Authors - Rehab Abdulmonem Ali Alshireef Abstract - The advent of large language models (LLMs) has marked a turning point in artificial intelligence applications within healthcare. Med-PaLM 2, developed by Google, stands out as a specialized model trained on medical data that has demonstrated expert-level performance on the USMLE. This literature review explores the educational potential of Med-PaLM 2 across different learner levels—medical students, residents, and practicing physicians. It evaluates the benefits, limitations, and contextual challenges of adopting such AI tools in the Arab world, particularly in remote education and clinical skills laboratories. While Med-PaLM 2 offers new opportunities for personalized learning and simulation-based training, its integration must be guided by ethical frameworks, policy development, and regional adaptation efforts to ensure equitable and effective implementation.
Authors - Aliou Ngor Diouf, Ibrahimma Fall, Lamine Diop Abstract - SenSCHOOL Ontology is an extension of the CIDOC-CRM (Conceptual Reference Model) designed to model and integrate information about Tariqas, Sufi religious brotherhoods present in West Africa in its diversity. CIDOC-CRM is a widely used generic data model for exchanging and integrating information from a variety of heterogeneous cultural heritage (CH) sources. The main objective of SenSCHOOL is to facilitate the management, preservation, and exchange of information about Tariqas in a structured and organized manner. We describe the methodology and steps we followed to design SenSCHOOL. We present the implementation of SenSCHOOL, which enables the integration and structuring of information from heterogeneous sources.
Authors - Toni Tani, Lasse Metso, Timo Karri Abstract - Digital transformation is reshaping business ecosystems through advances in artificial intelligence (AI), process automation, enhanced analytics, improved information visualization, and increased innovation. This study examines the impact of AI on ecosystems using traditional bibliometric analysis and a unique approach to processing large volumes of textual data. First, 232 documents published between 2014 and 2024 from the Scopus database were analyzed using Bibliometrix and Biblioshiny to identify influential authors, thematic clusters, and emerging research areas. In the second phase, a text network software called Infranodus was used to scan and analyze the 54 most relevant abstracts from 2023-2024, after which the extracted insights were refined using generative AI (genAI). Subsequently, the extracted information was further developed via prompt engineering from visual graphs and ChatGPT, revealing interesting results that demonstrated the potential of genAI in repeatedly conducting research and managing business ecosystems. Ultimately, this study shows a novel way of combining bibliometric data and visual prompt engineering to harness dynamic relations iteratively.
Authors - Syeda Sohail, Maurice van Keulen Abstract - Process mining enables organizations to gain actionable insights into their business processes by analyzing digital footprints extracted from information systems. These insights unravel inefficiencies and exude process enhancement through bottleneck detection and conformance checking. This paper presents a case study where process mining is applied to five real-world event logs of a Commerce Platform-as-a- Service provider to expedite the business process by reducing waiting times and minimizing multiple customer interactions. A comprehensive process mining project methodology was implemented to conduct the case study. The findings revealed key bottlenecks and underlying factors that contribute to delays and excessive customer interactions. In response, process enhancement recommendations were implemented with the organization’s template adjustments for an efficient business process optimization. The study also addresses the dilemma of privacy-utility tradeoff by ensuring that the event logs adhere to privacy-by-design requirements without compromising the utility of the data. Instead, the fulfilled requirements further refined process mining and data analysis by minimizing and abstracting event logs in this relatively less sensitive domain.
Authors - Salvatore Vella, Fatima Hussain, Salah Sharieh, Alex Ferworn Abstract - Policies and procedures coordinate the work of multiple knowledge workers. These are standardized workflows with specified inputs and outputs. AI agents can automate some or all of the steps in the workflow. The automation will greatly enhance efficiency, minimize human errors, enable the employees to focus on more strategic tasks and provide oversight for these more routine tasks. This paper examines the application of AI agents to understand and automate these workflows. We propose a framework where the policy or procedure is corrected via a large language model and translated to a simplified BPEL (Business Process Execution Language) form for later execution by AI agents. This two-step approach enables the creation of reusable policy and procedure libraries that the AI agents can reuse. We demonstrate that improved policies and procedures can be created from the code. Through case studies, we show the practical benefits in real-world office settings. Integrating AI agents in knowledge work professions is an important research topic; this framework shows how this can be done in a standardized way.We provide the source code and artifacts for these experiments.
Authors - Tsholofetso Taukobong, Audrey Naledi Masizana, George Anderson Abstract - The research contributes to the performance of X-Ray Transmission sensor-based sorting process during diamond sorting. The aim is to overcome the shortcomings of the current baseline methods for detection of small highly over-lapped and irregular shaped rock particles that could go undetected as part of the waste recovery process. Most methods work well when approximate shape and size is well known and particles are not highly overlapped. However due to challenges of over-segmentation or under-segmentation, several image segmentation techniques are explored in order to propose a new and improved segmentation process that aims to reduce false negatives and false positives, thus improving performance and efficiency of the waste recovery process. This reports on classical methods explorations and preliminary experiments on the ongoing research
Authors - Emilia-Loredana Pop, Augusta Ratiu, Daniela-Maria Cristea Abstract - In this article, we have performed an analysis related to the subjects Databases, Database Management Systems, and Web Programming for the students enrolled in Computer Science specializations. The data analyzed has been collected during one university year with the help of an anonymous survey. We have focused on students’ evaluation for these subjects, and comparisons related to gender and study lines (English and Romanian) have also been provided. The lectures and the labs were enjoyed for all the subjects, with small remarks, near to the interaction and communication. For databases, query optimization was harder, and for Web Programming, the solving of lab errors brought challenges. The evaluation was high for the subjects and exceeded 62%.
Authors - Ulrich Tedongmo Douanla, Jean Louis Kedieng Ebongue Fendji, Giquel Therance Sassa, Marcellin Atemkeng Abstract - The Internet has become an essential tool for modern activities and a fundamental right for digital inclusion. However, many regions, particularly in Africa, remain underserved, with limited or unstable Internet access. To address this issue, communities jointly with support organizations have deployed community networks, which are wireless infrastructures that provide connectivity to local populations. Despite their benefits, these networks frequently experience outages that impact both network infrastructure and associated services. Ensuring their reliability requires effective monitoring and supervision solutions. In this work, we propose a supervision platform that leverages the Simple Network Management Protocol (SNMP), log file analysis, and Big Data technologies to enable real-time monitoring of community networks. SNMP is employed to collect device status data, while log files provide insights into the performance of network applications. To facilitate scalable and real-time processing, we integrate Spark Structured Streaming, enabling continuous data analysis and proactive issue detection. The platform also includes an alerting system that delivers notifications via SMS, email, or other channels in case of failures. By providing a comprehensive view of network health and automating incident response, our solution enhances the availability and resilience of community networks, ultimately improving Internet access in underserved regions.
Authors - Vishnu Kumar Abstract - Heart disease remains a leading cause of mortality in the United States, responsible for approximately 1 in 5 deaths in 2022. Modifiable behavioral and lifestyle factors, such as smoking, physical activity, and diet, play a critical role in cardiovascular risk. This study applies a machine learning (ML) approach to predict heart disease risk in the U.S. using data from the 2022 Behavioral Risk Factor Surveillance System (BRFSS). Three ML based classification models were developed using ten key behavioral and lifestyle features: general health perception, days of poor physical and mental health, time since the last checkup, physical activity engagement, average sleep duration, smoking status, e-cigarette use, body mass index (BMI), and alcohol consumption. Among the three ML based classification models, XGBoost exhibited superior performance, achieving an F1-score of 0.92 with balanced precision and recall across both classes. SHAP (Shapley Additive Explanations) was then used to identify the impact of behavioral and lifestyle factors on heart disease risk. Global SHAP analysis revealed that general health, poor mental health, and BMI were the most influential features affecting heart disease risk. Local SHAP analysis showed that the importance of individual features varied across different observations, with factors such as: time since the last checkup, and smoking status significantly influencing heart disease risk for certain individuals. These findings demonstrate the potential of explainable ML techniques to identify actionable, personalized cardiovascular risk factors. The insights gained can help healthcare providers tailor interventions and prevention strategies, prioritize high-risk individuals for early detection, and allocate resources more effectively to reduce the burden of heart disease.
Authors - Fawzy Alsharif, Hasan Kaan Aldemir, Akay Deliorman Abstract - This paper presents image processing techniques for detecting eye diseases such as Vessel Tortuosity (VT), Glaucoma, Central Serous Retinopathy (CSR), and Diabetic Retinopathy (DR). The system supports early symptom detection, condition monitoring, and timely intervention. For VT, green channel extraction, Gaussian blurring, and Otsu thresholding isolate vessels, followed by morphological operations and thinning for curvature analysis. In Glaucoma, contrast enhancement and multi-level Otsu thresholding segment the optic disc and cup, enabling Cup-to-Disc Ratio calculation. For CSR, green channel processing and Gaussian blurring highlight fluid accumulation. In DR, lesion visibility is improved through green channel extraction, blurring, and morphological filtering. This integrated approach enhances image clarity and segmentation, achieving 97%–99% accuracy in early disease detection.
Authors - Doaa Abdelrahman, Heba Aslan, Mahmoud M. Nasreldin, Ghada Elkabbany, Mohamed Rasslan Abstract - The Bitcoin economy has grown significantly and rapidly, reaching an estimated market capitalization of around $1.87 trillion. As a type of cryptocurrency—essentially digital money—Bitcoin enables direct transactions between users without relying on a central authority or intermediary. These transactions are validated by network participants using cryptographic techniques and are permanently stored in a decentralized public ledger known as the blockchain. New Bitcoins are introduced into circulation through a process that is called mining, and they can be traded for conventional currencies, goods, or services. The dramatic increase in Bitcoin’s value has drawn the attention of both cybercriminals aiming to exploit system flaws for profit and researchers working to identify these vulnerabilities, devise protective measures, and anticipate future trends. It outlines the Bitcoin protocol by describing its main components, their functions, and how they interact. Moreover, it explores the foundational cryptographic concepts and existing weaknesses within the Bitcoin infrastructure and concludes by assessing the strength and effectiveness of current security approaches.
Authors - Faycal Fedouaki, Mouhsene Fri, Kaoutar Douaioui, Ayoub El Khairi Abstract - The merging of blockchain and the industrial internet of things (IIoT) will reshape how smart manufacturing systems operate. This paper proposes a conceptual framework for using blockchain's decentralization architecture, cryptographic integrity, and smart contract automation to improve process monitoring in industrial environments. With real-time data collection from IIoT devices and secure transparent Blockchain ledgers, the proposed model addresses important issues such as tampering, interoperability, and latency when it comes to decision making. It also supports real-time analytics through incorporating reduced latencies using edge processing and message queuing. Additional design principles will address scalability issues by layered Blockchain structures and fog computing nodes, allowing the framework to keep pace with rising data volumes and increasing device densities. Even though the model is built on the latest breakthroughs and conforms to the Industry 4.0 paradigms, a prototype and experimental simulations are planned to value its empirical viability. Notably, this work intends to establish a resilient and efficient digital infrastructure for Next Generation industrial process monitoring.
Authors - Davide Paglia, Lorenzo Rigatti, Andrea Sabatini, Fabrizio Venettoni Abstract - In the context of the new emerging trend in capital markets of tokenization of financial assets, the paper explains how financial security can be registered on a market blockchain settled in ECB central bank digital currency, in T+0 time and in compliance with the current European regulatory framework. A particular use case is explored through the design, implementation and use of a new market infrastructure DLT based, an enterprise application called DLT Bond Platform for the issuance of a digital bond settled in European Central Bank wholesale digital currency via a delivery versus payment process, using a layer 2 permissionless blockchain. After introducing the context and the problem statement in the first two sections, a general description of the solution proposed, and its novel contributions are provided in the third section. In the forth section the main components of the DLT Bond Platform are described in detail, both web2 and web3 as well as the related business processes, namely: (i) Management of the entire life cycle of a bond in digital form; (ii) Management of all the settlement phases envisaged by the bond also through atomic transactions for the simultaneous transfer of the securities and the corresponding cash flows (delivery versus payment or "DvP") through the use of the solution made available by the Bank of Italy as part of the European Central Bank initiative called New Technologies for Wholesale Settlement (iii) Identification, authorization and management of users, profiles and the respective roles on chain; (iv) real-time monitoring and audit trail. The final section focuses on the results obtained and on the completion of the validation process, ultimately dwelling on potential future developments.
Authors - Abdulrahman Azab, Paul De Geest, Sanjay K. Srikakulam, Tomas Vondrak, Mira Kuntz, Bjorn Gruning Abstract - Effective resource scheduling is critical in high-performance (HPC) and high-throughput computing (HTC) environments, where traditional scheduling systems struggle with resource contention, data locality, and fault tolerance. Meta-scheduling, which abstracts multiple schedulers for unified job allocation, addresses these challenges. Galaxy, a widely used platform for data-intensive computational analysis, employs the Total Perspective Vortex (TPV) system for resource scheduling. With over 550,000 users, Galaxy aims to optimize scheduling efficiency in large-scale environments. While TPV offers flexibility, its decision-making can be enhanced by incorporating real-time resource availability and job status. This paper introduces the TPV Broker, a meta-scheduling framework that integrates real-time resource data to enable dynamic, data-aware scheduling. TPV Broker enhances scalability, resource utilization, and scheduling efficiency in Galaxy, offering potential for further improvements in distributed computing environments.
Authors - Nina Valchkova, Vasil Tsvetkov Abstract - This paper investigates the dynamic characteristics of a collaborative robotic mobile platform with enhanced manipulability. It`s motion parameters, such as linear velocities and accelerations, and their influence on platform control are analyzed. The experiments performed include monitoring acceleration processes, constant lateral movement, and deceleration and braking phases. The presented graphical analyses demonstrate key features of the platform dynamics that can be used to optimize the control of a collaborative robotic mobile platform.
Authors - Sakshi Tiwari, Snigdha Bisht, Kanchan Sharma Abstract - Effective waste management is critical to achieving sustainability in urban regions like Delhi-NCR, where heterogeneous waste streams pose a classification challenge. In this research, we propose WasteIQNet, an intelligent deep hybrid model designed for precise waste classification across 18 categories under a well-defined hierarchy: Wet (Compostable, Special_Disposal) and Dry (Recycle, Reduce, Reuse). Leveraging the WEDR dataset, we first standardized over 1.75 lakh images via JPEG conversion, 256×256 resizing, and RGB formatting. SMOTE+ENN was applied to balance class distributions to 20,000 images each. Feature extraction was achieved through simulated DASC-like global vector embeddings using MobileNetV3. Our baseline hybrid model integrated MobileNetV3Large and GraphSAGE, achieving an initial accuracy of 80.56%. After optimizing the model for multi-label learning through sigmoid activation, threshold-based decoding, and hierarchical label interpretation, we conducted extensive enhancements. Hyperparameter tuning with Optuna, Feature-wise Attention (FWA), and Top-K Mixture of Experts (TopK-MoE) improved accuracy to 83.33%. Subsequent normalization and activation function experiments (Mish, Swish, GELU) led to a peak accuracy of 94.44% using GELU. We further introduced Dynamic Sparse Training (DST) and Model-Agnostic Meta-Learning (MAML), raising accuracy to 95.04%. Final enhancements included label smoothing and early stopping, culminating in a best-in-class accuracy of 97.87%. WasteIQNet demonstrates a scalable, interpretable, and high-performance solution for automated waste classification, supporting smart city initiatives and responsible environmental management.
Authors - Abdullah I Alshoshan Abstract - Verification and/or identification (VI) of the individual writer-identity is one of the most common secure personal biometric authentications, particularly in banks for verification and in sensitive data storages for both VI. A writer-identity VI system (WIVIS) is proposed using the writer-invariant features of his/her script using two approaches: offline approaches, which rely on the script information in a static format, such as an image or shape, and online approaches, which require the collection of information in a dynamic format, such as speed and acceleration, using a tablet with a stylus pen to capture both of these dynamic information. Both offline script VI methods, such as normalized Fourier transform descriptor (NFTD) and normalized central moment (NCM), and online script VI methods, such as normalized script speed and acceleration, will be discussed. These features are compared individually and then as a combination. In the combination mode, the neural network (NN) is used for classification. Implementation and testing of the WIVIS is done and analyzed, and the effectiveness of each invariant algorithm, regardless of the language or form of the script shape, is discussed. A set of data on the online (dynamic) and offline (static) script is also discussed.
Authors - Dzung Lai Ngoc, Maria Luojus, Jukka Heikkonen, Rajeev Kanth Abstract - The development of cancer therapeutic therapies has made significant advancements in recent years. Numerous innovative solutions have emerged, achieving notable success, including immunotherapy, targeted drugs, and, among them, oncolytic viruses. Oncolytic virus therapy represents the first instance in which humans have employed a biological logic program, rather than a conventional drug, to treat a disease. Despite its promising potential, clinical trials involving oncolytic viruses have not yielded the anticipated outcomes, due to our incomplete understanding of the underlying biological logic and mechanisms. This paper will describe a treatment approach from a biological algorithmic standpoint, encompassing biological logic programs, molecules that carry biological logic (Logical Biomolecular Complexes - LBC), and the existing tools that can be used to design such treatment programs. Our proposal is based on a review of oncolytic virus studies, but the logic framework behind LBCs is tailored specifically for cancer treatment, rather than focusing on replication and spreading. This sets LBCs apart from their viral counterparts and can be considered a new concept.
Authors - Julian Andres Duarte Suarez, Leonardo Juan Ramirez Lopez Abstract - Given the continuous need of beds available for hospitalization in health institutions, and even more so during pandemic periods, it is necessary to have alternatives that allow patients to be transferred to their homes and from there to carry out continuous monitoring of their health. For this, DATALOG was developed, which is an Internet of Medical Things platform that acquires, processes, transmits, stores and manages the medical signals of patients from their home to a central hospital. Following this, the stored data is processed by applying the Standard Intersectoral Process methodology for data mining that al-lows them to provide the medical staff with the behavior of six physiological variables and visualize it in a unique control table developed in php. Initial tests show an acceptance of the medical staff as a service and support tool for medical decisions, and, in addition, it has been proven that hospitalization at home con-tributes significantly to the rapid improvement of the patient, thus decongesting the hospital system. Among the most recent innovations, DATALOG includes an early warning system that allows to warn about the patient's condition between normal, attention, alert and critical, with the possibility of sending the alert to mobile systems. It is concluded that DATALOG is a useful service and tool for e-health, and machine learning techniques allow for the prediction of patient states and alerts.
Authors - K P N S Dayarathne, U Thayasivam Abstract - Deep learning has achieved amazing success in multiple areas, such as Image Classification, speech recognition, Object Detection & Segmentation, natural language processing, audio-visual recognition, adaptive testing, etc., and is gaining major interest from the research community. The application for deep learning is growing day by day. Predicting interest rate as univariate analysis is important given that total spectrum of the interest rates is not available to apply yield curve analysis. This paper investigates the applicability of deep learning models such as RNN, LSTM, CNN and TCN to interest rates in Asina frontier countries such as Sri Lanka, Pakistan and Bangladesh. The deep learning approach for interest rate perdition is still under the radar, and this is the first attempt on the Asian Fronter market. Interest rates associates with Government securities were considered to have uniqueness for all three countries, where data range from 2010-2022 for Sri Lanka and Pakistan, whereas Bangladesh analysis was based on the same from 2015-2022. The results revealed CNN was the best model for Sri Lanka and Bangladesh, while LSTM was the best model for Pakistan based on the lowest RMSE. The study further investigates the applicability of different activation function for output layers and hidden layers, but found ReLU is the most viable activation function along with Max pooling. Further, it was found that CNN works better for countries with stable term structures of interest rates, and the immediate dynamics of the interest rate influence the near future interest rate.
Authors - Taeko Onodera, Koutaro Hachiya, Yuhei Hatakenaka Abstract - Numerous studies have applied machine learning to diagnosis and screening in the medical and welfare fields. However, it is rare for the resulting machine learning models to be widely adopted in clinical practice. This study proposes a method for deriving diagnostic rules from machine learning models that can be applied manually without the use of computers. The proposed method involves inputting all possible patterns into a trained model, generating a truth table with the corresponding prediction results, and then using the Quine– McCluskey method to derive logical expressions that serve as manual diagnostic rules. In the experiments, the proposed method was compared with conventional methods for deriving manual diagnostic rules from datasets: the point score system, a method based on likelihood ratios, and a logic derivation method based on rough set theory. Only the proposed method achieved a positive clinical utility index of 0.81 or higher—classified as “excellent”—even when the number of rules was limited to just two or three.
Authors - El Hadji Bassirou TOURE, Ibrahima FALL, Mandicou BA, Alassane BAH Abstract - Microservice architectures offer scalability and deployment benefits but introduce significant data consistency challenges due to distributed data ownership and the necessity of data duplication across services. Current approaches either compromise service autonomy with strong consistency mechanisms or shift complexity to developers. This paper presents Micro-Flex, a novel megamodel-based consistency management framework that treats data entities as component models with formally tracked consistency states.We extend the Modified-Shared-Invalid protocol with differentiated Shared states (Shared+ and Shared−) to accommodate varying consistency requirements while maintaining formal guarantees. Our approach formalizes Global Operation Models (GOMs) as application specifications where consistency states are actively managed through well-defined transition rules. We validate Micro-Flex through an e-commerce case study, demonstrating its effectiveness in balancing consistency guarantees with service autonomy while addressing data duplication challenges. 4 By applying Model-Driven Engineering principles to distributed consistency challenges, our framework contributes to more disciplined data management in microservice architectures.
Authors - Nikolaos Lazaropoulos, Ioannis Vaxevanakis, Ioannis Sigalas, Ioannis Lamprou, Vassilis Zissimopoulos Abstract - Cloud Native Functions (CNFs) support automated and dynamic orchestration of containerized network services, replacing traditional hardware-based architectures. These deployments consist of modular microservices that enable elastic scalability and collaborative service delivery. This paper presents an approximation framework for capacity constrained CNF resource allocation, modeled as variants of the Group Generalized Assignment Problem (Group GAP). The main contributions are: (1) a 1 2 -approximation algorithm for CNF placement when each function’s footprint is at most half the cluster capacity and (2) a 1 2 (1 − e−1/d)-approximation for shared microservices among multiple CNFs, where d is the degree of sharing, supported by experimental evaluation of the algorithm relative error.
Authors - Abdulaziz Alkhajeh, Sara Alhashmi, Alya Al Ali, Rakan Alhosani, Suhail Alshehhi, Deepa Pavithran, Joseph Anajemba Abstract - Healthcare has always been a crucial part of human life, with people investing resources to get the best services available. Ensuring patient confidentiality has always been crucial, but the digital era introduces new security risks. Hospitals now store patient information in computerized databases, which are vulnerable to cyberattacks. One major threat is ransomware attacks, where hackers capture sensitive and confidential patient data and demand large sums of money to prevent it from being leaked or sold. This puts patient privacy at risk and can disrupt healthcare services. Also, unauthorized access to the patients’ information compromising the data confidentiality has been a growing concern because health care has always been sensitive and personal information that should not be utilized for commercial purposes. Blockchain technology offers a solution by providing a secure way to store patient files. Using an Interplanetary File System (IPFS) on the blockchain, healthcare providers can save patient records in a decentralized and protected system, reducing the risks linked to traditional databases. This method helps protect patient information from cyber threats, ensuring privacy and security. In this paper, we are using blockchain-based architecture coupled with pinata IPFS cloud to secure the patient’s valuable information from any kind of cyber-attack, including ransomware.
Authors - Shreya Joshi, Vijay Ukani, Priyank Thakkar, Mrudangi Thakker, Dhruvang Thakker Abstract - This study introduces a hybrid malware detection approach that combines machine learning (ML) and deep learning (DL) techniques to enhance detection accuracy. By applying models like K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Gradient Boosting, we achieved 99.47% accuracy during training and 99.21% accuracy during testing, focusing on analyzing Portable Executable (PE) header data. Additionally, incorporating Convolutional Neural Networks (CNN) with Long Short-Term Memory (LSTM) networks improved performance, achieving 99% accuracy, 97% precision, and 98% recall after 30 epochs. The proposed hybrid method reduces false positives and negatives while demonstrating scalability across various datasets, offering a reliable and efficient solution for contemporary malware detection.
Authors - Balvinder Pal Singh, Thangaraju B Abstract - Modern computing systems face the dual challenge of meeting escalating performance demands, driven especially by AI and ML workloads, while operating under stringent thermal constraints. As energy consumption continues to rise, conventional thermal mitigation techniques like frequency throttling often compromise performance and increase cooling costs, threatening long-term sustainability. This paper proposes a multi-level, software-centric approach to address this challenge through intelligent, temperature-aware scheduling. Leveraging evolutionary techniques, specifically Genetic Algorithm (GA), the proposed model reorders jobs at the OS scheduler level based on thermal impact and energy profiles. A secondary optimization phase further fine-tunes job execution using dynamic slice adjustment for thermally intensive tasks. Simulation results obtained using custom-integrated simulation framework leveraging GEM5, McPAT, and Hotspot tools, demonstrate 28% overall performance improvement, 53% reduction in thermal violations and a 15% decrease in energy consumption with 80% of the tasks executed without performance degradation. This approach was validated using representative benchmark workloads, optimizing both energy and temperature profiles. This AI-augmented, multi-level scheduling strategy significantly enhances thermal efficiency and performance, offering a scalable solution for next-generation high-performance computing environments.
Authors - Kostiantyn Hrishchenko, Oleksii Pysarchuk, Danylo Baran Abstract - This paper introduces a batch processing method for constraint programming to improve solution performance. The method is demonstrated in an example of a production scheduling problem. Transformations from mass production of standardized products toward small-scale, customized orders in the manufacturing sector introduce a new challenge of handling extensive input data. Modern production scheduling systems struggle to handle BigData loads caused by the limitations of state-of-the-art scheduling algorithms. Therefore, the development of highly adaptive algorithms and models capable of efficiently managing numerous unique orders while maintaining the ability to adapt to dynamic constraints and objectives becomes critically important. The baseline discrete constraint programming approach is chosen for its flexibility and extensibility, allowing it to model various realistic manufacturing scenarios. The proposed method splits large input into smaller subsets, each scheduled independently by repeatedly involving the constraint solver in different portions of the input data, significantly improving performance compared to allocating the whole input in a single step. Computational experiments with Google's OR-Tools CP-SAT solver evaluated the method's effectiveness. Time and memory usage reductions were shown. The proposed method demonstrates the possibility of solving problems of much larger size using the same constraint model and solver. It combines the advantages of the greedy algorithm and the exact integer programming approach.
Authors - V. Machaca, J. Grados, K. Lazarte, R. Escobedo, C. Lopez Abstract - The interaction between peptides and the Major Histocompatibility Complex (MHC) is a critical factor in the immune response against various threats. In this work, we fine-tuned protein language models like TAPE, ProtBert-BFD, ESM2(t6), ESM2(t12), ESM2(t30), and ESM2(t33) by adding a BiLSTM block in cascade for the task of peptide-MHC class-I binding prediction. Additionally, we addressed the vanishing gradient problem by employing LoRA, distillation, hyperparameter guidelines, and a layer freezing methodology. After experimentation, we found that TAPE and a distilled version of ESM2(t33) achieved the best results outperforming state-of-the-art tools such as NetMHCpan4.1, MHC urry2.0, Anthem, ACME, and MixMHCpred2.2 in terms of AUC, accuracy, recall, F1 score, and MCC.
Authors - Nafiz Eashrak, Mohammad Ikbal Hossain, Md Omum Siddique Auyon, Md Abdullah Al Adnan Abstract - The proliferation of cryptocurrencies and blockchain technology has significantly reshaped the financial sector, introducing decentralized and transparent digital transactions. Despite substantial advantages such as reduced transaction costs, enhanced transparency, and financial inclusion, the anonymous and decentralized nature of cryptocurrencies has also facilitated illicit financial activities, including fraud, money laundering, and tax evasion. This literature review systematically examines forensic methodologies, regulatory challenges, and theoretical frameworks relevant to cryptocurrency investigations. This study highlights advancements in blockchain analytics, AI-driven monitoring tools, and regulatory frameworks such as the FATF Travel Rule and EU MiCA. However, it also underscores persistent challenges posed by privacy-focused technologies, decentralized finance (DeFi), cross-border jurisdictional inconsistencies, and technical limitations in forensic methodologies. This paper proposes an integrated forensic framework incorporating AI analytics, international regulatory collaboration, specialized forensic training, and privacy-preserving investigative techniques. Through this comprehensive review, it provides critical insights for enhancing forensic investigation capabilities, regulatory compliance, and policymaking, while outlining future research opportunities addressing emerging threats in cryptocurrency-based financial systems.
Authors - Victor Sineglazov, Alexander Ruchkin Abstract - The article proposes a new combined approach to investigation the problem of classification on real-world noisy dataset using a multi-stage semi-supervised learning method. The main idea of this approach is based on combining two methods: self-supervised learning on unlabeled data using Contrastive Loss Nets and semi-supervised learning label propagation using an enhanced Poisson Seidel learning technique. The proposed approach offers significant advantages, as it allows for preliminary classification without labels, strengthening the distinctions between classes, and then using a minimal amount of labeled data for final classification. This is demonstrated through the analysis of synthetic data from different intersecting ”Two moons” and real medical dataset on heart disease - ”Cardio Vascular” Accuracy in the first case exceeds 82%, and for the second example - 73%, which is one of the best result on the Kaggle database when compared to any other known 20 methods.
Authors - Amir Aieb, Alexander Jacob, Antonio Liotta, Muhammad Azfar Yaqub Abstract - Predicting soil moisture under dynamic climate conditions is challenging due to intricate dependencies within the data. This study presents a surrogate deep learning (SDL) model with a multitask learning (MTL) approach to improve daily soil moisture predictions across spatiotemporal scales. The model employs a two-level encoding process, first compressing climate parameters into a single feature and then applying sequential encoding to capture long-term temporal patterns within a one-year timeframe for better generalization. Seasonality detection using autocorrelation facilitates data resampling into homogeneous samples, enhancing the SDL model by optimizing hyperparameters through efficient weight sharing between layers. To evaluate the effectiveness of MTL, three different SDL architectures such as LSTM, ConvLSTM, and BiLSTM were implemented for a comprehensive analysis. All models struggle with soil moisture prediction, particularly during dry periods, where LSTM experiences the most significant accuracy drop. While BiLSTM demonstrates better performance, its effectiveness remains constrained. However, integrating MTL enhances model stability and spatio-temporal accuracy, reducing errors across various conditions and achieving a 10% improvement due to better data representation, enabling SDL models to capture regional heterogeneity more effectively.
Authors - V. Machaca, D. Lopez, J. Mamani, S. Ramos, Y. Tupac Abstract - Cancer immunotherapy offers a promising option to conventional cancer therapies, in this field, neoantigen detection is a rapidly evolving field; however, as it requires multiple bioinformatics stages, such as quality control, alignment, variant calling, annotation, and neoantigen prioritization. Each stage depends on specific software tools, which can create technical and compatibility challenges, often requiring significant expertise to integrate and manage. To address these challenges, we introduce NeoArgosTools, a novel flowchart-based platform, with an intuitive graphical interface, designed to simplify and streamline neoantigen detection pipelines in cancer immunotherapy research.
