- Machine-learning-nederland

BeNeRL Reinforcement Learning Seminar: Daniel Palenicek from TU Darmstadt (September 12, this Thursday)
by Yang, Z. (Zhao) 09 Sep '24

09 Sep '24

Dear colleagues, Our next BeNeRL Reinforcement Learning Seminar (Sep 12) is coming: Speaker: Daniel Palenicek (https://www.ias.informatik.tu-darmstadt.de/Team/DanielPalenicek), PhD student from TU Darmstadt. Title: Sample Efficiency in Deep RL: Quo Vadis? Date: September 12, 16.00-17.00 (CET) Please find full details about the talk below this email and on the website of the seminar series:https://www.benerl.org/seminar-series The goal of the online BeNeRL seminar series is to invite RL researchers (mostly advanced PhD or early postgraduate) to share their work. In addition, we invite the speakers to briefly share their experience with large-scale deep RL experiments, and their style/approach to get these to work. We would be very glad if you forward this invitation within your group and to other colleagues that would be interested (also outside the BeNeRL region). Hope to see you on September 12! Kind regards, Zhao Yang & Thomas Moerland Leiden University —————————————————————— Upcoming talk: Date: September 12, 16.00-17.00 (CET) Speaker: Daniel Palenicek (https://www.ias.informatik.tu-darmstadt.de/Team/DanielPalenicek) Title: Sample Efficiency in Deep RL: Quo Vadis? Zoom: https://universiteitleiden.zoom.us/j/65411016557?pwd=MzlqcVhzVzUyZlJKTEE0Nk… Abstract: Deep reinforcement learning (RL) has shown remarkable successes but is often hindered by low sample efficiency and high computational costs. This talk presents two complementary studies that challenge conventional wisdom in deep RL. Both studies offer a fresh perspective on accelerating RL algorithms and highlight some fundamental limitations. First, we explore the limits of value expansion methods in model-based RL, revealing surprising insights about the diminishing returns of longer rollout horizons and increased model accuracy. Our findings suggest that pursuing perfect models may not be as crucial as previously thought. Second, we introduce CrossQ, a novel approach that dramatically improves sample efficiency in off-policy RL by leveraging batch normalization and eliminating target networks. Contrary to other approaches, CrossQ does not increase the update-to-data ratio and, thus, achieves its state-of-the-art performance at just 5% of the computational cost of other current methods. We conclude by discussing implications for future research directions, including applications in robotics and large-scale RL systems. Bio: Daniel Palenicek is a PhD student at the Intelligent Autonomous System Group, TU Darmstadt, where Prof. Jan Peters advises him. He is also a part of the 3AI project with hessian.AI. Daniel’s research lies at the intersection of reinforcement learning and robotics. He is interested in increasing sample efficiency and scaling model-free and model-based reinforcement learning algorithms. Before starting his Ph.D., Daniel completed his B.Sc. and M.Sc. in Wirtschaftsinformatik at TU Darmstadt. He wrote his Master's thesis entitled "Dyna-Style Model-Based Reinforcement Learning with Value Expansion" under the supervision of Dr. Michael Lutter and Prof. Jan Peters. Prior, Daniel did two research internships. At the Bosch Center for AI he focused on model-free RL, and at Huawei Noah’s Ark Lab in London, he worked on safe model-based RL and active exploration.

1 0

PhD position at TU Eindhoven
by Bahram Zonooz 29 Aug '24

29 Aug '24

Dear All, Mykola Pechenizkiy and I are looking for a PhD candidate to work on Foundation Models for Embedded and Embodied AI in Autonomous Driving. This project is part of a large program in collaboration with several academic and industrial partners, including major OEMs in Europe, with a significant impact on the future of safe autonomous driving deployment. You will be developing foundation models for both embedded and embodied AI, focusing on LLM-guided sparsity and learning agents. Your work will involve language-based navigation, training, labeling, and data curation, as well as developing more intuitive LLM-based active learning approaches. Ultimately, you will work towards creating LLM-based scenario descriptions for actor generation in simulations. https://jobs.tue.nl/en/vacancy/phd-on-foundation-models-for-embedded-and-em… Best, Bahram

1 0

PhD position in AI-based satellite data assimilation at TU Delft and BFH
by Angela Meyer 22 Aug '24

22 Aug '24

Dear all, We invite applications for a 4-year PhD position in AI-based satellite data assimilation at TU Delft and BFH to improve weather forecasts. Our group offers a stimulating collaborative research environment and a strong network with industry partners. The PhD student will be enrolled at TU Delft. The project is in collaboration with a weather company. What you'll be doing * Develop machine learning approaches for the assimilation of satellite data in numerical weather prediction to improve weather forecasts * Carry out high-quality research at the intersection of machine learning, satellite remote sensing, and numerical weather prediction * Publish your results in leading international journals and complete your doctorate at TU Delft * Present your work at international conferences * Have the opportunity to assist in teaching and co-supervise undergraduate students Job profile * MSc degree (or equivalent) in physics, computer science, atmospheric/geoscience, data science, engineering, applied mathematics, or a related field * Practical experience in programming (e.g., Python), machine learning and data analysis * Strong analytical and programming skills * Fluency in English, good scientific writing and communication skills * Being self-motivated and proactive, you enjoy working with complex topics and thrive on challenges * Experience in atmospheric research, data assimilation or numerical weather prediction is not necessarily required but will be an advantage Desired start date is October 2024. This may be adjusted by mutual agreement. The position is a full-time position. The annual salary corresponds to approximately 56’000 EUR. To apply, submit your CV, BSc and MSc transcripts, names and contact details of 2–3 references, BSc/MSc theses and publications (if any) to angela.meyer(a)bfh.ch<mailto:angela.meyer@bfh.ch>. Best regards, Angela

1 0

PhD position at TU Delft
by Angela Meyer 22 Aug '24

22 Aug '24

Dear all, We invite applications for a 4-year PhD position in solar energy forecasting at TU Delft and BFH to improve solar energy forecasts. Our group offers a stimulating collaborative research environment and a strong network with industry partners. This project will be performed in collaboration with an energy company. Desired start date is October 2024. This may be adjusted by mutual agreement. The position is a full-time position. The annual salary corresponds to approximately 56’000 EUR. To apply, submit your CV, BSc and MSc transcripts, names and contact details of 2–3 references, BSc/MSc theses and publications (if any) to angela.meyer(a)bfh.ch<mailto:angela.meyer@bfh.ch>. What you'll be doing * Develop advanced solar energy forecasting methods to improve the predictability of solar power production * Carry out high-quality research at the intersection of solar energy, machine learning, and satellite remote sensing * Publish your results in leading international journals and complete your doctorate at TU Delft * Present your work at international conferences * Have the opportunity to assist in teaching and co-supervise undergraduate students Job profile * MSc degree (or equivalent) in physics, atmospheric/geoscience, computer science, data science, engineering, applied mathematics, or a related field * Practical experience in programming (e.g., Python), machine learning and data analysis * Strong analytical and programming skills * Fluency in English, good scientific writing and communication skills * Being self-motivated and proactive, you enjoy working with complex topics and thrive on challenges * Experience with atmospheric and energy-related research or solar forecasting is not necessarily required but will be an advantage Best regards, Angela

1 0

Fwd: [TIAI-seminar] Sanjoy Dasgupta, July 11
by Tim van Erven 11 Jul '24

11 Jul '24

Dear all, There is an online seminar today by Sanjoy Dasgupta <https://cseweb.ucsd.edu/~dasgupta/> at 15.00, which may be of broader interest. Sanjoy is internationally well known for his work on clustering and unsupervised learning, and has a reputation as an excellent speaker. His talk today will be about new results on interpretable clustering. See below for details. This is part of a new online seminar about Theory for Interpretable AI. For further seminar info, see tverven.github.io/tiai-seminar/ <http://tverven.github.io/tiai-seminar/> Best, Tim -------- Forwarded Message -------- Subject: [TIAI-seminar] Sanjoy Dasgupta, July 11 Date: Fri, 5 Jul 2024 21:37:36 +0300 From: Michal Moshkovitz via TIAI-seminar <tiai-seminar(a)list.uva.nl> Reply-To: Michal Moshkovitz <michal.moshkovitz(a)mail.huji.ac.il> To: tiai-seminar(a)list.uva.nl Dear all, Unsupervised learning is a significant branch of machine learning, with k-means clustering being a key subproblem. However, it often lacks interpretability. Can we make it interpretable? Apparently we can. On this topic, our next speaker in the Theory of Interpretable AI Seminar, Sanjoy Dasgupta, will review recent progress on interpretable k-means clustering. *Speaker*: Sanjoy Dasgupta <https://cseweb.ucsd.edu/~dasgupta/> *Date*: Thursday, July 11, 15.00 Central European Time (CET) / 9.00 am Eastern Standard Time (EST) *Zoom link*:https://uva-live.zoom.us/j/87120549999 *Title*: Recent progress on interpretable clustering The widely-used k-means procedure returns k clusters that have arbitrary convex shapes. In high dimension, such a clustering might not be easy to understand. A more interpretable alternative is to constrain the clusters to be the leaves of a decision tree with axis-parallel splits; then each cluster is a hyper-rectangle given by a small number of features. Is it always possible to find clusterings that are interpretable in this sense and yet have k-means cost that is close to the unconstrained optimum? A recent line of work has answered this in the affirmative and moreover shown that these interpretable clusterings are easy to construct. I will give a survey of these results: algorithms, methods of analysis, and open problems. *General Seminar info*: Web:tverven.github.io/tiai-seminar/ <http://tverven.github.io/tiai-seminar/> Google calendar:Google calendar<https://calendar.google.com/calendar/u/1?cid=NTlhNjNhZDQ5ZmUxYmM5MmRmZTMwNz…> *Upcoming speakers*: - September 5: Lesia Semenova - October 10: Ulrike von Luxburg Best Regards, Michal Moshkovitz Also on behalf of TIAI co-organizers Suraj Srinivas and Tim van Erven

1 0

Call for online participation: IAS Workshop on Logic and AI (July 16-17, 2024)
by Balder ten Cate 10 Jul '24

10 Jul '24

*Workshop on Logic and AI* *July 16-17, 2024* *Amsterdam Institute for Advanced Study (IAS)* This workshop is part of an IAS research project on the topic of “Logic and AI”. It will bring together international experts to explore the promising interaction of logic and modern artificial intelligence (AI). While AI struggles with explainability, interpretability, and verifiability, logic excels at this. So can logic help AI? And if so, how? The workshop has sessions on expressive and computational power of machine learning, neuro-symbolic integration, and causality, logic, and machine learning. Speakers include Giuseppe Marra (KU Leuven), Levin Hornischer (LMU Munich), Martin Grohe (RWTH Aachen), Lena Strobl (Umeå University), Herbert Jaeger (University of Groningen), Atticus Geiger (Pr(Ai)²R), Thomas Icard (Stanford University). Due to limited space, on-site participation is by invitation, but the talks can be followed online. If you would like to follow the talks online, please submit the registration form indicating online participation and you will receive a Zoom link. For more information, see: https://ias.uva.nl/content/events/2024/07/logic-and-ai.html

1 0

PhD position University of Twente
by Heide, R. de (Rianne) 17 Jun '24

17 Jun '24

Dear all, The University of Twente has a vacancy for a fully funded 4-year PhD position on Mathematical Statistics and/or Machine Learning Theory, under my supervision. If you want to work on e-values, multiple testing, bandits, Bayesian inference, or combine all of those, please apply before August 1st. Feel free to e-mail me if you have any questions. Please share with anyone who might be interested. https://utwentecareers.nl/en/vacancies/1822/phd-position-on-mathematical-st… Best regards, Rianne de Heide r.deheide(a)utwente.nl<mailto:r.deheide@utwente.nl> https://riannedeheide.github.io -- Per September 1st I will move to the University of Twente. My new e-mail address, which I already use, is: r.deheide(a)utwente.nl<mailto:r.deheide@utwente.nl>

1 0

BeNeRL Reinforcement Learning Seminar: Nicklas Hansen from UC San Diego (June 13, this Thursday)
by Yang, Z. (Zhao) 10 Jun '24

10 Jun '24

Dear colleagues, Our next BeNeRL Reinforcement Learning Seminar (June 13) is coming: Speaker: Nicklas Hansen (https://www.nicklashansen.com<https://www.nicklashansen.com/>), PhD student from University of California San Diego. Title: Data-Driven World Models for Robots Date: June 13, 16.00-17.00 (CET) Please find full details about the talk below this email and on the website of the seminar series: https://www.benerl.org/seminar-series The goal of the online BeNeRL seminar series is to invite RL researchers (mostly advanced PhD or early postgraduate) to share their work. In addition, we invite the speakers to briefly share their experience with large-scale deep RL experiments, and their style/approach to get these to work. We would be very glad if you forward this invitation within your group and to other colleagues that would be interested (also outside the BeNeRL region). Hope to see you on June 13! Kind regards, Zhao Yang & Thomas Moerland Leiden University —————————————————————— Upcoming talk: Date: June 13, 16.00-17.00 (CET) Speaker: Nicklas Hansen (https://www.nicklashansen.com<https://www.nicklashansen.com/>) Title: Data-Driven World Models for Robots Zoom: https://universiteitleiden.zoom.us/j/65411016557?pwd=MzlqcVhzVzUyZlJKTEE0Nk… Abstract: Recent progress in AI can be attributed to the emergence of large models trained on large datasets. However, teaching AI agents to reliably interact with our physical world has proven challenging and, consequently, this new paradigm has not materialized as much in robotics as in related areas. In this talk, I will share my perspective on why it is challenging, what an agent that "understands" our physical world may look like, and how to build it. Concretely, I will discuss our work on TD-MPC, a highly data-driven approach to world models (models of the physical world) that can learn from diverse data, improve autonomously through real-world interaction, and scale with data and model size. I will discuss its algorithmic foundations, practical considerations, applications to diverse robot embodiments (manipulation, locomotion, humanoids) in both simulation and the real world, and conclude by sharing my perspective on future research directions. Bio: Nicklas Hansen is a PhD student at University of California San Diego advised by Prof. Xiaolong Wang and Prof. Hao Su. His research focuses on developing generalist AI agents that learn from interaction with the physical and digital world. He has spent time at Meta AI (FAIR) and University of California Berkeley (BAIR), and received his BS and MS degrees from Technical University of Denmark. He is a recipient of the 2024 NVIDIA Graduate Fellowship, and his work has been featured at top venues in machine learning and robotics.

1 0

Final call to register for the AIM workshop on June 13-14
by Tim van Erven 03 Jun '24

03 Jun '24

*Register now! AIM workshop on June 13 - 14* On June 13 – 14 the *3^rd AIM cluster workshop* is taking place at Utrecht Universityin the Martinus Ruppert Building <https://www.uu.nl/organisatie/praktische-zaken/facilitair/zaalhuur/zalen-ui…>. Whether you're a researcher, student, or industry professional, this workshop is a great opportunity to learn about the latest developments in AI and mathematics, and connect with peers and experts in the field. In the programme attached, you can find the *list of speakers* and sessions. You can register via the website https://aimath.nl/index.php/2024/03/29/aim-2024-uu/ Registrations are possible until *Friday, June 7*. We hope to see a lot of you in Utrecht! The AIM Board: Christoph Brune Leo van Iersel Tim van Erven Tristan van Leeuwen Olivia Muthsam -- Tim van Erven<tim(a)timvanerven.nl> www.timvanerven.nl

1 0

[2nd CfP] Multi-Objective Decision Making Workshop (MODeM 2024) at ECAI 2024
by Radulescu, R.T. (Roxana) 27 May '24

27 May '24

2nd CfP: Multi-Objective Decision Making Workshop (MODeM 2024) at ECAI 2024 ------------------------------------------------------------------------------------------------------------- TLDR * Following on from successful previous editions that were held virtually in 2021<https://modem2021.cs.universityofgalway.ie/> and in person at ECAI in 2023<https://modem2023.vub.ac.be/>, the Multi-Objective Decision Making 2024 (MODeM 2024) workshop will be held in person again this year at ECAI 2024, on 19/20 October in Santiago de Compostela, Spain. * MODeM 2024 covers all aspects of multi-objective/multi-criteria/multi-attribute decision making in relation to intelligent systems and autonomous agents * After the workshop, there will be a topical collection on Multi-Objective Decision Making 2024 in the journal Neural Computing & Applications<https://link.springer.com/journal/521> (NC&A, impact factor 6.0, one of the top 10 journals in AI on Google Scholar Metrics): https://link.springer.com/journal/521/updates/26958874 * Further details are available on the workshop website: https://modem2024.vub.ac.be/ * MODeM 2024 Workshop paper deadline: 15 June 2024 (23:59 AoE) ------------------------------------------------------------------------------------------------------------- Explicitly modelling multiple objectives can be essential for decision support, explainability, and human-AI alignment. As such, there has been a growing awareness of the need for automated and assistive decision making systems to move beyond single-objective formulations when dealing with complex real-world issues, which invariably involve multiple competing objectives. In this workshop, we aim to provide a platform for original research in multi-objective decision making using AI. The workshop targets high-quality original papers covering all aspects of multi-objective decision making, including, but not limited to, the list of topics below. Topics The following is a non-exhaustive list of topics that we would like to cover in the workshop: * Multi-objective/multi-criteria/multi-attribute decision making * Multi-objective reinforcement learning * Multi-objective planning and scheduling * Multi-objective multi-agent decision making * Multi-objective game theory * Multi-objective/multi-criteria/multi-attribute utility theory * Preference elicitation for MODeM * Social choice and MODeM * Multi-objective decision support systems * Multi-objective metaheuristic optimisation (e.g. evolutionary algorithms) for autonomous agents and multi-agent systems * Multi-objectivisation * Human-AI alignment through multi-objective modelling * Ethical AI through multi-objective modelling * Explainable AI through multi-objective modelling * Interactive systems for MODeM * Applications of MODeM * Interdisciplinary work (MODeM research that relates to other fields) * New benchmark problems for MODeM Important Dates Submissions deadline: 15 June 2024 (23:59 AoE) Notification of acceptance: 06 August 2024 Camera-ready copies: 13 August 2024 Workshop: 19 or 20 October 2024 (TBA) Submission details Papers should be formatted according to the ECAI 2024 guidelines<https://www.ecai2024.eu/calls/main-track>, and should be a maximum of 7 pages in length (with additional pages containing references only). Shorter papers (up to 5 pages) reporting preliminary results and work-in-progress are also welcome. Submissions should report original work that has not previously been published elsewhere. Finally, we also welcome summaries of recently published journal papers in the form of a 2-page abstract. All submissions will be peer-reviewed (double-blind). Accepted work will be allocated time for oral and/or poster presentation during the workshop. Papers can be submitted through Microsoft CMT: https://cmt3.research.microsoft.com/MODeM2024 Topical Collection on Multi-Objective Decision Making 2024 After the workshop, all original contributions presented at MODeM 2024 will be invited to submit substantially improved and extended versions of their work, for consideration to be published in a post-proceedings topical collection (TC) of the Springer journal Neural Computing & Applications<https://www.springer.com/journal/521> (NC&A, impact factor 6.0, one of the top 10 journals in AI on Google Scholar Metrics): https://link.springer.com/journal/521/updates/26958874 Organising Committee Patrick Mannion (University of Galway, IE) Roxana Rădulescu (Utrecht University, NL; Vrije Universiteit Brussel, BE) Willem Röpke (Vrije Universiteit Brussel, BE) Pieter Libin (Vrije Universiteit Brussel, BE) Senior Advisory Committee Ali E. Abbas (University of Southern California, USA) Carlos A. Coello Coello (CINVESTAV-IPN, MX) Richard Dazeley (Deakin University, AU) Enda Howley (University of Galway, IE) Ann Nowé (Vrije Universiteit Brussel, BE) Patrice Perny (UPMC, FR) Marcello Restelli (Politecnico di Milano, IT) Diederik M. Roijers (Vrije Universiteit Brussel, BE; City of Amsterdam, NL) Peter Vamplew (Federation University Australia, AU) Nic Wilson (University College Cork, IE) Program Committee If you are interested in serving on the programme committee please get in touch with the organisers at: modem.organisers AT gmail.com

1 0

2024

2023

2022

2021

2020

Machine-learning-nederland