Javier Vazquez

Today we presented three papers on #quantum #reinforcement learning at #ICAART 2025 in the special #QAIO workshop. The papers were presented by Theodora-Augustina Drăgan and Rodrigo Coelho In our paper “Quantum Multi-Agent Reinforcement Learning for Aerial Ad-hoc Networks”, https://lnkd.in/ehvX85rK, together with Akshat Tandon from Airbus, we proposed a quantum-enhanced multi-agent reinforcement learning algorithm to dynamically optimize the connectivity between airplanes. We could show that the quantum-enhanced solution reaches as good results as a classical baseline while requiring less probes to the environment. In another work, “Continuous Quantum Reinforcement Learning for Robot Navigation”, we have used quantum reinforcement learning to navigate robots. In comparison to previous work available in the community, we expanded the model to being able to deal with local signals received from LiDAR sensors instead of relaying on global coordinates. The paper will appear on arXiv shortly. In general, it is quite unclear still which QRL model is really promising and where QRL can show an advantage. The third work, “Benchmarking Quantum Reinforcement Learning”, https://lnkd.in/eHrcfnTv, with the first author Georg Kruse thus proposes a benchmarking procedure on the gridworld environments. Some of this work has been done in the context of the Munich Quantum Valley. Fraunhofer IKS Fraunhofer IISB

51

📢 The June issue of ACM Transactions on Evolutionary Learning and Optimization is out with 2 regular research papers and Part II of the Special Issue on Explainable #AI in Evolutionary Computation: https://lnkd.in/dZwRWsHm The two regular papers are: ⭐Multi-Objectivising Acquisition Functions in Bayesian Optimisation. Chao Jiang, Miqing Li. https://lnkd.in/d7QZ7uuq Bayesian optimisation (BO) is an efficient approach for solving expensive optimisation problems, where acquisition functions play a major role in achieving the tradeoff between exploitation and exploration. The exploitation–exploration tradeoff is ... ⭐First Steps Toward a Runtime Analysis When Starting With a Good Solution. Denis Antipov, Maxim Buzdalov, Benjamin Doerr. https://lnkd.in/dKzY5NcX The mathematical runtime analysis of evolutionary algorithms traditionally regards the time an algorithm needs to find a solution of a certain quality when initialized with a random population. In practical applications it may be possible to guess ... The SI papers are: ⭐Introduction to the Special Issue on Explainable AI in Evolutionary Computation—Part 2. Jaume Bacardit, Sandy Brownlee, Stefano Cagnoni, Giovanni Iacca, John McCall, David Walker. https://lnkd.in/dC34Gtmw ⭐Explainable Optimisation through Online and Offline Hyper-heuristics. Ed Keedwell, William B. Yates, Ahmed Kheiri. https://lnkd.in/d4gNkBN2 Research in the explainability of optimisation techniques has largely focused on metaheuristics and their movement of solutions around the search landscape. Hyper-heuristics create a different challenge for explainability as they make use of many more ... ⭐EVOTER: Evolution of Transparent Explainable Rule-sets. Hormoz Shahrzad, Babak Hodjat, Risto Miikkulainen. https://lnkd.in/dmKrfN7R Most AI systems are black boxes generating reasonable outputs for given inputs. Some domains, however, have explainability and trustworthiness requirements that cannot be directly met by these approaches. Various methods have therefore been developed to ... ⭐Towards Explainable Metaheuristics: Feature Mining of Search Trajectories through Principal Component Projection. Martin Fyvie, John McCall, Lee A. Christie. https://lnkd.in/d7X9JWpF While population-based metaheuristics have proven useful for refining and improving explainable AI systems, they are seldom the focus of explanatory approaches themselves. This stems from their inherently stochastic, population-driven searches, which ... ⭐Explainable Benchmarking for Iterative Optimization Heuristics. Niki van Stein, Diederick Vermetten, Anna Kononova, Thomas Bäck. https://lnkd.in/distr2Fk Benchmarking heuristic algorithms is vital to understand under which conditions and on what kind of problems certain algorithms perform well. In most current research into heuristic optimization algorithms, only a very limited number of scenarios, ... #telo #optimization #ai #explainable

54

4 comentarios

📢 New Open Access Article in Mathematical and Computer Modelling of Dynamical Systems! We are pleased to announce the publication of a new article: 📄 Spivey’s type recurrence relation for Lah-Bell polynomials 🔗 Read the full text here: https://lnkd.in/dcUcDNUu As Editor-in-Chief, I am delighted to share this contribution with our research community. We continue to welcome high-quality work on mathematical modelling, computational methods, and their diverse applications. Taylor & Francis Group Bailey Prescott Paul Naish

38

Excited to share our recent paper explaining the importance of error mitigation (EM) for quantum computing (now available on arXiv: https://lnkd.in/dxmGUTsq ). Our main message: error mitigation will continue to play a critical in quantum computing for the foreseeable future. Undoubtedly, large-scale quantum computing will ultimately require error correction. Our paper explains why error mitigation will continue to provide very significant boosts in executable circuit volumes - 100x or more for obtaining high-accuracy results - over what's possible only with error correction.    Recently, several papers criticized EM claiming that asymptotically, it cannot yield exponential quantum advantage. While these claims are probably correct, they do not exclude dramatic finite quantum advantages that can be reached with EM. These finite quantum advantages are of key practical importance: they correspond to concrete circuit executions for which QPUs compute much faster than a classical computer, for a given problem size.   Already today, error mitigation significantly extends the reach of existing quantum hardware. Demonstrations of quantum advantage using error mitigation are around the corner and are likely to occur well before they can be achieved using error correction.    At Qedma we believe that EM is a key enabler for the future of quantum computing, and we are devoted to building EM tools for boosting quantum computing across a variety of applications. Stay tuned!   #QuantumComputing #ErrorMitigation #QuantumAdvantage #Qedma 

102

2 comentarios