Search

This study reports the synthesis and characterization of nanocrystalline Ni0.5Co0.2Zn0.3BixFe2-xO4 x varis by 0.0, 0.025, 0.050, 0.075, 0.100 ferrites synthesized via the sol-gel auto combustion method.The low coercivity values 23.68 to 87.71 Oe are observed,classifying the investigated materials as soft ferromagnetic.The increased magnetic anisotropy K through Bi3+ doping indicates tunable stability in magnetic orientations,making them suitable for multifunctional applications.

Nowadays, cybersecurity has grown into a more significant and difficult scientific issue. The recog-nition of threats and attacks meant for knowledge and safety on the internet is growing harder to detect. Since cybersecurity guarantees the privacy and security of data sent via the Internet, it is essential, while also providing protection against malicious attacks. Encrypt has grown into an an-swer that has become an essential element of information security systems. To ensure the security of shared data, including text, images, or videos, it is essential to employ various methods and strategies. This study delves into the prevalent cryptographic methods and algorithms utilized for prevention and stream encryption, examining their encoding techniques such as advanced encryp-tion standard (AES), Blowfish, Twofish, Salsa20, and ChaCha20. The primary objective of this re-search is to identify the optimal times and throughputs (speeds) for data encryption and decryption processes. The methodology of this study involved selecting five distinct types of images to com-pare the outcomes of the techniques evaluated in this research. The assessment focused on pro-cessing time and speed parameters, examining visual encoding and decoding using Java as the pri-mary platform. A comparative analysis of several symmetric key ciphers was performed, focusing on handling large datasets. Despite this limitation, comparing different images helped evaluate the techniques' novelty. The results showed that ChaCha20 had the best average time for both encryp-tion and decryption, being over 50% faster than some other algorithms. However, the Twofish algo-rithm had lower throughput during testing. The paper concludes with findings and suggestions for future improvements.

Fine-tuning large language models on private data for downstream applications poses significant privacy risks in potentially exposing sensitive information. Several popular community platforms now offer convenient distribution of a large variety of pre-trained models, allowing anyone to publish without rigorous verification. This scenario creates a privacy threat, as pre-trained models can be intentionally crafted to compromise the privacy of fine-tuning datasets. In this study, we introduce a novel poisoning technique that uses model-unlearning as an attack tool. This approach manipulates a pre-trained language model to increase the leakage of private data during the fine-tuning process. Our method enhances both membership inference and data extraction attacks while preserving model utility. Experimental results across different models, datasets, and fine-tuning setups demonstrate that our attacks significantly surpass baseline performance. This work serves as a cautionary note for users who download pre-trained models from unverified sources, highlighting the potential risks involved.

In this paper, a compact microstrip forward broadside coupler (MFBC) with high coupling level is proposed in the frequency band of 3.5-3.8 GHz. The coupler is composed of two parallel pixelated transmission lines. To validate the designstrategy, the proposed MFBC is fabricated and measured. The measured results demonstrate a forward coupler with 3 dB coupling, and a compact size of 0.12 {\lambda}g x 0.10{\lambda}g. Binary Particle Swarm Optimization (BPSO) design methodology and flexibility of pixelation enable us to optimize the proposed MFBC with desired coupling level and operating frequency within a fixed dimension. Also, low sensitivity to misalignment between two coupled TLs makes the proposed coupler a good candidate for near-field Wireless Power Transfer (WPT) application and sensors.

Robotic Process Automation (RPA) has emerged as a game-changing technology in data extraction, revolutionizing the way organizations process and analyze large volumes of documents such as invoices, purchase orders, and payment advices. This study investigates the use of RPA for structured data extraction and evaluates its advantages over manual processes. By comparing human-performed tasks with those executed by RPA software bots, we assess efficiency and accuracy in data extraction from invoices, focusing on the effectiveness of the RPA system. Through four distinct scenarios involving varying numbers of invoices, we measure efficiency in terms of time and effort required for task completion, as well as accuracy by comparing error rates between manual and RPA processes. Our findings highlight the significant efficiency gains achieved by RPA, with bots completing tasks in significantly less time compared to manual efforts across all cases. Moreover, the RPA system consistently achieves perfect accuracy, mitigating the risk of errors and enhancing process reliability. These results underscore the transformative potential of RPA in optimizing operational efficiency, reducing human labor costs, and improving overall business performance., Comment: This manuscript has been accepted for publication in the journal Revue d'Intelligence Artificielle

In this work, we studied X-ray source SRGe J021932.4$-$040154 (SRGe J02193), which we associated with a single X-ray active star of spectral class G2V$-$G4V and the rotational period $\rm P_{rot} < 9.3$ days. Additional analysis of TESS light-curves allowed for the rotational period estimation of $3.2 \pm 0.5$ days. SRGe J02193 was observed with the SRG/eROSITA during eUDS survey in 2019 in a much dimmer state compared to the XMM-Newton catalogue 4XMM-DR12. Detailed analysis revealed that the archival XMM-Newton observations captured the source during a flaring event in 2017. The XMM-Newton light curve demonstrates a strong flare described with the Gaussian rise and exponential decay, typical for stellar flares, characterized by timescales of ${\sim}400$ s and ${\sim}1300$ s, respectively. The spectral analysis of the quiescent state reveals ${\sim}10$ MK plasma at luminosity of $(1.4\pm0.4) \times 10^{29} \rm~erg~s^{-1}$ (0.3$-$4.5 keV). The spectrum of the flare is characterized by temperature of ${\sim}40$ MK and luminosity $(5.5\pm0.6)\times 10^{30} \rm~erg~s^{-1}$. The total energy emitted during the flare ${\sim}1.7 \times 10^{34}$ erg exceeds the canonical threshold of $10^{33}$ erg, allowing us to classify the observed event as a superflare on a Sun-like star. Additionally, we present the upper limit on the surface starspot area based on the brightness variations and consider the hypothesis of the object being a binary system with G-type and M-type stars, suggested by two independent estimations of radial velocity variations from APOGEE-2 and Gaia.

Recent advancements in text-to-speech (TTS) models have aimed to streamline the two-stage process into a single-stage training approach. However, many single-stage models still lag behind in audio quality, particularly when handling Kurdish text and speech. There is a critical need to enhance text-to-speech conversion for the Kurdish language, particularly for the Sorani dialect, which has been relatively neglected and is underrepresented in recent text-to-speech advancements. This study introduces an end-to-end TTS model for efficiently generating high-quality Kurdish audio. The proposed method leverages a variational autoencoder (VAE) that is pre-trained for audio waveform reconstruction and is augmented by adversarial training. This involves aligning the prior distribution established by the pre-trained encoder with the posterior distribution of the text encoder within latent variables. Additionally, a stochastic duration predictor is incorporated to imbue synthesized Kurdish speech with diverse rhythms. By aligning latent distributions and integrating the stochastic duration predictor, the proposed method facilitates the real-time generation of natural Kurdish speech audio, offering flexibility in pitches and rhythms. Empirical evaluation via the mean opinion score (MOS) on a custom dataset confirms the superior performance of our approach (MOS of 3.94) compared with that of a one-stage system and other two-staged systems as assessed through a subjective human evaluation., Comment: 19 pages

Micro-objects levitated in a vacuum are an exciting platform for precision sensing due to their low dissipation motion and the potential for control at the quantum level. Arrays of such sensors would allow noise cancellation, directionality, increased sensitivity and in the quantum regime the potential to exploit correlation and entanglement. We use neuromorphic detection via a single event-based camera to record the motion of an array of levitated microspheres. We present the first truly scalable method for multiparticle control by implementing real-time feedback to cool the motion of three objects simultaneously.

Machine learning (ML) is a rapidly developing area of medicine that uses significant resources to apply computer science and statistics to medical issues. ML's proponents laud its capacity to handle vast, complicated, and erratic medical data. It's common knowledge that attackers might cause misclassification by deliberately creating inputs for machine learning classifiers. Research on adversarial examples has been extensively conducted in the field of computer vision applications. Healthcare systems are thought to be highly difficult because of the security and life-or-death considerations they include, and performance accuracy is very important. Recent arguments have suggested that adversarial attacks could be made against medical image analysis (MedIA) technologies because of the accompanying technology infrastructure and powerful financial incentives. Since the diagnosis will be the basis for important decisions, it is essential to assess how strong medical DNN tasks are against adversarial attacks. Simple adversarial attacks have been taken into account in several earlier studies. However, DNNs are susceptible to more risky and realistic attacks. The present paper covers recent proposed adversarial attack strategies against DNNs for medical imaging as well as countermeasures. In this study, we review current techniques for adversarial imaging attacks, detections. It also encompasses various facets of these techniques and offers suggestions for the robustness of neural networks to be improved in the future.

Multimodal machine learning models that combine visual and textual data are increasingly being deployed in critical applications, raising significant safety and security concerns due to their vulnerability to adversarial attacks. This paper presents an effective strategy to enhance the robustness of multimodal image captioning models against such attacks. By leveraging the Fast Gradient Sign Method (FGSM) to generate adversarial examples and incorporating adversarial training techniques, we demonstrate improved model robustness on two benchmark datasets: Flickr8k and COCO. Our findings indicate that selectively training only the text decoder of the multimodal architecture shows performance comparable to full adversarial training while offering increased computational efficiency. This targeted approach suggests a balance between robustness and training costs, facilitating the ethical deployment of multimodal AI systems across various domains., Comment: Accepted into KDD 2024 workshop on Ethical AI

We present foundation language models developed to power Apple Intelligence features, including a ~3 billion parameter model designed to run efficiently on devices and a large server-based language model designed for Private Cloud Compute. These models are designed to perform a wide range of tasks efficiently, accurately, and responsibly. This report describes the model architecture, the data used to train the model, the training process, how the models are optimized for inference, and the evaluation results. We highlight our focus on Responsible AI and how the principles are applied throughout the model development.

Generative models based on variational autoencoders are a popular technique for detecting anomalies in images in a semi-supervised context. A common approach employs the anomaly score to detect the presence of anomalies, and it is known to reach high level of accuracy on benchmark datasets. However, since anomaly scores are computed from reconstruction disparities, they often obscure the detection of various spurious features, raising concerns regarding their actual efficacy. This case study explores the robustness of an anomaly detection system based on variational autoencoder generative models through the use of eXplainable AI methods. The goal is to get a different perspective on the real performances of anomaly detectors that use reconstruction differences. In our case study we discovered that, in many cases, samples are detected as anomalous for the wrong or misleading factors., Comment: World Conference on eXplainable Artificial Intelligence

Data clustering involves identifying latent similarities within a dataset and organizing them into clusters or groups. The outcomes of various clustering algorithms differ as they are susceptible to the intrinsic characteristics of the original dataset, including noise and dimensionality. The effectiveness of such clustering procedures directly impacts the homogeneity of clusters, underscoring the significance of evaluating algorithmic outcomes. Consequently, the assessment of clustering quality presents a significant and complex endeavor. A pivotal aspect affecting clustering validation is the cluster validity metric, which aids in determining the optimal number of clusters. The main goal of this study is to comprehensively review and explain the mathematical operation of internal and external cluster validity indices, but not all, to categorize these indices and to brainstorm suggestions for future advancement of clustering validation research. In addition, we review and evaluate the performance of internal and external clustering validation indices on the most common clustering algorithms, such as the evolutionary clustering algorithm star (ECA*). Finally, we suggest a classification framework for examining the functionality of both internal and external clustering validation measures regarding their ideal values, user-friendliness, responsiveness to input data, and appropriateness across various fields. This classification aids researchers in selecting the appropriate clustering validation measure to suit their specific requirements.

Stop location detection, within human mobility studies, has an impacts in multiple fields including urban planning, transport network design, epidemiological modeling, and socio-economic segregation analysis. However, it remains a challenging task because classical density clustering algorithms often struggle with noisy or incomplete GPS datasets. This study investigates the application of classification algorithms to enhance density-based methods for stop identification. Our approach incorporates multiple features, including individual routine behavior across various time scales and local characteristics of individual GPS points. The dataset comprises privacy-preserving and anonymized GPS points previously labeled as stops by a sequence-oriented, density-dependent algorithm. We simulated data gaps by removing point density from select stops to assess performance under sparse data conditions. The model classifies individual GPS points within trajectories as potential stops or non-stops. Given the highly imbalanced nature of the dataset, we prioritized recall over precision in performance evaluation. Results indicate that this method detects most stops, even in the presence of spatio-temporal gaps and that points classified as false positives often correspond to recurring locations for devices, typically near previous stops. While this research contributes to mobility analysis techniques, significant challenges persist. The lack of ground truth data limits definitive conclusions about the algorithm's accuracy. Further research is needed to validate the method across diverse datasets and to incorporate collective behavior inputs., Comment: paper produced during the Complexity72h conference

It has been known for many years that methods inspired by string theory, such as the worldline formalism, allow one to write down integral representations that combine large numbers of Feynman diagrams of different topologies. However, to make this fact useful for state-of-the-art calculations one has to confront non-standard integration problems where neither the known integration techniques for Feynman diagrams nor algebraic manipulation programs are of much help. Here I will give a progress report on this long-term project focussing on photon amplitudes at one and two loops, in vacuum and in external fields., Comment: 13 pages, 3 figures, talk given by Christian Schubert at Loops and Legs in Quantum Field Theory (LL2024), 14-19 April 2024, Wittenberg, Germany

In this paper, a swarm intelligence optimization algorithm is proposed as the Shrike Optimization Algorithm (SHOA). Many creatures living in a group and surviving for the next generation randomly search for food; they follow the best one in the swarm, called swarm intelligence. Swarm-based algorithms are designed to mimic creatures' behaviours, but in multimodal problem competition, they cannot find optimal solutions in some difficult cases. The main inspiration for the proposed algorithm is taken from the swarming behaviours of shrike birds in nature. The shrike birds are migrating from their territory to survive. However, the SHOA mimics the surviving behaviour of shrike birds for living, adaptation, and breeding. Two parts of optimization exploration and exploitation are designed by modelling shrike breeding and searching for foods to feed nestlings until they get ready to fly and live independently. This paper is a mathematical model for the SHOA to perform optimization. The SHOA benchmarked 19 well-known mathematical test functions, 10 from CEC-2019, and 12 from CEC-2022 most recent test functions, a total of 41 competitive mathematical test functions benchmarked and four real-world engineering problems with different conditions, both constrained and unconstrained. The statistical results obtained from the Wilcoxon sum ranking and Fridman test show that SHOA has a significant statistical superiority in handling the test benchmarks compared to competitor algorithms in multi-modal problems. The results for engineering optimization problems show the SHOA outperforms other nature-inspired algorithms in many cases., Comment: 25 pages

Writing secure code is challenging and so it is expected that, following the release of code-generative AI tools, such as ChatGPT and GitHub Copilot, developers will use these tools to perform security tasks and use security APIs. However, is the code generated by ChatGPT secure? How would the everyday software or security engineer be able to tell? As we approach the next decade we expect a greater adoption of code-generative AI tools and to see developers use them to write secure code. In preparation for this, we need to ensure security-by-design. In this paper, we look back in time to Saltzer & Schroeder's security design principles as they will need to evolve and adapt to the challenges that come with a world of AI-generated code.

Bat Algorithm (BA) is a nature-inspired metaheuristic search algorithm designed to efficiently explore complex problem spaces and find near-optimal solutions. The algorithm is inspired by the echolocation behavior of bats, which acts as a signal system to estimate the distance and hunt prey. Although the BA has proven effective for various optimization problems, it exhibits limited exploration ability and susceptibility to local optima. The algorithm updates velocities and positions based on the current global best solution, causing all agents to converge towards a specific location, potentially leading to local optima issues in optimization problems. On this premise, this paper proposes the Modified Bat Algorithm (MBA) as an enhancement to address the local optima limitation observed in the original BA. MBA incorporates the frequency and velocity of the current best solution, enhancing convergence speed to the optimal solution and preventing local optima entrapment. While the original BA faces diversity issues, both the original BA and MBA are introduced. To assess MBAs performance, three sets of test functions (classical benchmark functions, CEC2005, and CEC2019) are employed, with results compared to those of the original BA, Particle Swarm Optimization (PSO), Genetic Algorithm (GA), and Dragonfly Algorithm (DA). The outcomes demonstrate the MBAs significant superiority over other algorithms. Additionally, MBA successfully addresses a real-world assignment problem (call center problem), traditionally solved using linear programming methods, with satisfactory results.

Beam search is a standard tree search algorithm when it comes to finding sequences of maximum likelihood, for example, in the decoding processes of large language models. However, it is myopic since it does not take the whole path from the root to a leaf into account. Moreover, it is agnostic to prior knowledge available about the process: For example, it does not consider that the objective being maximized is a likelihood and thereby has specific properties, like being bound in the unit interval. Taking a probabilistic approach, we define a prior belief over the LLMs' transition probabilities and obtain a posterior belief over the most promising paths in each iteration. These beliefs are helpful to define a non-myopic Bayesian-optimization-like acquisition function that allows for a more data-efficient exploration scheme than standard beam search. We discuss how to select the prior and demonstrate in on- and off-model experiments with recent large language models, including Llama-2-7b, that our method achieves higher efficiency than beam search: Our method achieves the same or a higher likelihood while expanding fewer nodes than beam search., Comment: 10 pages

Cygnus X-3 is an enigmatic X-ray binary, that is both an exceptional accreting system and a cornerstone for the population synthesis studies. Prominent X-ray and radio properties follow a well-defined pattern, yet the physical reasons for the state changes observed in this system are not known. Recently, the presence of an optically thick envelope around the central source in the hard state was revealed using the X-ray polarization data obtained with Imaging X-ray Polarimetry Explorer (IXPE). In this work, we analyse IXPE data obtained in the ultrasoft (radio quenched) state of the source. The average polarization degree (PD) of $11.9\pm0.5\%$ at a polarization angle (PA) of $94^{\circ}\pm1^{\circ}$ is inconsistent with the simple geometry of the accretion disc viewed at an intermediate inclination. The high PD, the blackbody-like spectrum, and the weakness of fluorescent iron line imply that the central source is hidden behind the optically thick outflow and its beamed radiation is scattered towards our line of sight. In this picture the observed PD is directly related to the source inclination, which we conservatively determine to lie in the range $26^{\circ}

Multi-Objective Optimization (MOO) techniques have become increasingly popular in recent years due to their potential for solving real-world problems in various fields, such as logistics, finance, environmental management, and engineering. These techniques offer comprehensive solutions that traditional single-objective approaches fail to provide. Due to the many innovative algorithms, it has been challenging for researchers to choose the optimal algorithms for solving their problems. This paper examines recently developed MOO-based algorithms. MOO is introduced along with Pareto optimality and trade-off analysis. In real-world case studies, MOO algorithms address complicated decision-making challenges. This paper examines algorithmic methods, applications, trends, and issues in multi-objective optimization research. This exhaustive review explains MOO algorithms, their methods, and their applications to real-world problems. This paper aims to contribute further advancements in MOO research. No singular strategy is superior; instead, selecting a particular method depends on the natural optimization problem, the computational resources available, and the specific objectives of the optimization tasks., Comment: 21 pages

Artificial neural networks play a crucial role in machine learning and there is a need to improve their performance. This paper presents FOXANN, a novel classification model that combines the recently developed Fox optimizer with ANN to solve ML problems. Fox optimizer replaces the backpropagation algorithm in ANN; optimizes synaptic weights; and achieves high classification accuracy with a minimum loss, improved model generalization, and interpretability. The performance of FOXANN is evaluated on three standard datasets: Iris Flower, Breast Cancer Wisconsin, and Wine. The results presented in this paper are derived from 100 epochs using 10-fold cross-validation, ensuring that all dataset samples are involved in both the training and validation stages. Moreover, the results show that FOXANN outperforms traditional ANN and logistic regression methods as well as other models proposed in the literature such as ABC-ANN, ABC-MNN, CROANN, and PSO-DNN, achieving a higher accuracy of 0.9969 and a lower validation loss of 0.0028. These results demonstrate that FOXANN is more effective than traditional methods and other proposed models across standard datasets. Thus, FOXANN effectively addresses the challenges in ML algorithms and improves classification performance., Comment: 12 pages

Using $(27.12\pm 0.14)\times10^8$ $\psi(3686)$ decays and data samples of $e^+e^-$ collisions with $\sqrt{s}$ from 4.130 to 4.780~GeV collected with the BESIII detector, we report the first observation of the electromagnetic Dalitz transition $h_c\to e^+e^-\eta_c$ with a statistical significance of $5.4\sigma$. We measure the ratio of the branching fractions $\frac{\mathcal{B}(h_c\rightarrow e^+e^-\eta_c)}{\mathcal{B}(h_c\rightarrow \gamma \eta_c)}$ separately for the $h_c$ samples produced via $\psi(3686)\to\pi^0h_c$ and $e^+e^-\to\pi^+\pi^-h_c$. The average ratio is determined to be $(0.59\pm0.10(\text{stat.})\pm0.04(\text{syst.}))\%$, where the uncertainty includes both statistical and systematic components.

Fitness dependent optimizer (FDO) is considered one of the novel swarm intelligent algorithms. Recently, FDO has been enhanced several times to improve its capability. One of the improvements is called improved FDO (IFDO). However, according to the research findings, the variants of FDO are constrained by two primary limitations that have been identified. Firstly, if the number of agents employed falls below five, it significantly diminishes the algorithm's precision. Secondly, the efficacy of FDO is intricately tied to the quantity of search agents utilized. To overcome these limitations, this study proposes a modified version of IFDO, called M-IFDO. The enhancement is conducted by updating the location of the scout bee to the IFDO to move the scout bees to achieve better performance and optimal solutions. More specifically, two parameters in IFDO, which are alignment and cohesion, are removed. Instead, the Lambda parameter is replaced in the place of alignment and cohesion. To verify the performance of the newly introduced algorithm, M-IFDO is tested on 19 basic benchmark functions, 10 IEEE Congress of Evolutionary Computation (CEC-C06 2019), and five real-world problems. M-IFDO is compared against five state-of-the-art algorithms: Improved Fitness Dependent Optimizer (IFDO), Improving Multi-Objective Differential Evolution algorithm (IMODE), Hybrid Sampling Evolution Strategy (HSES), Linear Success-History based Parameter Adaptation for Differential Evolution (LSHADE) and CMA-ES Integrated with an Occasional Restart Strategy and Increasing Population Size and An Iterative Local Search (NBIPOP-aCMAES). The verification criteria are based on how well the algorithm reaches convergence, memory usage, and statistical results. The results show that M-IFDO surpasses its competitors in several cases on the benchmark functions and five real-world problems.

Artificial intelligence (AI) has become a crucial instrument for streamlining processes in various industries, including electrical power systems, as a result of recent digitalization. Algorithms for artificial intelligence are data-driven models that are based on statistical learning theory and are used as a tool to take use of the data that the power system and its users generate. Initially, we perform a thorough literature analysis of artificial intelligence (AI) applications related to renewable energy (RE). Next, we present a thorough analysis of renewable energy factories and assess their suitability, along with a list of the most widely used and appropriate AI algorithms. Nine AI-based strategies are identified here to assist Renewable Energy (RE) in contemporary power systems. This survey paper comprises an extensive review of the several AI techniques used for renewable energy as well as a methodical analysis of the literature for the study of various intelligent system application domains across different disciplines of renewable energy. This literature review identifies the performance and outcomes of nine different research methods by assessing them, and it aims to distill valuable insights into their strengths and limitations. This study also addressed three main topics: using AI technology for renewable power generation, utilizing AI for renewable energy forecasting, and optimizing energy systems. Additionally, it explored AI's superiority over conventional models in controllability, data handling, cyberattack prevention, smart grid implementation, robotics- AI's significance in shaping the future of the energy industry. Furthermore, this article outlines future directions in the integration of AI for renewable energy.

Reduced environmental effect, lower operating costs, and a stable and sustainable energy supply for current and future generations are the main reasons why power optimization is important. Power optimization makes ensuring that energy is used more effectively, cutting down on waste and optimizing the utilization of resources.In today's world, power optimization and artificial intelligence (AI) integration are essential to changing the way energy is produced, used, and distributed. Real-time monitoring and analysis of power usage trends is made possible by AI-driven algorithms and predictive analytics, which enable dynamic modifications to effectively satisfy demand. Efficiency and sustainability are increased when power consumption is optimized in different sectors thanks to the use of intelligent systems. This survey paper comprises an extensive review of the several AI techniques used for power optimization as well as a methodical analysis of the literature for the study of various intelligent system application domains across different disciplines of power consumption.This literature review identifies the performance and outcomes of 17 different research methods by assessing them, and it aims to distill valuable insights into their strengths and limitations. Furthermore, this article outlines future directions in the integration of AI for power consumption optimization.

This study sequentially investigates the enhancement of fracture behavior in adhesive joints. Initially, mechanical sanding techniques were explored, revealing significant improvements of 55% in fracture energy and 38% in load-bearing capacity with optimal 240-grit sanding. Subsequently, in a second stage, Multi-Walled Carbon Nanotubes (MWCNTs) were introduced at varying weight percentages (0.1%, 0.3%, and 0.5% by weight) into the adhesive joints. This combined approach aimed to synergistically examine the impact of surface pretreatments and nanoparticle integration on fracture behavior. Results highlight the potential for fracture resistance and enhanced load-carrying capabilities in reinforced specimens, with maximum load and fracture energy improvements of up to 92 % and 50 %, respectively, compared to unreinforced specimens and those subjected solely to sanding treatments. Furthermore, the fracture mode shifted from adhesive failure to cohesive substrate failure in reinforced specimens. This comprehensive investigation underscores the intricate interplay between mechanical preparation and nanomaterial incorporation in significantly improving adhesive joint performance

Cyber Essentials (CE) comprise a set of controls designed to protect organisations, irrespective of their size, against cyber attacks. The controls are firewalls, secure configuration, user access control, malware protection & security update management. In this work, we explore the extent to which CE remains robust against an ever-evolving threat landscape. To that end, we reconstruct 45 breaches mapped to MiTRE ATT&CK using an Incident Fault Tree ( IFT ) approach. Our method reveals the intersections where the placement of controls could have protected organisations. Then we identify appropriate Cyber Essential controls and/or Additional Controls for these vulnerable intersections. Our results show that CE controls can effectively protect against most attacks during the initial attack phase. However, they may need to be complemented with additional Controls if the attack proceeds further into organisational systems & networks. The Additional Controls (AC) we identify include back-ups, security awareness, logging and monitoring. Our analysis brings to the fore a foundational issue as to whether controls should exclude recovery and focus only on pre-emption. The latter makes the strong assumption that a prior identification of all controls in a dynamic threat landscape is indeed possible. Furthermore, any potential broadening of technical controls entails re-scoping the skills that are required for a Cyber Essentials (CE) assessor. To that end, we suggest human factors and security operations and incident management as two potential knowledge areas from Cyber Security Body of Knowledge (CyBOK) if there is any broadening of CE based on these findings., Comment: We have submitted this paper in ACM Digital Threats: Research and Practice (DTRAP) Journal. The paper is currently in the review process

There has been a growing need to devise processes that can create comprehensive datasets in the world of Computer Algebra, both for accurate benchmarking and for new intersections with machine learning technology. We present here a method to generate integrands that are guaranteed to be integrable, dubbed the LIOUVILLE method. It is based on Liouville's theorem and the Parallel Risch Algorithm for symbolic integration. We show that this data generation method retains the best qualities of previous data generation methods, while overcoming some of the issues built into that prior work. The LIOUVILLE generator is able to generate sufficiently complex and realistic integrands, and could be used for benchmarking or machine learning training tasks related to symbolic integration., Comment: To appear in proc. CASC (2024)

This paper presents the development of a prototype Automatic Speech Recognition (ASR) system specifically designed for Bengali biomedical data. Recent advancements in Bengali ASR are encouraging, but a lack of domain-specific data limits the creation of practical healthcare ASR models. This project bridges this gap by developing an ASR system tailored for Bengali medical terms like symptoms, severity levels, and diseases, encompassing two major dialects: Bengali and Sylheti. We train and evaluate two popular ASR frameworks on a comprehensive 46-hour Bengali medical corpus. Our core objective is to create deployable health-domain ASR systems for digital health applications, ultimately increasing accessibility for non-technical users in the healthcare sector.

Large language models (LLMs) can significantly be improved by aligning to human preferences -- the so-called reinforcement learning from human feedback (RLHF). However, the cost of fine-tuning an LLM is prohibitive for many users. Due to their ability to bypass LLM finetuning, tokenwise reward-guided text generation (RGTG) methods have recently been proposed. They use a reward model trained on full sequences to score partial sequences during a tokenwise decoding, in a bid to steer the generation towards sequences with high rewards. However, these methods have so far been only heuristically motivated and poorly analyzed. In this work, we show that reward models trained on full sequences are not compatible with scoring partial sequences. To alleviate this issue, we propose to explicitly train a Bradley-Terry reward model on partial sequences, and autoregressively sample from the implied tokenwise policy during decoding time. We study the property of this reward model and the implied policy. In particular, we show that this policy is proportional to the ratio of two distinct RLHF policies. We show that our simple approach outperforms previous RGTG methods and achieves similar performance as strong offline baselines but without large-scale LLM finetuning.

Query expansion has been employed for a long time to improve the accuracy of query retrievers. Earlier works relied on pseudo-relevance feedback (PRF) techniques, which augment a query with terms extracted from documents retrieved in a first stage. However, the documents may be noisy hindering the effectiveness of the ranking. To avoid this, recent studies have instead used Large Language Models (LLMs) to generate additional content to expand a query. These techniques are prone to hallucination and also focus on the LLM usage cost. However, the cost may be dominated by the retrieval in several important practical scenarios, where the corpus is only available via APIs which charge a fee per retrieved document. We propose combining classic PRF techniques with LLMs and create a progressive query expansion algorithm ProQE that iteratively expands the query as it retrieves more documents. ProQE is compatible with both sparse and dense retrieval systems. Our experimental results on four retrieval datasets show that ProQE outperforms state-of-the-art baselines by 37% and is the most cost-effective.

Training agents to behave as desired in complex 3D environments from high-dimensional sensory information is challenging. Imitation learning from diverse human behavior provides a scalable approach for training an agent with a sensible behavioral prior, but such an agent may not perform the specific behaviors of interest when deployed. To address this issue, we draw an analogy between the undesirable behaviors of imitation learning agents and the unhelpful responses of unaligned large language models (LLMs). We then investigate how the procedure for aligning LLMs can be applied to aligning agents in a 3D environment from pixels. For our analysis, we utilize an academically illustrative part of a modern console game in which the human behavior distribution is multi-modal, but we want our agent to imitate a single mode of this behavior. We demonstrate that we can align our agent to consistently perform the desired mode, while providing insights and advice for successfully applying this approach to training agents. Project webpage at https://adamjelley.github.io/aligning-agents-like-llms .

A cyber security organisation needs to ensure that its workforce possesses the necessary knowledge to fulfil its cyber security business functions. Similarly, where an organisation chooses to delegate their cyber security tasks to a third party provider, they must ensure that the chosen entity possesses robust knowledge capabilities to effectively carry out the assigned tasks. Building a comprehensive cyber security knowledge profile is a distinct challenge; the field is ever evolving with a range of professional certifications, academic qualifications and on-the-job training. So far, there has been a lack of a well-defined methodology for systematically evaluating an organisation's cyber security knowledge, specifically derived from its workforce, against a standardised reference point. Prior research on knowledge profiling across various disciplines has predominantly utilised established frameworks such as SWEBOK. However, within the domain of cyber security, the absence of a standardised reference point is notable. In this paper, we advance a framework leveraging CyBOK, to construct an organisation's knowledge profile. The framework enables a user to identify areas of coverage and where gaps may lie, so that an organisation can consider targeted recruitment or training or, where such expertise may be outsourced, drawing in knowledge capability from third parties. In the latter case, the framework can also be used as a basis for assessing the knowledge capability of such a third party. We present the knowledge profiling framework, discussing three case studies in organisational teams underpinning its initial development, followed by its refinement through workshops with cyber security practitioners.

Low radio frequency spectral index measurements are a powerful tool to distinguish between different emission mechanisms and, in turn, to understand the nature of the sources. Besides the standard method of estimating the ``broadband" spectral index of sources from observations in two different frequency ``bands", if the observations were made with large instantaneous bandwidth, the ``in-band" spectral index can be determined, either using images of emission at multiple frequency ranges within a band or using the novel Multi Term-Multi Frequency Synthesis (MT-MFS) imaging algorithm. Here, using simulated upgraded Giant Metrewave Radio Telescope (uGMRT) data, we have systematically studied the reliability of various methods of spectral index estimation for sources with a wide range of signal-to-noise ratio (SNR). It is found that, for synthetic uGMRT point source data, the MT-MFS imaging algorithm produces in-band spectral indices for SNR~$\lesssim100$ that have errors $\gtrsim 0.2$, making them unreliable. However, at a similar SNR, the sub-band splitting method produces errors $\lesssim 0.2$, which are more accurate and unbiased in-band spectral indices. The broadband spectral indices produce errors $\lesssim 0.2$ even for SNR $\gtrsim 15$, and hence, they are most reliable if there are no higher-order variations in the spectral index. These results may be used to improve the uGMRT observation and data analysis strategies depending on the brightness of the target source., Comment: 10 pages, 5 figures, accepted by ApJ on 24th May 2024

In an era where digital platforms increasingly mediate public discourse, grasping the complexities and nuances in affective polarization--especially as influenced by key figures on social media--has never been more vital. This study delves into the intricate web of interactions on Twitter, now rebranded as 'X', to unravel how influencer-led conversations catalyze shifts in public sentiment, laying bare the complex dynamics that underpin online polarization. Employing a novel methodological framework that includes counterfactual analysis, we analyze scenarios with and without specific influencer-led conversations. Our findings illuminate the significant role influencers play in shaping public discourse, offering insights into the mechanisms of online polarization and suggesting pathways for future research to mitigate divisiveness and explore new methods for quantifying affective polarization. This research contributes to the broader understanding of digital communication's impact on societal polarization, underscoring the importance of detailed analysis in developing strategies to foster a more cohesive digital public sphere., Comment: 8 pages, 4 figures

Coded caching scheme originally proposed by Maddah-Ali and Niesen (MN) considered a broadcast network consisting of a single server connected to a set of users each having a cache memory. Motivated by practical scenarios, Karamchandani \textit{et al.} in [16] proposed a coded caching scheme for a two-layer hierarchical network consisting of a single server connected to multiple mirror sites and each mirror site connected to a distinct set of users, in which both mirror sites and users having cache memories. Low subpacketization level coded caching schemes are desirable for practical implementations. Placement delivery array (PDA) was proposed as a tool to design coded caching schemes with reduced subpacketization level by Yan \textit{et al.} in [4]. Schemes with reduced subpacketization levels are studied extensively in the literature for single-layer networks. Kong \textit{et al.} in [17] proposed a structure called hierarchical placement delivery arrays (HPDA), which characterizes a hierarchical coded caching system and also proposed a class of HPDAs that gives low subpacketization level schemes by using two PDAs. Low subpacketization level hierarchical schemes using combinatorial $t$-designs is proposed in [20]. Apart from that there is no other existing work that discusses the subpacketization problem in a hierarchical network. This paper proposes a class of HPDA construction that gives low subpacketization level hierarchical coded caching schemes, by first constructing a new class of PDAs. Compared with the existing schemes, in cases where the system parameters and subpacketization level are the same, the proposed hierarchical scheme has a better coding delay. Further, the new class of PDAs constructed either subsumes several known PDA constructions or achieves better transmission load for the same system parameters., Comment: Added Section IV - (performance analysis of proposed HPDA construction). The term 'coding delay' is formally defined (page no. 5). 14 pages, 10 figures and 9 tables

The advancement of sophisticated artificial intelligence (AI) algorithms has led to a notable increase in energy usage and carbon dioxide emissions, intensifying concerns about climate change. This growing problem has brought the environmental sustainability of AI technologies to the forefront, especially as they expand across various sectors. In response to these challenges, there is an urgent need for the development of sustainable AI solutions. These solutions must focus on energy-efficient embedded systems that are capable of handling diverse data types even in environments with limited resources, thereby ensuring both technological progress and environmental responsibility. Integrating complementary multimodal data into tiny machine learning models for edge devices is challenging due to increased complexity, latency, and power consumption. This work introduces TinyM$^2$Net-V3, a system that processes different modalities of complementary data, designs deep neural network (DNN) models, and employs model compression techniques including knowledge distillation and low bit-width quantization with memory-aware considerations to fit models within lower memory hierarchy levels, reducing latency and enhancing energy efficiency on resource-constrained devices. We evaluated TinyM$^2$Net-V3 in two multimodal case studies: COVID-19 detection using cough, speech, and breathing audios, and pose classification from depth and thermal images. With tiny inference models (6 KB and 58 KB), we achieved 92.95% and 90.7% accuracies, respectively. Our tiny machine learning models, deployed on resource limited hardware, demonstrated low latencies within milliseconds and very high power efficiency., Comment: Accepted at AAAI 2024 Workshop SAI

We describe the design and results from the BraTS 2023 Intracranial Meningioma Segmentation Challenge. The BraTS Meningioma Challenge differed from prior BraTS Glioma challenges in that it focused on meningiomas, which are typically benign extra-axial tumors with diverse radiologic and anatomical presentation and a propensity for multiplicity. Nine participating teams each developed deep-learning automated segmentation models using image data from the largest multi-institutional systematically expert annotated multilabel multi-sequence meningioma MRI dataset to date, which included 1000 training set cases, 141 validation set cases, and 283 hidden test set cases. Each case included T2, T2/FLAIR, T1, and T1Gd brain MRI sequences with associated tumor compartment labels delineating enhancing tumor, non-enhancing tumor, and surrounding non-enhancing T2/FLAIR hyperintensity. Participant automated segmentation models were evaluated and ranked based on a scoring system evaluating lesion-wise metrics including dice similarity coefficient (DSC) and 95% Hausdorff Distance. The top ranked team had a lesion-wise median dice similarity coefficient (DSC) of 0.976, 0.976, and 0.964 for enhancing tumor, tumor core, and whole tumor, respectively and a corresponding average DSC of 0.899, 0.904, and 0.871, respectively. These results serve as state-of-the-art benchmarks for future pre-operative meningioma automated segmentation algorithms. Additionally, we found that 1286 of 1424 cases (90.3%) had at least 1 compartment voxel abutting the edge of the skull-stripped image edge, which requires further investigation into optimal pre-processing face anonymization steps., Comment: 16 pages, 11 tables, 10 figures, MICCAI

The ability to build and leverage world models is essential for a general-purpose AI agent. Testing such capabilities is hard, in part because the building blocks of world models are ill-defined. We present Elements of World Knowledge (EWOK), a framework for evaluating world modeling in language models by testing their ability to use knowledge of a concept to match a target text with a plausible/implausible context. EWOK targets specific concepts from multiple knowledge domains known to be vital for world modeling in humans. Domains range from social interactions (help/hinder) to spatial relations (left/right). Both, contexts and targets are minimal pairs. Objects, agents, and locations in the items can be flexibly filled in enabling easy generation of multiple controlled datasets. We then introduce EWOK-CORE-1.0, a dataset of 4,374 items covering 11 world knowledge domains. We evaluate 20 openweights large language models (1.3B--70B parameters) across a battery of evaluation paradigms along with a human norming study comprising 12,480 measurements. The overall performance of all tested models is worse than human performance, with results varying drastically across domains. These data highlight simple cases where even large models fail and present rich avenues for targeted research on LLM world modeling capabilities., Comment: 21 pages (11 main), 7 figures. Authors Anna Ivanova, Aalok Sathe, Benjamin Lipkin contributed equally

We present the first search for the leptonic decays $D^{*+}\to e^+\nu_e$ and $D^{*+}\to \mu^+\nu_\mu$ by analyzing a data sample of electron-positron collisions recorded with the BESIII detector at center-of-mass energies between 4.178 and 4.226 GeV, corresponding to an integrated luminosity of 6.32~fb$^{-1}$. No significant signal is observed. The upper limits on the branching fractions for $D^{*+}\to e^+\nu_e$ and $D^{*+}\to \mu^+\nu_\mu$ are set to be $1.1 \times 10^{-5}$ and $4.3 \times 10^{-6}$ at 90\% confidence level, respectively., Comment: 14 pages, 7 figures

Following the recent release of AI assistants, such as OpenAI's ChatGPT and GitHub Copilot, the software industry quickly utilized these tools for software development tasks, e.g., generating code or consulting AI for advice. While recent research has demonstrated that AI-generated code can contain security issues, how software professionals balance AI assistant usage and security remains unclear. This paper investigates how software professionals use AI assistants in secure software development, what security implications and considerations arise, and what impact they foresee on secure software development. We conducted 27 semi-structured interviews with software professionals, including software engineers, team leads, and security testers. We also reviewed 190 relevant Reddit posts and comments to gain insights into the current discourse surrounding AI assistants for software development. Our analysis of the interviews and Reddit posts finds that despite many security and quality concerns, participants widely use AI assistants for security-critical tasks, e.g., code generation, threat modeling, and vulnerability detection. Their overall mistrust leads to checking AI suggestions in similar ways to human code, although they expect improvements and, therefore, a heavier use for security tasks in the future. We conclude with recommendations for software professionals to critically check AI suggestions, AI creators to improve suggestion security and capabilities for ethical security tasks, and academic researchers to consider general-purpose AI in software development., Comment: 20 pages, 2 figures, 3 tables

We study nonequilibrium Andreev resonances in a voltage-biased graphene three-terminal Josephson junction (JJ). We observe periodic oscillations of resistance with maxima at multiples of the magnetic flux quantum (noninversion regime). As we increase the bias voltage, we further observe a transition point beyond which oscillations exhibit a $\pi$ phase shift (inversion regime) with maxima of resistance occuring at multiples of half-flux quantum. At this transition point, the frequency of the oscillations is doubled. We develop a model based on the coupling of the static Andreev bound states (ABSs) to the nonequilibrium Fermi surface of graphene to explain the observed noninversion to inversion crossovers. Our model associates these crossovers to microscopic phase-sensitive Andreev reflections which couple the normal and superfluid components of the current. Our findings show that multiterminal JJs can be used to engineer unconventional energy-phase relations such as those expected in the $\pi$-shifted ABSs without relying on quartet and Floquet physics. These nonequilibrium ABSs could potentially find applications in superconducting $\pi$ qubits.