Powerful Learning Research Articles

Rethreading the needle: A novel molecular index of soil health (MISH) using microbial functional genes to predict soil health management

Soil health relies on the actions and interactions of an abundant and diverse biological community. Current soil health assessments rely heavily on a suite of soil biological, chemical, and physical indicators, often excluding molecular information. Soil health is critical for sustainable agricultural production, and a comprehensive understanding of how microbial communities provide ecosystem services can help guide management practices. To explore the role of microbial function in soil health, 536 soil samples were collected from 26 U.S. states, representing 52 different crops and grazing lands, and analyzed for various soil health indicators. The bacterial functional profile was characterized using 16S ribosomal RNA gene sequencing paired with PICRUSt2 to predict metagenome functions. Functional data were used as predictors in eXtreme Gradient Boosting (XGBoost), a powerful machine learning algorithm, and enzymes important to soil health indicators were compiled into a Molecular Index of Soil Health (MISH). The overall MISH score significantly correlated with non-molecular measures of soil health and management practice adoption. Additionally, several new enzymes were identified as potential targets to better understand microbial mediation of soil health. This low-cost, DNA-based approach to measuring soil health is robust and generalizable across climates.

Prediction of Two Year Survival Following Elective Repair of Abdominal Aortic Aneurysms at A Single Centre Using A Random Forest Classification Algorithm

The decision to electively repair an abdominal aortic aneurysm (AAA) involves balancing the risk of rupture, periprocedural mortality, and life expectancy. Random forest classifiers (RFCs) are powerful machine learning algorithms. The aim of this study was to construct and validate a random forest machine learning tool to predict two year survival following elective AAA repair. All patients who underwent elective open or endovascular repair of AAA from 1 January 2008 to 31 March 2021 were reviewed. They were assessed using the Vascular Surgery Quality Improvement Program pathway involving cardiopulmonary exercise testing, contrast enhanced computerised tomography scan, and multidisciplinary assessment. Patients were followed up for at least two years. A RFC was developed using 70% of the dataset and validated using 30% to predict survival for at least two years following AAA repair. A total of 925 patients (n = 836 men; n = 89 women) underwent elective repair of AAA; 126 (13.6%) died during the first two years; 11 (1.2%) died from periprocedural mortality. Variable importance analysis suggested that anaerobic threshold, pre-operative haemoglobin, maximal O2 consumption, body mass index, risk category, and forced expiratory volume in 1 second - forced vital capacity ratio were the most important contributors to the model. Sensitivity and specificity of the RFC for prediction of two year survival following surgery was 96.7% (95% CI 94.4 - 99%) and 67.1% (95% CI 61 - 72%); overall accuracy: 92.6% (95% CI 88 - 95%) (positive predictive value: 0.93, negative predictive value: 0.80); 10-fold cross validation revealed area under the receiver operator characteristic curve of 0.88. RFCs based on readily available clinical data can successfully predict survival in the first two years following elective repair of AAA. Such information can contribute to the risk benefit assessment when deciding to electively repair AAAs.

The role of big data in driving advancements in deep learning: Opportunities, challenges, and future directions

The advent of big data has significantly shaped the trajectory of machine learning, particularly in the field of deep learning. The vast amounts of data generated every day across various sectors—from healthcare to finance—have provided unprecedented opportunities for training more powerful and accurate deep learning models. This review paper explores the critical role big data plays in driving advances in deep learning, analyzing how these two fields intersect and fuel each other. The paper also examines the challenges associated with leveraging big data in deep learning, including data quality, scalability, and computational constraints. Finally, the paper discusses future directions in the convergence of big data and deep learning, emphasizing emerging trends and the potential of this intersection to revolutionize industries.

RAG-based explainable prediction of road users behaviors for automated driving using knowledge graphs and large language models

The prediction of road user behaviors in the context of autonomous driving has attracted considerable attention from the scientific community in recent years. Most works focus on predicting behaviors based on kinematic information alone, a simplification of reality since road users are humans, and as such they are highly influenced by their surrounding context. In addition, a large plethora of research works rely on powerful Deep Learning techniques, which exhibit high-performance metrics in prediction tasks but may lack the ability to fully understand and exploit the contextual semantic information contained in the road scene, not to mention their inability to provide explainable predictions that can be understood by humans. In this work, we propose an explainable road users’ behavior prediction system that integrates the reasoning abilities of Knowledge Graphs (KG) and the expressiveness capabilities of Large Language Models (LLM) by using Retrieval Augmented Generation (RAG) techniques. For that purpose, Knowledge Graph Embeddings (KGE) and Bayesian inference are combined to allow the deployment of a fully inductive reasoning system that enables the issuing of predictions that rely on legacy information contained in the graph, as well as on current evidence gathered in real-time by onboard sensors. Two use cases have been implemented following the proposed approach: 1) Prediction of pedestrians’ crossing actions; and 2) Prediction of lane change maneuvers. In both cases, the performance attained exceeds the current state-of-the-art in terms of anticipation and F1 score, showing a promising avenue for future research in this field.

Unveiling the Secrets of the Mammogram: A Statistical Journey through Machine Learning Approach

Mammograms, once mere X-ray images, are now powerful allies in the fight against breast cancer. Armed with sophisticated statistical image classification techniques, these images reveal hidden clues that can lead to early detection. Beneath the surface of a mammogram lies a complex tapestry of patterns, each one a potential marker of malignancy. By decoding these intricate patterns, algorithms can identify abnormalities that might otherwise go unnoticed. The Mammographic Image Analysis Society (MIAS) has been at the forefront of this revolution, creating a vast database of digital mammograms. From this treasure trove, we've selected 322 images to put our statistical prowess to the test. Using a powerful machine learning tool known as Support Vector Machine (SVM), we've classified these images into seven categories: • Calcifications (CALC): Tiny, hard deposits. • Circumscribed (CIRC) masses: Well-defined lumps. • Speculated (SPIC) masses: Lumps with spiky edges. • Ill-defined (MISC) masses: Irregular, indistinct lumps. • Architectural distortions (ARCH): Changes in the breast's structure. • Asymmetry(s) (ASYM): Differences between the two breasts. • Normal (NORM): Healthy breast tissue. Our goal is to shed light on the intricate process of mammogram classification and share the insights gained from our analysis presented in the user-friendly language and the analysis is carried out using MATLAB.

Engineering Safety and Ethical Challenges in 2045 Artificial Intelligence Singularity

Artificial intelligence (AI) has rapidly advanced, increasingly showcasing its powerful learning and computational capabilities. This progress has resulted in significant breakthroughs in areas such as image processing, speech recognition, and autonomous driving. Scientists predict that by around 2045, AI will overcome existing technological barriers, allowing strong AI to surpass human intelligence. However, it will inevitably affect human social relationships and order. Ethical issues associated with AI technology, such as unemployment, privacy breaches, and discrimination, generate a sense of threat among people, resulting in a loss of confidence in AI, which hampers its sustainable progress. Therefore, AI ethical issues are not only significant topics in academia but also become critical concerns for individuals, society, and nations. This article aims to address the challenges of AI ethics safety and the erosion of human confidence, while promoting the sustainable development of AI. It presents an AI ethics safety framework that analyzes engineering ethics and human trust within the context of sustainable AI development, and it recommends governance methods and strategies informed by case studies. Furthermore, we propose evaluation criteria and methods, establishing early-warning thresholds to keep potential AI risks within acceptable limits. Finally, the future prospects for AI ethics safety are highlighted. We hope our research contributes to the sustainable development of AI, ensuring that the arrival of the AI singularity has a positive impact on society with a long-term harmonious coexistence between AI and humanity.

A Caching-based Framework for Scalable Temporal Graph Neural Network Training

Representation learning over dynamic graphs is critical for many real-world applications such as social network services and recommender systems. Temporal graph neural networks (T-GNNs) are powerful representation learning methods and have demonstrated remarkable effectiveness on continuous-time dynamic graphs. However, T-GNNs still suffer from high time complexity, which increases linearly with the number of timestamps and grows exponentially with the model depth, making them not scalable to large dynamic graphs. To address the limitations, we propose Orca , a novel framework that accelerates T-GNN training by caching and reusing intermediate embeddings. We design an optimal caching policy, named MRD , for the uniform cache replacement problem, where embeddings at different intermediate layers have identical dimensions and recomputation costs. MRD not only improves the efficiency of training T-GNNs by maximizing the number of cache hits but also reduces the approximation errors by avoiding keeping and reusing extremely stale embeddings. For the general cache replacement problem, where embeddings at different intermediate layers can have different dimensions and recomputation costs, we solve this NP-hard problem by presenting a novel two-stage framework with approximation guarantees on the achieved benefit of caching. Furthermore, we have developed profound theoretical analyses of the approximation errors introduced by reusing intermediate embeddings, providing a thorough understanding of the impact of our caching and reuse schemes on model outputs. We also offer rigorous convergence guarantees for model training, adding to the reliability and validity of our Orca framework. Extensive experiments have validated that Orca can obtain two orders of magnitude speedup over state-of-the-art T-GNNs while achieving higher precision on various dynamic graphs.

TorchSISSO: A PyTorch-based implementation of the sure independence screening and sparsifying operator for efficient and interpretable model discovery

Symbolic regression (SR) is a powerful machine learning approach that searches for both the structure and parameters of algebraic models, offering interpretable and compact representations of complex data. Unlike traditional regression methods, SR explores progressively complex feature spaces, which can uncover simple models that generalize well, even from small datasets. Among SR algorithms, the Sure Independence Screening and Sparsifying Operator (SISSO) has proven particularly effective in the natural sciences, helping to rediscover fundamental physical laws as well as discover new interpretable equations for materials property modeling. However, its widespread adoption has been limited by performance inefficiencies and the challenges posed by its FORTRAN-based implementation, especially in modern computing environments. In this work, we introduce TorchSISSO, a native Python implementation built in the PyTorch framework. TorchSISSO leverages GPU acceleration, easy integration, and extensibility, offering a significant speed-up and improved accuracy over the original. We demonstrate that TorchSISSO matches or exceeds the performance of the original SISSO across a range of tasks, while dramatically reducing computational time and improving accessibility for broader scientific applications.

Automated Forensic Examination of Virtual Assets Using XGBoost

The issues of identification of virtual assets and digital currencies have become more acute in recent years because of increased numbers of transactions, their increased volumes, and complexity in forensic examination. This paper is devoted to a new method of the automated forensic investigation of virtual assets with the help of a powerful machine learning algorithm called XGBoost. XGBoost has very high speed and it is quite useful when dealing with large amounts of data and used for finding patterns that may suggest unlawful operations. The given framework is based on XGBoost by identifying the transactions using their features activities, amount, frequencies, volumes and addresses. Through training the model that is given the historical data, the model is, therefore, able to separate normal and suspicious transactions and alert the decision-making process for further review. Network analysis integration takes the system capability to a new level by enabling the identification of intricate transaction characteristics and the interconnectedness of addressing. Aside from adding automation to the forensic examinations to make it more efficient and accurate, such a procedure assists in identifying the developments of virtual asset transactions. The results show the effectiveness of the XGBoost in supporting the advance of forensic investigators helping them to meet the growing roles that cybercriminals portray in today’s digital asset markets.

Music Sequence Creation Technique in Piano Music Production Based on Embedded Neural Network

With the rapid development of computer technology and artificial intelligence, especially the rise of embedded neural network technology. The traditional arrangement methods are facing increasing challenges in the field of music production, and their complexity and time-consuming nature are no longer sufficient to meet the urgent needs of modern audiences for efficient and high-quality computer piano music works. This paper proposes an innovative solution — a piano music production matching method based on embedded neural network fusion, and particularly focused on intelligent creation technology for music sequences. In the system, embedded neural networks are core applied in the process of creating music sequences. By utilizing their powerful feature learning and generalization abilities, neural networks can capture complex patterns and dynamic changes in music sequences, providing infinite possibilities for melody generation and innovation. At the same time, combined with association rule algorithms, the system can discover potential association patterns between notes, further optimizing the fluency and harmony of the melody; The clustering algorithm helps extract common features from massive music data, providing a rich material library for melody innovation.

PID control algorithm based on multistrategy enhanced dung beetle optimizer and back propagation neural network for DC motor control.

Traditional Proportional-Integral-Derivative (PID) control systems often encounter challenges related to nonlinearity and time-variability. Original dung beetle optimizer (DBO) offers fast convergence and strong local exploitation capabilities. However, they are limited by poor exploration capabilities, imbalance between exploration and exploitation phases, and insufficient precision in global search. This paper proposes a novel adaptive PID control algorithm based on enhanced dung beetle optimizer (EDBO) and back propagation neural network (BPNN). Firstly, the diversity of exploration is increased by incorporating a merit-oriented mechanism into the rolling behavior. Then, a sine learning factor is introduced to balance the global exploration and local exploitation capabilities. Additionally, a dynamic spiral search strategy and adaptive [Formula: see text]-distribution disturbance are presented to enhance search precision and global search capability. The BPNN is employed to fine-tune both PID and network parameters, leveraging its powerful generalization and learning ability to model nonlinear system dynamics. In the simplified motor experiments, the proposed controller achieved the lowest overshoot (0.5%) and the shortest response time (0.012s), with a settling time of 0.02s and a steady-state error of just 0.0010. In another set of experiments, the proposed controller recorded an overshoot and response time of 0.7% and 0.0010s, across five DC motor tests. These results demonstrate the proposed adaptive PID control algorithm has superior performance in optimizing control system parameters, as well as improving system robustness and stability.

Permutation-equivariant quantum convolutional neural networks

Abstract The Symmetric group S n manifests itself in large classes of quantum systems as the invariance of certain characteristics of a quantum state with respect to permuting the qubits. Subgroups of S n arise, among many other contexts, to describe label symmetry of classical images with respect to spatial transformations, such as reflection or rotation. Equipped with the formalism of geometric quantum machine learning, in this study we propose the architectures of equivariant quantum convolutional neural networks (EQCNNs) adherent to S n and its subgroups. We demonstrate that a careful choice of pixel-to-qubit embedding order can facilitate easy construction of EQCNNs for small subgroups of S n . Our novel EQCNN architecture corresponding to the full permutation group S n is built by applying all possible QCNNs with equal probability, which can also be conceptualized as a dropout strategy in quantum neural networks. For subgroups of S n , our numerical results using MNIST datasets show better classification accuracy than non-equivariant QCNNs. The S n -equivariant QCNN architecture shows significantly improved training and test performance than non-equivariant QCNN for classification of connected and non-connected graphs. When trained with sufficiently large number of data, the S n -equivariant QCNN shows better average performance compared to S n -equivariant QNN . These results contribute towards building powerful quantum machine learning architectures in permutation-symmetric systems.

Consistent positive correlation sample distribution: Alleviating the negative sample noise issue in contrastive adaptation

Contrastive learning has been widely applied in domain adaptation tasks due to its powerful representation learning capabilities in recent years. Constructing positive and negative sample pairs through different augmented views of instances promotes the effective application of contrastive learning in domain adaptation. However, most existing methods are based on Noise Contrastive Estimation (NCE), which primarily focuses on instance-level information and lacks attention to the relationships between different instances. These methods treat all negative samples as noise interference, leading to an increased distance between negative samples and anchor samples, disregarding instances in the negative samples that share the same category as the anchor sample. Consequently, these methods do not offer significant advantages in domain adaptation for image classification tasks. To address this issue, we propose a method called Consistent Positive Correlation Sample Distribution (CPCSD) to mitigate the problem of class collision among negative samples by leveraging the semantic similarity between instances. We introduce a positive correlation distribution loss within the contrastive adaptation framework to compute the similarity distribution between batches of augmented views and align them. Additionally, we sharpen the target similarity distribution to obtain a more emphasized relationship distribution, thereby alleviating the issue of negative sample noise in contrastive adaptation. Extensive experiments demonstrate that our proposed model has significant advantages in contrastive adaptation algorithms and improves the performance of downstream domain adaptation tasks.

Storytelling: A learning tool to enhance medical students' empathy, attentive listening, clinical curiosity and reflection.

Listening to stories of sickness lies at the heart of the patient-doctor relationship. Storytelling serves as a powerful learning tool fostering empathy, attentive listening, clinical curiosity and reflection, key elements of narrative competence. The patient's story, their history, forms the fundamental core of diagnosis, and their emotions are central to crafting appropriate treatment plans that address their true concerns. This study reviews the benefits of storytelling as a learning tool, particularly in enhancing empathy, attentive listening, clinical curiosity and reflection within medical education. It also describes a practical workshop that demonstrates how storytelling can be utilised to inspire students to develop narrative competence in a supportive learning environment. Since the stories draw on participants' clinical experience, in both primary and secondary care, the article aims to engage clinicians involved in medical education and the teaching of communication skills.

Random forest models reveal academic and financial factors outweigh demographics in predicting completion of a year-round veterinary program.

The purpose of this study was to develop random forest classifier models (a type of supervised machine learning algorithm) that could (1) predict students who will or will not complete the DVM degree requirements and (2) identify the top predictors for academic success and completion of the DVM degree. The study utilized Ross University School of Veterinary Medicine student records from 2013 to 2022. Twenty-four variables encompassing demographic (eg, age, race), academic (eg, grade point average), and financial aid (eg, outstanding balances) data were assessed in 11 cross-validated random forest machine learning models. One model was built assessing all years of data and 10 individual models were developed for each enrollment year to compare how the top predictors of success varied among the years. Consistently, only academic and financial factors were identified as being features of importance (predictors) in all models. Demographic factors such as race were not important for predicting student success. All models performed very well to excellently based on multiple performance metrics including accuracy, ranging from 96.1% to 99%, and the areas under the receiver operating characteristic curves, ranging from 98.1% to 99.9%. The random forest algorithm is a powerful machine learning prediction model that performs well with veterinary student academic records and is customizable such that variables important to each veterinary school's student population can be assessed. Identifying predictors of success as well as at-risk students is essential for providing targeted curricular interventions to increase retention and achieve timely completion of a DVM degree.

Quantum machine learning enhanced laser speckle analysis for precise speed prediction

Laser speckle contrast imaging (LSCI) is an optical technique used to assess blood flow perfusion by modeling changes in speckle intensity, but it is generally limited to qualitative analysis due to difficulties in absolute quantification. Three-dimensional convolutional neural networks (3D CNNs) enhance the quantitative performance of LSCI by excelling at extracting spatiotemporal features from speckle data. However, excessive downsampling techniques can lead to significant information loss. To address this, we propose a hybrid quantum–classical 3D CNN framework that leverages variational quantum algorithms (VQAs) to enhance the performance of classical models. The proposed framework employs variational quantum circuits (VQCs) to replace the 3D global pooling layer, enabling the model to utilize the complete 3D information extracted by the convolutional layers for feature integration, thereby enhancing velocity prediction performance. We perform cross-validation on experimental LSCI speckle data and demonstrate the superiority of the hybrid models over their classical counterparts in terms of prediction accuracy and learning stability. Furthermore, we evaluate the models on an unseen test set and observe that the hybrid models outperform the classical models with up to 14.8% improvement in mean squared error (MSE) and up to 26.1% improvement in mean absolute percentage error (MAPE) evaluation metrics. Finally, our qualitative analysis shows that the hybrid models offer substantial improvements over classical models in predicting blood flow at both low and high velocities. These results indicate that the hybrid models possess more powerful learning and generalization capabilities.

Revolutionizing Cardio Vascular Disease Diagnosis Through Lasso Regression Model

Cardiovascular diseases (CVDs) are among the most prevalent and fatal health conditions globally, necessitating early and accurate diagnosis to mitigate risk and enhance treatment outcomes. Traditional diagnostic methods often rely on a combination of clinical assessments and static risk models, which can be constrained by their inability to handle the complexity and interrelationships of multiple risk factors. In this research, we propose a novel approach to revolutionizing cardiovascular disease diagnosis through the application of the Lasso (Least Absolute Shrinkage and Selection Operator) regression model, a powerful machine learning technique for feature selection and predictive modeling. The Lasso regression model is particularly advantageous for high-dimensional medical datasets, where multiple clinical features may be correlated. By applying Lasso, we aim to address two critical challenges in cardiovascular disease prediction: identifying the most relevant predictors from large and complex datasets and reducing model over-fitting, which is common when working with a large number of co-variates. Our results demonstrated that the Lasso model significantly improves prediction accuracy when compared to traditional logistic regression and other machine learning models. Lasso regression has the potential to revolutionize cardiovascular disease diagnosis by providing a data-driven, efficient, and interpretable solution that bridges complex medical datasets with practical clinical decision-making. Keywords: Cardio vascular diseases (CVD), Machine learning models, Losso regression model.

Generative AI

The integration of Generative AI (GenAI) technology in higher education demands active and creative pedagogical responses. In this paper we report on an emerging creative educational design approach which introduces activities for working with GenAI in imaginative ways. Through multidisciplinary research, we are iteratively designing activities to engage educators and students in creative and critical inquiry of GenAI with metaphors. First, we introduce the background and rationale for using metaphor to explore GenAI in education. Our approach to generative AI draws on postdigital ideas that consider technology as entwined in our lives, rather than as separate tools. Put simply, GenAI is neither inherently good nor bad for education. Research on metaphor as a powerful learning tool of analogy is reviewed. A detailed account of workshop learning activities is then provided, including their design, implementation, and preliminary findings. The aim is to broaden approaches to using and integrating such technologies, and to serve as an alternative model for other educators to adapt for their own practices. Finally, we discuss future, creative possibilities with GenAI and metaphor.

SSML: Semi-supervised metric learning with hard samples for hyperspectral image classification

Deep learning is widely used in hyperspectral image (HSI) classification due to its powerful learning capabilities. However, its excellent performance typically requires a large number of samples, which can be time-consuming and labor-intensive to produce. The limitation of available samples greatly constrains the model's generalization performance. To alleviate the sample pressure, we propose a semi-supervised metric learning method for HSI classification that focuses on hard samples and can obtain reliable pseudo-labels through multiscale prediction. Additionally, we employ a hard sample learning strategy for network training, which concentrates on enhancing network discrimination by adaptively optimizing intra-class and inter-class distances. Performance tests on three public datasets indicate that the proposed method surpasses other state-of-the-art methods across multiple validation metrics.

Exploring Diversity and Abundance of Stingless Bee using Clustering Approach

Stingless bees are paramount in food chain as they are important pollinators of field crops. Recent studies revealed that these bees are seriously threatened by climate change and rapid urbanization across the world. It is thus important to study the relationship between the stingless bee’s diversity and the characteristics of the locations they inhibit. At the same time, clustering algorithms is a powerful machine learning approach in exploring unsupervised data. Consequently, this study aims to explore the stingless bee diversity in Malaysia through hierarchical, k-means and DBSCAN clustering. The dataset of this study consists of individual stingless bees collected from 12 locations. It comprises 14 environmental features, 3 physical characteristics, 35 species count, 12 genera counts and 3 diversity-and-abundance weights. A four-stage methodology is employed in the study. The results show that DBSCAN effectively groups data into clusters that are well-defined, but the results are less informative. In contrast, hierarchical and k-means clustering are found producing results that provide clearer insights, with hierarchical clustering delivering notably richer results.