Scalable Analysis Research Articles

Automating Information Retrieval from Biodiversity Literature Using Large Language Models: A Case Study

Recently, Large Language Models (LLMs) have transformed information retrieval, becoming widely adopted across various domains due to their ability to process extensive textual data and generate diverse insights. Biodiversity literature, with its broad range of topics, is no exception to this trend (Boyko et al. 2023, Castro et al. 2024). LLMs can help in information extraction and synthesis, text annotation and classification, and many other natural language processing tasks. We leverage LLMs to automate the information retrieval task from biodiversity publications, building upon data sourced from our previous work (Ahmed et al. 2024). In our previous work (Ahmed et al. 2023, Ahmed et al. 2024), we assessed the reproducibility of deep learning (DL) methods used in biodiversity research. We developed a manual pipeline to extract key information on DL pipelines—dataset, source code, open-source frameworks, model architecture, hyperparameters, software and hardware specs, randomness, averaging result and evaluation metrics from 61 publications (Ahmed et al. 2024). While this allowed analysis, it required extensive manual effort by domain experts, limiting scalability. To address this, we propose an automatic information extraction pipeline using LLMs with the Retrieval Augmented Generation (RAG) technique. RAG combines the retrieval of relevant documents with the generative capabilities of LLMs to enhance the quality and relevance of the extracted information. We employed an open-source LLM, Hugging Face implementation of Mixtral 8x7B (Jiang et al. 2024), a mixture of expert models in our pipeline (Fig. 1) and adapted the RAG pipeline from earlier work (Kommineni et al. 2024). The pipeline was run on a single NVIDIA A100 40GB graphics processing unit with 4-bit quantization. To evaluate our pipeline, we compared the expert-assisted manual approach with the LLM-assisted automatic approach. We measured their consistency using the inter-annotator agreement (IAA) and quantified it with the Cohen Kappa score (Pedregosa et al. 2011), where a higher score indicates more reliable and aligned outputs (1: maximum agreement, -1: no agreement). The Kappa score among human experts (annotators 1 and 2) was 0.54 (moderate agreement), while the scores comparing human experts with the LLM were 0.16 and 0.12 (slight agreement). The difference is partly due to human annotators having access to more information (including code, dataset, figures, tables and supplementary materials) than the LLM, which was restricted to the text itself. Given these restrictions, the results are promising but also show the potential to improve them by adding further modalities to the LLM inputs. Future work will involve several key improvements to our LLM-assisted information retrieval pipeline: Incorporating multimodal data (e.g., figures, tables, code, etc.) as input to the LLM, alongside text, to enhance the accuracy and comprehensiveness of the information retrieved from publications. Optimizing the retrieval component of the RAG framework with advanced techniques like semantic search, hybrid search or relevance feedback can improve the quality of outputs. Expanding the evaluation to a larger corpus of biodiversity literature could provide a more comprehensive understanding of pipeline capabilities, and this paves the way for pipeline optimization. A human-in-the-loop approach for evaluating the LLM-generated outputs by matching the ground truth values from the respective publications, will increase the quality of the overall pipeline. Employing more metrics for the evaluation beyond the Cohen Kappa score to better understand the LLM-assisted outputs. Incorporating multimodal data (e.g., figures, tables, code, etc.) as input to the LLM, alongside text, to enhance the accuracy and comprehensiveness of the information retrieved from publications. Optimizing the retrieval component of the RAG framework with advanced techniques like semantic search, hybrid search or relevance feedback can improve the quality of outputs. Expanding the evaluation to a larger corpus of biodiversity literature could provide a more comprehensive understanding of pipeline capabilities, and this paves the way for pipeline optimization. A human-in-the-loop approach for evaluating the LLM-generated outputs by matching the ground truth values from the respective publications, will increase the quality of the overall pipeline. Employing more metrics for the evaluation beyond the Cohen Kappa score to better understand the LLM-assisted outputs. Leveraging LLMs to automate information retrieval from biodiversity publications signifies a notable advancement in the scalable and efficient analysis of biodiversity literature. Initial results show promise, yet there is substantial potential for enhancement through the integration of multimodal data, optimized retrieval mechanisms, and comprehensive evaluation. By addressing these areas, we aim to improve the accuracy and utility of our pipeline, ultimately enabling broader and more in-depth analysis of biodiversity literature.

Optimising microplastics analysis for quantifying and identifying microplastic fibres in laundry wastewater

Current methods for measuring microplastic fibres (MPF) are cumbersome, time consuming and unscalable for routine high throughput analysis. This study reports a method for rapidly extracting, quantifying and analysing MPFs in laundry wastewater with several key improvements which vastly enhance overall efficiency and scalability of analysis. FT-IR surveying is employed as a preliminary step in analysis to quickly determine what polymers are present in a sample prior to fluorescence treatment. Using random quadrating, whole 25 mm filter membranes were surveyed in <30 min with high recovery rates. In industrial laundry wastewater samples, polyester was the most common MPF, however acrylic, nylon, cotton and rayon were all ubiquitous. The study also demonstrates that an excitation wavelength of 365 nm was optimal for fluorescing PET fibres like polyester which were stained with Nile Red, but not 495 nm, which is commonly used in microplastic analysis. Finally, a custom ImageJ macro was written to automatically enumerate and describe MPFs on filter membranes using just a single stitched fluorescence image. In just a few seconds, concentrations of up to 40,000 fibres/L were analysed in industrial laundry wastewater samples with a lower particle size limit of 20 μm. This study highlights the need for more optimised and scalable analysis workflows which maintain high levels of reliability and accuracy.

Quantitative assessment of human motion for health and rehabilitation: A novel fuzzy comprehensive evaluation approach

In the pursuit of advancing health and rehabilitation, the quintessence of human motion recognition technology has been underscored through its quantitative contributions to physical performance assessment. This research delineates the inception of a novel fuzzy comprehensive evaluation-based recognition method that stands at the forefront of such innovative endeavours. By synergistically fusing multi-sensor data and advanced classification algorithms, the proposed system offers a granular quantitative analysis with implications for health and fitness monitoring, particularly rehabilitation processes. Our methodological approach, grounded in the modal separation technique and Empirical Mode Decomposition (EMD), effectively distills the motion acceleration component from raw accelerometer data, facilitating the extraction of intricate motion patterns. Quantitative analysis revealed that our integrated framework significantly amplifies the accuracy of motion recognition, achieving an overall recognition rate of 90.03 %, markedly surpassing conventional methods, such as Support Vector Machines (SVM), Decision Trees (DT), and K-Nearest Neighbors (KNN), which hovered around 80 %. Moreover, the system demonstrated an unprecedented accuracy of 97 % in discerning minor left-right swaying motions, showcasing its robustness in evaluating subtle movement nuances—a paramount feature for rehabilitation and patient monitoring. This marked precision in motion recognition heralds a new paradigm in health assessment, enabling objective and scalable analysis pertinent to individualized therapeutic interventions. The experimental evaluation accentuates the system's adeptness at navigating the dichotomy between complex, intense motions and finer, subtler movements with a high fidelity rate. It substantiates the method's utility in delivering sophisticated, data-driven insights for rehabilitation trajectory monitoring.

ScParser: sparse representation learning for scalable single-cell RNA sequencing data analysis

The rapid rise in the availability and scale of scRNA-seq data needs scalable methods for integrative analysis. Though many methods for data integration have been developed, few focus on understanding the heterogeneous effects of biological conditions across different cell populations in integrative analysis. Our proposed scalable approach, scParser, models the heterogeneous effects from biological conditions, which unveils the key mechanisms by which gene expression contributes to phenotypes. Notably, the extended scParser pinpoints biological processes in cell subpopulations that contribute to disease pathogenesis. scParser achieves favorable performance in cell clustering compared to state-of-the-art methods and has a broad and diverse applicability.

Scalable and Non-Destructive Analysis of Battery Degradation and Performance

Electrochemical processes occurring during battery operation are highly complex. Thermodynamic, kinetic and transport-based phenomena all play a significant role in the dynamic evolution of the time-based electrochemical response of any battery system. Researchers have long sought to quantify these various interconnected phenomena, but to-date elucidating the role these mechanisms play in the performance of battery systems has been challenging or impossible. To aid researchers in uncovering the mechanistic behavior of their systems, we have developed a novel testing protocol and corresponding data analysis method that is scalable and non-destructive (SAND). This protocol and analysis can be applied to any battery system, in any form factor, at any point in the battery life cycle; including R&D, manufacturing, and second life battery screening to name a few.The testing protocol is based on a series of galvanostatic current pulses at a variety of currents and/or times. By utilizing the natural differences in time-constant of the various electrochemical phenomena inside a battery, this technique can probe each mechanism to glean novel insight using standard battery cycling equipment. Specifically, the insights this technique can provide include: the electrochemical potential of each phase during lithiation and delithiation quantification of electrode kinetics at various states-of-charge (including exchange current density estimation from the Butler-Volmer model)quantitative estimation of transport-based polarization losses from both the electrodes and the electrolyte. This analysis provides an un-parallelled level of mechanistic insight into the inner workings of any battery, at any point, spanning small-scale R&D cells to large format cells in mass production.To demonstrate the capabilities of this technique this presentation will focus on the analysis of multiple cells (same model) from a Tier 1 supplier containing a lithium transition metal oxide cathode material cycled at various temperatures. These results have led to an extensive set of conclusions. For brevity a few of these conclusions are: The initial exchange current density of these cells is calculated to be 25.5 A/m2. The H2→H3 phase transition during charging is the least efficient phase, exhibiting the largest power loss during chargingThe inefficiencies in the H2→H3 phase transition are dominated by solid-state lithium atom diffusion within the electrodes. At low current densities the overpotential associated with transport is the dominant source of losses, while at higher currents kinetic overpotentials become the dominant source. In summary, we have developed a novel characterization technique requiring only standard battery cycling equipment that can be easily and quickly performed by researchers around the world. Whether on small-scale coin cells or large-format pouch cells, the analysis from this technique will give the battery community insight into the thermodynamic, kinetic, and transport properties of their system. Moreover, this technique can be applied in a variety of contexts related to battery development, including understanding degradation mechanisms, performing root-cause analysis of manufacturing defects, or providing more accurate parameters for computational modeling.

Personalized Recommendation Systems Powered By Large Language Models: Integrating Semantic Understanding and User Preferences

This study proposes a novel personalized recommendation system leveraging Large Language Models (LLMs) to integrate semantic understanding with user preferences [1]. The system addresses critical challenges in traditional recommendation approaches by harnessing LLMs' advanced natural language processing capabilities. We introduce a framework combining a fine-tuned Roberta semantic analysis model with a multi-modal user preference extraction mechanism.The LLM component undergoes domain adaptation using Masked Language Modeling on a corpus of 112,000 user reviews from the MyAnimeList dataset, followed by task-specific fine-tuning using contrastive learning. User preferences are modeled through a weighted combination of explicit ratings, review sentiment, and implicit feedback, incorporating temporal dynamics through a time-decay function. Experimental results demonstrate significant improvements over state-of-the-art baselines, including Matrix Factorization, Neural Collaborative Filtering, BERT4Rec, and LightGCN. Our LLM-powered system achieves an 8.6%increase in NDCG@10 and a 10.5% improvement in Mean Reciprocal Rank compared to the best-performing baseline. Ablation studies reveal the synergistic effect of integrating LLM-based semantic understanding with user preference modeling. Case studies highlight the system's ability to recommend long-tail items and provide cross-genre suggestions, showcasing its capacity for nuanced content understanding. Scalability analysis indicates that while the LLM-based approach has higher initial computational costs, its performance scales comparably to other deep learning models for larger datasets. This research contributes to the field by demonstrating the effectiveness of LLMs in enhancing recommendation accuracy and diversity. Future work will explore advanced LLM architectures, multi-modal data integration, and techniques to improve computational efficiency and interpretability of recommendations.

Improving Prenatal Detection of Congenital Heart Disease With a Scalable Composite Analysis of 6 Fetal Cardiac Ultrasound Biometrics

Improving Prenatal Detection of Congenital Heart Disease With a Scalable Composite Analysis of 6 Fetal Cardiac Ultrasound Biometrics

Deep sight: enhancing periprocedural adverse event recording in endoscopy by structuring text documentation with privacy-preserving large language models

Background and aimsThe assessment of adverse events from endoscopic procedures is essential for successful interventions, ensuring accurate follow-up, adverse event management and processing for quality control. Despite the critical need for structured documentation, the current practice often relies on free-text recordings, which poses challenges for scalable intervention analysis; however, the introduction of Large Language Models (LLMs) offers a promising solution by enabling the automatic extraction of adverse event details from procedural reports, without altering existing documentation practices. MethodsWe analyzed 672 endoscopy reports, using OpenAI’s GPT-4 and Llama-2 based models to structure the data in JSON for efficient analysis. We used an automated LLM-pipeline to extract adverse events such as bleeding, perforation, and aspiration. The dataset was divided into a proof of concept set (PoC-S) with n=171 reports, on which we explored prompt engineering to improve the performance of the models. The final analysis was run on an additional external test set of 501 reports (TS). ResultsGPT-4 showed high accuracy, with a sensitivity of 97% and specificity of 92% in the PoC-S and 91% and 96% respectively in the TS. GPT-4 use in real-world settings is limited by privacy concerns. Conversely, Llama-2 based models, especially the Llama-2 variants fine-tuned for German language, demonstrated comparable performance (PoC-S: sensitivity 94%, specificity 92%, CS: sensitivity 89%, specificity 93%) and offered a viable, privacy-compliant alternative. These models' effectiveness was further influenced by the method of prompt engineering, with experiments showing that the specificity and sensitivity could vary substantially based on the inclusion of specific prompt features, underscoring the importance of tailored prompt design. ConclusionApplying LLMs to extract structured medical information, particularly from endoscopy reports, offers an efficient, scalable, and adaptable documentation method that captures adverse events accurately with low error rate. It facilitates immediate quality reporting and reduces manual documentation efforts.

FInCH: FIJI plugin for automated and scalable whole-image analysis of protein expression and cell morphology

Study of morphogenesis and its regulation requires analytical tools that enable simultaneous assessment of processes operating at cellular level, such as synthesis of transcription factors (TF), with their effects at the tissue scale. Most current studies conduct histological, cellular and immunochemical (IHC) analyses in separate steps, introducing inevitable biases in finding and alignment of areas of interest at vastly distinct scales of organization, as well as image distortion associated with image repositioning or file modifications. These problems are particularly severe for longitudinal analyses of growing structures that change size and shape. Here we introduce a python-based application for automated and complete whole-slide measurement of expression of multiple TFs and associated cellular morphology. The plugin collects data at customizable scale from the cell-level to the entire structure, records each data point with positional information, accounts for ontogenetic transformation of structures and variation in slide positioning with scalable grid, and includes a customizable file manager that outputs collected data in association with full details of image classification (e.g., ontogenetic stage, population, IHC assay). We demonstrate the utility and accuracy of this application by automated measurement of morphology and associated expression of eight TFs for more than six million cells recorded with full positional information in beak tissues across 12 developmental stages and 25 study populations of a wild passerine bird. Our script is freely available as an open-source Fiji plugin and can be applied to IHC slides from any imaging platforms and transcriptional factors.

Corporate network anomaly detection methodology utilizing machine learning algorithms

ABSTRACT This study addresses the critical need for securing corporate networks against anomalies, a pressing concern in ensuring the comprehensive security of these networks. It aims to develop and validate a new machine learning-based methodology for anomaly detection that is adaptable across various corporate network environments, highlighting the method’s potential practical applications. Employing a systematic approach, the research integrates system analysis of anomaly detection methodologies with an analytical review of machine learning techniques tailored for high-security measures and attack prevention in corporate networks. This dual approach ensures a robust framework for identifying and addressing network anomalies efficiently. The methodology demonstrated notable efficacy, with the proposed machine learning-based anomaly detection techniques achieving an efficiency rate upwards of 90% in identifying and categorizing network traffic types. This high level of precision allows for the effective tracking of network anomalies across diverse corporate networks and their respective devices and equipment. The findings underscore the substantial practical value of the developed methodology, offering a promising avenue for enhancing corporate network security. The implementation of this machine learning-based approach not only facilitates the timely detection of anomalies but also significantly contributes to the improvement of machine learning applications within the realm of network security. Future research could further refine these techniques, exploring scalability and real-time data analysis enhancements to bolster their effectiveness across various network configurations.

Multi-modal generative modeling for joint analysis of single-cell T cell receptor and gene expression data

Recent advances in single-cell immune profiling have enabled the simultaneous measurement of transcriptome and T cell receptor (TCR) sequences, offering great potential for studying immune responses at the cellular level. However, integrating these diverse modalities across datasets is challenging due to their unique data characteristics and technical variations. Here, to address this, we develop the multimodal generative model mvTCR to fuse modality-specific information across transcriptome and TCR into a shared representation. Our analysis demonstrates the added value of multimodal over unimodal approaches to capture antigen specificity. Notably, we use mvTCR to distinguish T cell subpopulations binding to SARS-CoV-2 antigens from bystander cells. Furthermore, when combined with reference mapping approaches, mvTCR can map newly generated datasets to extensive T cell references, facilitating knowledge transfer. In summary, we envision mvTCR to enable a scalable analysis of multimodal immune profiling data and advance our understanding of immune responses.

Visibility graph-based covariance functions for scalable spatial analysis in non-convex partially Euclidean domains.

We present a new method for constructing valid covariance functions of Gaussian processes for spatial analysis in irregular, non-convex domains such as bodies of water. Standard covariance functions based on geodesic distances are not guaranteed to be positive definite on such domains, while existing non-Euclidean approaches fail to respect the partially Euclidean nature of these domains where the geodesic distance agrees with the Euclidean distances for some pairs of points. Using a visibility graph on the domain, we propose a class of covariance functions that preserve Euclidean-based covariances between points that are connected in the domain while incorporating the non-convex geometry of the domain via conditional independence relationships. We show that the proposed method preserves the partially Euclidean nature of the intrinsic geometry on the domain while maintaining validity (positive definiteness) and marginal stationarity of the covariance function over the entire parameter space, properties which are not always fulfilled by existing approaches to construct covariance functions on non-convex domains. We provide useful approximations to improve computational efficiency, resulting in a scalable algorithm. We compare the performance of our method with those of competing state-of-the-art methods using simulation studies on synthetic non-convex domains. The method is applied to data regarding acidity levels in the Chesapeake Bay, showing its potential for ecological monitoring in real-world spatial applications on irregular domains.

A robust CETSA data analysis automation workflow for routine screening

The Cellular Thermal Shift Assay (CETSA) enables the study of protein-ligand interactions in a cellular context. It provides valuable information on the binding affinity and specificity of both small and large molecule ligands in a relevant physiological context, hence forming a unique tool in drug discovery. Though high-throughput lab protocols exist for scaling up CETSA, subsequent data analysis and quality control remain laborious and limit experimental throughput. Here, we introduce a scalable and robust data analysis workflow which allows integration of CETSA into routine high throughput screening (HT-CETSA). This new workflow automates data analysis and incorporates quality control (QC), including outlier detection, sample and plate QC, and result triage. We describe the workflow and show its robustness against typical experimental artifacts, show scaling effects, and discuss the impact of data analysis automation by eliminating manual data processing steps.

Custom Biomedical FAIR Data Analysis in the Cloud Using CAVATICA.

The historically fragmented biomedical data ecosystem has moved towards harmonization under the findable, accessible, interoperable, and reusable (FAIR) data principles, creating more opportunities for cloud-based research. This shift is especially opportune for scientists across diverse domains interested in implementing creative, nonstandard computational analytic pipelines on large and varied datasets. However, executing custom cloud analyses may present difficulties, particularly for investigators lacking advanced computational expertise. Here, we present an accessible, streamlined approach for the cloud compute platform CAVATICA that offers a solution. We outline how we developed a custom workflow in the cloud, for analyzing whole genome sequences of case-parent trios to detect sex-specific genetic effects on orofacial cleft risk, which required several programming languages and custom software packages. The approach involves just three components: Docker to containerize software environments, tool creation for each analysis step, and a visual workflow editor to weave the tools into a Common Workflow Language (CWL) pipeline. Our approach should be accessible to any investigator with basic computational skills, is readily extended to implement any scalable high-throughput biomedical data analysis in the cloud, and is applicable to other commonly used compute platforms such as BioData Catalyst. We believe our approach empowers versatile data reuse and promotes accelerated biomedical discovery in a time of substantial FAIR data.

Multiscale Equation-Oriented Optimization Decreases the Carbon Intensity of Shale Gas to Liquid Fuel Processes.

Shale gas is revolutionizing the U.S. energy and chemical commodity landscape and can ease the transition to a sustainable decarbonized economy. This work develops an equation-oriented (EO) multiscale modeling framework using the open-source IDAES-PSE platform that tractably incorporates microkinetic detail in process design via reduced-order kinetic (ROK) models. Using multiobjective optimization with embedded heat integration and life-cycle analysis, we simultaneously minimize the minimum selling price of liquid hydrocarbons (e.g., liquid fuels/additives from shale gas) and process emissions (via a CO2 tax). Optimization reduces greenhouse gas emissions per MJ of fuel produced by over 35% compared to the literature and achieves a carbon efficiency of 87%. The optimizer changes the recycling rate, temperatures, and pressures to mitigate the effect of ROK model-form uncertainty on product portfolio predictions. Moreover, we show that the optimal process design is insensitive to changing CO2 tax rates. Finally, the EO framework enables a fast sensitivity analysis of shale gas composition variability across 12 regions of the Eagle Ford basin. These results highlight the benefits of the open-source EO framework: fast, scalable, customized, and reproducible system analysis and optimization for sustainable energy technologies beyond shale utilization.

INFLUTRUST: Trust-Based Influencer Marketing Campaigns in Online Social Networks

This paper introduces the INFLUTRUST framework that is designed to address challenges in trust-based influencer marketing campaigns on Online Social Networks (OSNs). The INFLUTRUST framework enables the influencers to autonomously select products across the OSN platforms for advertisement by employing a reinforcement learning algorithm. The Stochastic Learning Automata reinforcement algorithm considers the OSN platforms’ provided monetary rewards, the influencers’ advertising profit, and the influencers’ trust levels towards the OSN platforms to enable the influencers to autonomously select an OSN platform. The trust model for the influencers incorporates direct and indirect trust, which are derived from past interactions and social ties among the influencers and the OSN platforms, respectively. The OSN platforms allocate rewards through a multilateral bargaining model that supports competition among the influencers. Simulation-based results validate the INFLUTRUST framework’s effectiveness across diverse scenarios, with the scalability analysis demonstrating its robustness. Comparative evaluations highlight the INFLUTRUST framework’s superiority in considering trust levels and reward allocation fairness, benefiting both the influencers and the OSN platforms.

S-LDM: Server local dynamic map for 5G-based centralized enhanced collective perception

The automotive field is undergoing significant technological advances, which includes making the next generation of autonomous vehicles smarter, greener and safer through vehicular networks, which are often referred to as Vehicle-to-Everything (V2X) communications. Together with V2X, centralized maneuver management services for autonomous vehicles are increasingly gaining importance, as, thanks to their complete view over the road, they can optimally manage even the most complex maneuvers targeting L4 driving and beyond. These services face the challenge of strictly requiring a high reliability and low latency, which are tackled with the deployment at orchestrated Multi-Access Edge Computing (MEC) platforms. In order to properly manage safety-critical maneuvers, these services need to receive a large amount of data from vehicles, even though the useful subset of data is often related to a specific context on the road (e.g., to specific road users or geographical areas). Decoding and post-processing a large amount of raw messages, which are then for the most part filtered, increases the load on safety-critical services, which should instead focus on meeting the deadlines for the actual control and management operations. On this basis, we present an innovative open-source, 5G & MEC enabled service, called Server Local Dynamic Map (S-LDM). The S-LDM is a service that collects information about vehicles and other non-connected road objects using standard-compliant messages. Its primary purpose is to create a centralized dynamic map of the road that can be shared efficiently with other services managing L4 automation, when needed. By doing so, the S-LDM enables these services to widely and precisely understand the current situation of sections of the road, offloading them from the need of quickly processing a large number of messages. After a detailed description of the service architecture, we validate it through extensive laboratory and pilot trials, involving the MEC platforms and production 5G networks of three major European network operations and two Stellantis vehicles equipped with V2X On-Board Units (OBUs). We show how it can efficiently handle high update rates and process each messages in less than few tenths of microseconds. We also provide a complete scalability analysis with details on deployment options, providing insights on where new instances should be created in practical 5G-based V2X scenarios.

Efficient Static Vulnerability Analysis for JavaScript with Multiversion Dependency Graphs

While static analysis tools that rely on Code Property Graphs (CPGs) to detect security vulnerabilities have proven effective, deciding how much information to include in the graphs remains a challenge. Including less information can lead to a more scalable analysis but at the cost of reduced effectiveness in identifying vulnerability patterns, potentially resulting in classification errors. Conversely, more information in the graph allows for a more effective analysis but may affect scalability. For example, scalability issues have been recently highlighted in ODGen, the state-of-the-art CPG-based tool for detecting Node.js vulnerabilities. This paper examines a new point in the design space of CPGs for JavaScript vulnerability detection. We introduce the Multiversion Dependency Graph (MDG), a novel graph-based data structure that captures the state evolution of objects and their properties during program execution. Compared to the graphs used by ODGen, MDGs are significantly simpler without losing key information needed for vulnerability detection. We implemented Graph.js, a new MDG-based static vulnerability scanner specialized in analyzing npm packages and detecting taint-style and prototype pollution vulnerabilities. Our evaluation shows that Graph.js outperforms ODGen by significantly reducing both the false negatives and the analysis time. Additionally, we have identified 49 previously undiscovered vulnerabilities in npm packages.

Nonlinear optimization of optical camera multiparameter via triple integrated Gradient-based optimizer algorithm

Metaheuristic techniques are widely used to solve various complex real-world problems. Camera calibration is a critical task in machine vision, which aims to determine the optical camera parameters through 2D image features and corresponding 3D spatial features. To overcome the limitations of traditional methods, such as weak local optima avoidance and low calibration accuracy, a triple integrated Gradient-based optimizer (TIGBO_CC) is proposed for nonlinear optimization of optical camera parameters. Several improvements are made to the original Gradient-based optimizer (GBO). First, in the exploration phase, a fitness-distance balance is adopted to select the most contributing candidate to guide the update of the solution. Second, a random search technique is adopted in the exploitation phase. The switching between the original mode and the additional mode follows the probability. Finally, a neighborhood local search is also introduced. Small-step fine search is performed near the optimal solution to improve the quality of the final solution. TIGBO_CC is evaluated on 12 classical functions, the CEC2020 test suite, and the CEC2022 test suite. The experiments include sensitivity analysis, effectiveness analysis, scalability analysis, function value, convergence curve, box plot, Friedman test, Wilcoxon test, etc. Furthermore, a nonlinear camera calibration model is established. Taking the mean reprojection error as the objective function, TIGBO_CC solves the optimal optical camera parameters by minimizing this function. In different scenarios, the calibration method based on TIGBO_CC has high precision and strong reproducibility. Comprehensive analysis shows that TIGBO_CC has excellent optimization ability and practicality.

Making sense of student feedback and engagement using artificial intelligence

Making sense of student feedback and engagement is important for informing pedagogical decision-making and broader strategies related to student retention and success in higher education courses. Although learning analytics and other strategies are employed within courses to understand student engagement, the interpretation of data for larger data sets is more challenging and rarely pursued. This is concerning as data offers the potential for critical insights into engagement behaviour and the value students place on engagement. Artificial intelligence (AI) offers a revolutionary ability to make sense of data, with capacity for prediction and classification, by consuming vast amounts of structured and unstructured data sets. This paper reports on how AI methodologies (specifically, deep learning and natural language processing) were used to leverage labelled student feedback in terms of online engagement in five courses in a regional Australian university. This paper reinforces the value of AI as a viable and scalable multilayered analysis tool for analysing and interpreting student feedback, particularly for categorising student responses as to the types of engagement that they most valued to support their learning. The paper concludes with a discussion of suggested further refinement, including how the AI-derived data may add insights for informing pedagogical practice. Implications for practice or policy: AI offers an ability to make sense of large data sets in higher education courses. Teachers can use student feedback data categorised into types of engagement by AI to support reflection on what students value in their courses. Educators and key stakeholders can use the insights AI analysed data offers for informing pedagogical practice and decision-making in higher education to enhance student experiences.