26838 research outputs found
Sort by
Neural ODE and SDE Models for Adaptation and Planning in Model-Based Reinforcement Learning
Get PDFWe investigate neural ordinary and stochastic differential equations (neural ODEs and SDEs) to model stochastic dynamics in fully and partially observed environments within a modelbased reinforcement learning (RL) framework. Through a sequence of simulations, we show that neural SDEs more effectively capture transition dynamics’ inherent stochasticity, enabling high-performing policies with improved sample efficiency in challenging scenarios. We leverage neural ODEs and SDEs for efficient policy adaptation to changes in environment dynamics via inverse models, requiring only limited interactions with the new environment. To address partial observability, we introduce a latent SDE model that combines an ODE and a GAN-trained stochastic component in latent space. Policies derived from this model offer a strong baseline, outperforming or matching general model-based and model-free approaches across stochastic continuous-control benchmarks. This work illustrates the applicability of action-conditional latent SDEs for RL planning in environments with stochastic transitions. Our code is available at: https://github.com/ChaoHan-UoS/NeuralRL
Online matching on 3-uniform hypergraphs
No full textThe online matching problem was introduced by Karp, Vazirani and Vazirani (STOC 1990) on bipartite graphs with vertex arrivals. It is well-known that the optimal competitive ratio is for both integral and fractional versions of the problem. Since then, there has been considerable effort to find optimal competitive ratios for other related settings.
In this work, we go beyond the graph case and study the online matching problem on -uniform hypergraphs. For , we provide an optimal primal-dual fractional algorithm, which achieves a competitive ratio of . As our main technical contribution, we present a carefully constructed adversarial instance, which shows that this ratio is in fact optimal. It combines ideas from known hard instances for bipartite graphs under the edge-arrival and vertex-arrival models.
For , we give a simple integral algorithm which performs better than greedy when the online nodes have bounded degree. As a corollary, it achieves the optimal competitive ratio of 1/2 on 3-uniform hypergraphs when every online node has degree at most 2. This is because the special case where every online node has degree 1 is equivalent to the edge-arrival model on graphs, for which an upper bound of 1/2 is known
The future of human-robot synergy in interactive environments: The role of robots at the workplace
Get PDFThe increasing integration of robots into workplaces raises critical questions about human-robot synergy in interactive environments. While robots are designed to enhance productivity and safety, their successful deployment depends on effective collaboration, trust, and seamless interaction with human workers. However, existing research has primarily focused on either technical capabilities or human-centered concerns in isolation, leaving a gap in understanding how robots can be meaningfully integrated into dynamic workspaces. In this workshop, we bring together experts from robotics, HCI, and work sciences to explore the future of human-robot collaboration at the workplace. This workshop aims to identify key design principles, ethical considerations, and practical challenges. The insights gained will inform future research and policy recommendations, shaping a future in which robots act not as mere tools but as cooperative agents that enhance workplace efficiency, well-being, and innovation
Optimizing clinical pathways with federated data
No full textClinical pathways are currently difficult to optimize due to sensitive data that is typically distributed across organizations while rules and regulations constrain access and data processing. In this paper we describe a federated approach that can significantly reduce the efforts required to overcome these obstacles. First, we describe a standard conceptual workflow for optimizing clinical pathways, including all steps and involved stakeholders. This is followed by a translation of the workflow into a real-world scenario with an associated proof of principle to demonstrate how the scenario can be implemented on top of a federated framework. We present the most important results and conclude with an overview of the benefits for each of the stakeholders. Our most important outcomes are: the federated approach offers significant benefits for all relevant stakeholders and has little downsides. A policy-driven framework with embedded policy enforcement is crucial for successful adoption of a federated approach. Integration of safe statistics and synthetic data generation in a federated framework is straightforward and offers additional benefits, especially when setting up healthcare consortia. This solution is almost ready to be adopted by healthcare organizations as part of their regular operations
Nationale supercomputer kwam er mede door Walter - Noordhollands Dagblad ed. Dagblad Waterland - 19-06-2025
No full text
Modular ixml
Get PDFMost current ixml grammars are small. However there are examples of large grammars, and it is likely that in the future more large grammars will emerge as ixml usage increases. To make large grammars more manageable, and to enable reuse, it would be useful to have a way to modularise them. One of the requirements of modularisation for reuse in any notation is to have a method of specifying the contractual interface, such that it is possible for the producers of the modules to change their internal structure without breaking any existing usage of the module.
This paper describes a proposal for an ixml preprocessor that permits an ixml grammar to invoke other modules of ixml grammars, specifying their linkage. This involves the renaming of rules with name clashes in the modules, using ixml renaming, resulting in a single ixml grammar with no rule-name clashes, and so that the resultant XML serialisations remain the same. The invoking grammar remains unchanged. There is no change to the syntax or semantics of ixml proper