings of the Twenty-fourth European Conference on Artificial Intelligence (ECAI’20), pp. 427–434, 2020