Search

Loss of plasticity, trainability loss, and primacy bias have been identified as issues arising when training deep neural networks on sequences of tasks -- all referring to the increased difficulty in training on new tasks. We propose to use Parseval regularization, which maintains orthogonality of weight matrices, to preserve useful optimization properties and improve training in a continual reinforcement learning setting. We show that it provides significant benefits to RL agents on a suite of gridworld, CARL and MetaWorld tasks. We conduct comprehensive ablations to identify the source of its benefits and investigate the effect of certain metrics associated to network trainability including weight matrix rank, weight norms and policy entropy.

The spectroscopic quadrupole moment of the first excited state, $Q_{_S}(2^{+}_{1})$, at 1.634 MeV in $^{20}$Ne was determined from sensitive reorientation-effect Coulomb-excitation measurements using a heavy target and safe energies well below the Coulomb barrier. Particle-$\gamma$ coincidence measurements were collected at iThemba LABS with a digital data-acquisition system using the {\sc AFRODITE} array coupled to an annular, doubled-sided silicon detector. A precise value of $Q_{_S}(2^{+}_{1})=-0.22(2)$ eb was determined at backward angles in agreement with the only safe-energy measurement prior to this work, $Q_{_S}(2^{+}_{1})=-0.23(8)$ eb. This result adopts 1$\hbar\omega$ shell-model calculations of the nuclear dipole polarizability of the 2$^+_1$ state that contributes to the effective quadrupole interaction and determination of $Q_{_S}(2^{+}_{1})$. It disagrees, however, with the ideal rotor model for axially-symmetric nuclei by almost $3\sigma$. Larger discrepancies are computed by modern state-of-the-art calculations performed in this and prior work, including {\it ab initio} shell model with chiral effective interactions and the multi-reference relativistic energy density functional ({\sc MR-EDF}) model. The intrinsic nucleon density of the 2$^+_1$ state in $^{20}$Ne calculated with the {\sc MR-EDF} model illustrates the presence of $\alpha$ clustering, which explains the largest discrepancy with the rotor model found in the nuclear chart and motivates the explicit inclusion of $\alpha$ clustering for full convergence of $E2$ collective properties., Comment: 8 pages, 8 figures

The advent of next-generation radio interferometers like the Square Kilometer Array promises to revolutionise our radio astronomy observational capabilities. The unprecedented volume of data these devices generate requires fast and accurate image reconstruction algorithms to solve the ill-posed radio interferometric imaging problem. Most state-of-the-art reconstruction methods lack trustworthy and scalable uncertainty quantification, which is critical for the rigorous scientific interpretation of radio observations. We propose an unsupervised technique based on a conformalized version of a radio-augmented equivariant bootstrapping method, which allows us to quantify uncertainties for fast reconstruction methods. Noticeably, we rely on reconstructions from ultra-fast unrolled algorithms. The proposed method brings more reliable uncertainty estimations to our problem than existing alternatives., Comment: 14 pages, 7 figures. Accepted at the Machine Learning and the Physical Sciences Workshop, NeurIPS 2024

As machine intelligence evolves, the need to test and compare the problem-solving abilities of different AI models grows. However, current benchmarks are often simplistic, allowing models to perform uniformly well and making it difficult to distinguish their capabilities. Additionally, benchmarks typically rely on static question-answer pairs that the models might memorize or guess. To address these limitations, we introduce Dynamic Intelligence Assessment (DIA), a novel methodology for testing AI models using dynamic question templates and improved metrics across multiple disciplines such as mathematics, cryptography, cybersecurity, and computer science. The accompanying dataset, DIA-Bench, contains a diverse collection of challenge templates with mutable parameters presented in various formats, including text, PDFs, compiled binaries, visual puzzles, and CTF-style cybersecurity challenges. Our framework introduces four new metrics to assess a model's reliability and confidence across multiple attempts. These metrics revealed that even simple questions are frequently answered incorrectly when posed in varying forms, highlighting significant gaps in models' reliability. Notably, API models like GPT-4o often overestimated their mathematical capabilities, while ChatGPT-4o demonstrated better performance due to effective tool usage. In self-assessment, OpenAI's o1-mini proved to have the best judgement on what tasks it should attempt to solve. We evaluated 25 state-of-the-art LLMs using DIA-Bench, showing that current models struggle with complex tasks and often display unexpectedly low confidence, even with simpler questions. The DIA framework sets a new standard for assessing not only problem-solving but also a model's adaptive intelligence and ability to assess its limitations. The dataset is publicly available on the project's page: https://github.com/DIA-Bench.

This review paper provides a comprehensive analysis of recent advances in automatic speech recognition (ASR) with bidirectional encoder representations from transformers BERT and connectionist temporal classification (CTC) transformers. The paper first introduces the fundamental concepts of ASR and discusses the challenges associated with it. It then explains the architecture of BERT and CTC transformers and their potential applications in ASR. The paper reviews several studies that have used these models for speech recognition tasks and discusses the results obtained. Additionally, the paper highlights the limitations of these models and outlines potential areas for further research. All in all, this review provides valuable insights for researchers and practitioners who are interested in ASR with BERT and CTC transformers.

Chiral magnets have garnered significant interest due to the emergence of unique phenomena prohibited in inversion-symmetric magnets. While the equilibrium characteristics of chiral magnets have been extensively explored through the Dzyaloshinskii-Moriya interaction (DMI), non-equilibrium properties like magnetic damping have received comparatively less attention. We present the inaugural direct observation of chiral damping through Brillouin light scattering (BLS) spectroscopy. Employing BLS spectrum analysis, we independently deduce the Dzyaloshinskii-Moriya interaction (DMI) and chiral damping, extracting them from the frequency shift and linewidth of the spectrum peak, respectively. The resulting linewidths exhibit clear odd symmetry with respect to the magnon wave vector, unambiguously confirming the presence of chiral damping. Our study introduces a novel methodology for quantifying chiral damping, with potential ramifications on diverse nonequilibrium phenomena within chiral magnets.

Knowledge graphs (KGs) have become ubiquitous publicly available knowledge sources, and are nowadays covering an ever increasing array of domains. However, not all knowledge represented is useful or pertaining when considering a new application or specific task. Also, due to their increasing size, handling large KGs in their entirety entails scalability issues. These two aspects asks for efficient methods to extract subgraphs of interest from existing KGs. To this aim, we introduce KGPrune, a Web Application that, given seed entities of interest and properties to traverse, extracts their neighboring subgraphs from Wikidata. To avoid topical drift, KGPrune relies on a frugal pruning algorithm based on analogical reasoning to only keep relevant neighbors while pruning irrelevant ones. The interest of KGPrune is illustrated by two concrete applications, namely, bootstrapping an enterprise KG and extracting knowledge related to looted artworks., Comment: Accepted as a demo paper at ECAI 2024

We present significant improvements to our previous work on noise reduction in {\sl Herschel} observation maps by defining sparse filtering tools capable of handling, in a unified formalism, a significantly improved noise reduction as well as a deconvolution in order to reduce effects introduced by the limited instrumental response (beam). We implement greater flexibility by allowing a wider choice of parsimonious priors in the noise-reduction process. More precisely, we introduce a sparse filtering and deconvolution approach approach of type $l^2$-$l^p$, with $p > 0$ variable and apply it to a larger set of molecular clouds using {\sl Herschel} 250 $\mu $m data in order to demonstrate their wide range of application. In the {\sl Herschel} data, we are able to use this approach to highlight extremely fine filamentary structures and obtain singularity spectra that tend to show a significantly less $\log$-normal behavior and a filamentary nature in the less dense regions. We also use high-resolution adaptive magneto-hydrodynamic simulation data to assess the quality of deconvolution in such a simulated beaming framework., Comment: 24 pages, 32 figures, official acceptance date on 16/05/2024 (Astronomy & Astrophysics)

The rise of new complex attacks scenarios in Internet of things (IoT) environments necessitate more advanced and intelligent cyber defense techniques such as various Intrusion Detection Systems (IDSs) which are responsible for detecting and mitigating malicious activities in IoT networks without human intervention. To address this issue, deep reinforcement learning (DRL) has been proposed in recent years, to automatically tackle intrusions/attacks. In this paper, a comprehensive survey of DRL-based IDS on IoT is presented. Furthermore, in this survey, the state-of-the-art DRL-based IDS methods have been classified into five categories including wireless sensor network (WSN), deep Q-network (DQN), healthcare, hybrid, and other techniques. In addition, the most crucial performance metrics, namely accuracy, recall, precision, false negative rate (FNR), false positive rate (FPR), and F-measure, are detailed, in order to evaluate the performance of each proposed method. The paper provides a summary of datasets utilized in the studies as well.

Protecting Internet of things (IoT) devices against cyber attacks is imperative owing to inherent security vulnerabilities. These vulnerabilities can include a spectrum of sophisticated attacks that pose significant damage to both individuals and organizations. Employing robust security measures like intrusion detection systems (IDSs) is essential to solve these problems and protect IoT systems from such attacks. In this context, our proposed IDS model consists on a combination of convolutional neural network (CNN) and long short-term memory (LSTM) deep learning (DL) models. This fusion facilitates the detection and classification of IoT traffic into binary categories, benign and malicious activities by leveraging the spatial feature extraction capabilities of CNN for pattern recognition and the sequential memory retention of LSTM for discerning complex temporal dependencies in achieving enhanced accuracy and efficiency. In assessing the performance of our proposed model, the authors employed the new CICIoT2023 dataset for both training and final testing, while further validating the model's performance through a conclusive testing phase utilizing the CICIDS2017 dataset. Our proposed model achieves an accuracy rate of 98.42%, accompanied by a minimal loss of 0.0275. False positive rate(FPR) is equally important, reaching 9.17% with an F1-score of 98.57%. These results demonstrate the effectiveness of our proposed CNN-LSTM IDS model in fortifying IoT environments against potential cyber threats.

Red-teaming, or identifying prompts that elicit harmful responses, is a critical step in ensuring the safe and responsible deployment of large language models (LLMs). Developing effective protection against many modes of attack prompts requires discovering diverse attacks. Automated red-teaming typically uses reinforcement learning to fine-tune an attacker language model to generate prompts that elicit undesirable responses from a target LLM, as measured, for example, by an auxiliary toxicity classifier. We show that even with explicit regularization to favor novelty and diversity, existing approaches suffer from mode collapse or fail to generate effective attacks. As a flexible and probabilistically principled alternative, we propose to use GFlowNet fine-tuning, followed by a secondary smoothing phase, to train the attacker model to generate diverse and effective attack prompts. We find that the attacks generated by our method are effective against a wide range of target LLMs, both with and without safety tuning, and transfer well between target LLMs. Finally, we demonstrate that models safety-tuned using a dataset of red-teaming prompts generated by our method are robust to attacks from other RL-based red-teaming approaches.

This paper provides a comprehensive review of the future of cybersecurity through Generative AI and Large Language Models (LLMs). We explore LLM applications across various domains, including hardware design security, intrusion detection, software engineering, design verification, cyber threat intelligence, malware detection, and phishing detection. We present an overview of LLM evolution and its current state, focusing on advancements in models such as GPT-4, GPT-3.5, Mixtral-8x7B, BERT, Falcon2, and LLaMA. Our analysis extends to LLM vulnerabilities, such as prompt injection, insecure output handling, data poisoning, DDoS attacks, and adversarial instructions. We delve into mitigation strategies to protect these models, providing a comprehensive look at potential attack scenarios and prevention techniques. Furthermore, we evaluate the performance of 42 LLM models in cybersecurity knowledge and hardware security, highlighting their strengths and weaknesses. We thoroughly evaluate cybersecurity datasets for LLM training and testing, covering the lifecycle from data creation to usage and identifying gaps for future research. In addition, we review new strategies for leveraging LLMs, including techniques like Half-Quadratic Quantization (HQQ), Reinforcement Learning with Human Feedback (RLHF), Direct Preference Optimization (DPO), Quantized Low-Rank Adapters (QLoRA), and Retrieval-Augmented Generation (RAG). These insights aim to enhance real-time cybersecurity defenses and improve the sophistication of LLM applications in threat detection and response. Our paper provides a foundational understanding and strategic direction for integrating LLMs into future cybersecurity frameworks, emphasizing innovation and robust model deployment to safeguard against evolving cyber threats., Comment: 50 pages, 8 figures

In this paper, we introduce a large class of (so-called) conditional indicators, on a complete probability space with respect to a sub $\sigma$-algebra. A conditional indicator is a positive mapping, which is not necessary linear, but may share common features with the conditional expectation, such as the tower property or the projection property. Several characterizations are formulated. Beyond the definitions, we provide some non trivial examples that are used in finance and may inspire new developments in the theory of operators on Riesz spaces.

In this paper, a general framework is developed for continuous-time financial market models defined from simple strategies through conditional topologies that avoid stochastic calculus and do not necessitate semimartingale models. We then compare the usual no-arbitrage conditions of the literature, e.g. the usual no-arbitrage conditions NFL, NFLVR and NUPBR and the recent AIP condition. With appropriate pseudo-distance topologies, we show that they hold in continuous time if and only if they hold in discrete time. Moreover, the super-hedging prices in continuous time coincide with the discrete-time super-hedging prices, even without any no-arbitrage condition.

Despite its amazing quantitative successes and contributions to revolutionary technologies, physics currently faces many unsolved mysteries ranging from the meaning of quantum mechanics to the nature of the dark energy that will determine the future of the Universe. It is clearly prohibitive for the general reader, and even the best informed physicists, to follow the vast number of technical papers published in the thousands of specialized journals. For this reason, we have asked the leading experts across many of the most important areas of physics to summarise their global assessment of some of the most important issues. In lieu of an extremely long abstract summarising the contents, we invite the reader to look at the section headings and their authors, and then to indulge in a feast of stimulating topics spanning the current frontiers of fundamental physics from The Future of Physics by William D Phillips and What characterises topological effects in physics? by Gerard t Hooft through the contributions of the widest imaginable range of world leaders in their respective areas. This paper is presented as a preface to exciting developments by senior and young scientists in the years that lie ahead, and a complement to the less authoritative popular accounts by journalists., Comment: 54 pages, 13 figures

As more devices connect to the internet, it becomes crucial to address their limitations and basic security needs. While much research focuses on utilizing ML and DL to tackle security challenges, there is often a tendency to overlook the practicality and feasibility of implementing these methods in real-time settings. This oversight stems from the constrained processing power and memory of certain devices (IoT devices), as well as concerns about the generalizability of these approaches. Focusing on the detection of DNS-tunneling attacks in a router as a case study, we present an end-to-end process designed to effectively address these challenges. The process spans from developing a lightweight DNS-tunneling detection model to integrating it into a resource-constrained device for real-time detection. Through our experiments, we demonstrate that utilizing stateless features for training the ML model, along with features chosen to be independent of the network configuration, leads to highly accurate results. The deployment of this carefully crafted model, optimized for embedded devices across diverse environments, resulted in high DNS-tunneling attack detection with minimal latency. With this work, we aim to encourage solutions that strike a balance between theoretical advancements and the practical applicability of ML approaches in the ever-evolving landscape of device security.

This work investigates the role of the emerging Analog In-memory computing (AIMC) paradigm in enabling Medical AI analysis and improving the certainty of these models at the edge. It contrasts AIMC's efficiency with traditional digital computing's limitations in power, speed, and scalability. Our comprehensive evaluation focuses on brain tumor analysis, spleen segmentation, and nuclei detection. The study highlights the superior robustness of isotropic architectures, which exhibit a minimal accuracy drop (0.04) in analog-aware training, compared to significant drops (up to 0.15) in pyramidal structures. Additionally, the paper emphasizes IMC's effective data pipelining, reducing latency and increasing throughput as well as the exploitation of inherent noise within AIMC, strategically harnessed to augment model certainty.