26838 research outputs found
Sort by
MOMAland: Benchmarking multi-objective multi-agent reinforcement learning
No full textMany challenging tasks such as managing traffic systems, electricity grids, or supply chains involve complex decision-making processes that must balance multiple conflicting objectives and coordinate the actions of various independent decision-makers (DMs). One perspective for formalising and addressing such tasks is multi-objective multi-agent reinforcement learning (MOMARL). MOMARL broadens reinforcement learning (RL) to problems with multiple agents each needing to consider multiple objectives in their learning process. In reinforcement learning research, benchmarks are crucial in facilitating progress, evaluation, and reproducibility. The significance of benchmarks is underscored by the existence of numerous benchmark frameworks developed for various RL paradigms, including single-agent RL (e.g., Gymnasium), multi-agent RL (e.g., PettingZoo), and single-agent multi-objective RL (e.g., MO-Gymnasium). To support the advancement of the MOMARL field, we introduce MOMAland, the first collection of standardised environments for multi-objective multi-agent reinforcement learning. MOMAland addresses the need for comprehensive benchmarking in this emerging field, offering over 10 diverse environments that vary in the number of agents, state representations, reward structures, and utility considerations. To provide strong baselines for future research, MOMAland also includes algorithms capable of learning policies in such settings
The BlackGEM telescope array. I. Overview
Get PDFThe main science aim of the BlackGEM array is to detect optical counterparts to gravitational wave mergers. Additionally, the array will perform a set of synoptic surveys to detect Local Universe transients and short timescale variability in stars and binaries, as well as a six-filter all-sky survey down to ∼22nd mag. The BlackGEM Phase-I array consists of three optical wide-field unit telescopes. Each unit uses an f/5.5 modified Dall-Kirkham (Harmer-Wynne) design with a triplet corrector lens, and a 65 cm primary mirror, coupled with a 110Mpix CCD detector, that provides an instantaneous field-of-view of 2.7 square degrees, sampled at 0.″564 pixel−1. The total field-of-view for the array is 8.2 square degrees. Each telescope is equipped with a six-slot filter wheel containing an optimised Sloan set (BG-u, BG-g, BG-r, BG-i, BG-z) and a wider-band 440-720 nm (BG-q) filter. Each unit telescope is independent from the others. Cloud-based data processing is done in real time, and includes a transient-detection routine as well as a full-source optimal-photometry module. BlackGEM has been installed at the ESO La Silla observatory as of 2019 October. After a prolonged COVID-19 hiatus, science operations started on 2023 April 1 and will run for five years. Aside from its core scientific program, BlackGEM will give rise to a multitude of additional science cases in multi-colour time-domain astronomy, to the benefit of a variety of topics in astrophysics, such as infant supernovae, luminous red novae, asteroseismology of post-main-sequence objects, (ultracompact) binary stars, and the relation between gravitational wave counterparts and other classes of transients
First draft of P.IXC "Interactive test methods for subjective assessment of extended reality communications"
No full textRound-tripping invisible XML
Get PDFInvisible XML takes textual documents where the structure is implicit and produces documents with the structure made explicit. This paper addresses the question of the extent to which it is possible to recreate the original textual document from its structured version, how it can be done, and what if any the ramifications are for ixml
Open-Sourcing VR2Gather: A collaborative social VR system for adaptive multi-party real time communication
Get PDFSocial Virtual Reality is envisioned to transform how individu-
als communicate remotely, offering a sense of immersion and co-
presence within a virtual space. Current platforms enabling remote
social interactions rely on synthetic user representations. We ad-
dress this limitation by enabling realistic human representation
through volumetric content capture, encoding and transmission.
Specifically, we present an extended version of VR2Gather, now a
fully open source Unity package, available at https://github.com/
cwi-dis/VR2Gather-acmmm-oss. Our platform is a customisable
system to transmit volumetric content in a multi-party real-time
environment, easy to integrate into existing applications
Balanced resonate-and-fire neurons
Get PDFThe resonate-and-fire (RF) neuron, introduced over two decades ago, is a simple, efficient, yet biologically plausible spiking neuron model, which can extract frequency patterns within the time domain due to its resonating membrane dynamics. However, previous RF formulations suffer from intrinsic shortcomings that limit effective learning and prevent exploiting the principled advantage of RF neurons. Here, we introduce the balanced RF (BRF) neuron, which alleviates some of the intrinsic limitations of vanilla RF neurons and demonstrates its effectiveness within recurrent spiking neural networks (RSNNs) on various sequence learning tasks. We show that networks of BRF neurons achieve overall higher task performance, produce only a fraction of the spikes, and require significantly fewer parameters as compared to modern RSNNs. Moreover, BRF-RSNN consistently provide much faster and more stable training convergence, even when bridging many hundreds of time steps during backpropagation through time (BPTT). These results underscore that our BRF-RSNN is a strong candidate for future large-scale RSNN architectures, further lines of research in SNN methodology, and more efficient hardware implementations
Learning discretized Bayesian networks with GOMEA
Get PDFBayesian networks model relationships between random variables under uncertainty and can be used to predict the likelihood of events and outcomes while incorporating observed evidence. From an eXplainable AI (XAI) perspective, such models are interesting as they tend to be compact. Moreover, captured relations can be directly inspected by domain experts. In practice, data is often real-valued. Unless assumptions of normality can be made, discretization is often required. The optimal discretization, however, depends on the relations modelled between the variables. This complicates learning Bayesian networks from data. For this reason, most literature focuses on learning conditional dependencies between sets of variables, called structure learning. In this work, we extend an existing state-of-the-art structure learning approach based on the Gene-pool Optimal Mixing Evolutionary Algorithm (GOMEA) to jointly learn variable discretizations. The proposed Discretizing Bayesian Network GOMEA (DBN-GOMEA) obtains similar or better results than the current state-of-the-art when tasked to retrieve randomly generated ground-truth networks. Moreover, leveraging a key strength of evolutionary algorithms, we can straightforwardly perform DBN learning multi-objectively. We show how this enables incorporating expert knowledge in a uniquely insightful fashion, finding multiple DBNs that trade-off complexity, accuracy, and the difference with a pre-determined expert network
RADIUS/UDP considered harmful
Get PDFThe RADIUS protocol is the de facto standard lightweight protocol for authentication, authorization, and accounting (AAA) for networked devices. It is used to support remote access for diverse use cases including network routers, industrial control systems, VPNs, enterprise Wi-Fi including the Eduroam network, Linux Pluggable Authentication Modules, and mobile roaming and Wi-Fi offload. We have discovered a protocol vulnerability in RADIUS that has been present for decades. Our attack allows a man-in-the-middle attacker to authenticate itself to a device using RADIUS for user authentication, or to assign itself arbitrary network privileges. Our attack exploits an MD5 chosen-prefix collision on the ad hoc RADIUS packet authentication construction to produce Access-Accept and Access-Reject packets with identical Response Authenticators, allowing our attacker to transform a reject into an accept without knowledge of the shared secret between RADIUS client and server. We optimize the MD5 chosen-prefix attack to produce collisions online in less than five minutes, and show how to fit the collision blocks within RADIUS attributes that will be echoed back from the server. We demonstrate our attack in a variety of settings against popular RADIUS implementations. It is our hope that this attack will provide the impetus for vendors and the IETF to deprecate RADIUS over UDP, and to require RADIUS to run over secure channels with modern cryptographic privacy and integrity guarantees
A Cox rate-and-state model for monitoring seismic hazard in the Groningen gas field
Get PDFTo monitor the seismic hazard in the Groningen gas field, we modify the rate-
and-state model that relates changes in pore pressure to induced seismic hazard by allowing
for noise in pore pressure measurements and by explicitly taking into account gas production
volumes. We analyse the first and second-moment structure of the resulting Cox process,
propose an unbiased estimating equation approach for the unknown model parameters and
derive the posterior distribution of the driving random measure. We use a parallel Metropolis
adjusted Langevin algorithm for sampling from the posterior and to monitor the hazard
Multidimensional quantum walks,recursion, and quantum divide & conquer
Get PDFWe introduce an object called a subspace graph that formalizes the technique of multidimen-
sional quantum walks. Composing subspace graphs allows one to seamlessly combine quantum
and classical reasoning, keeping a classical structure in mind, while abstracting quantum parts into
subgraphs with simple boundaries as needed. As an example, we show how to combine a switching
network with arbitrary quantum subroutines, to compute a composed function. As another applica-
tion, we give a time-efficient implementation of quantum Divide & Conquer when the sub-problems
are combined via a Boolean formula. We use this to quadratically speed up Savitch’s algorithm for
directed st-connectivity