Information leakage in financial machine learning research

Image-based plant disease detection is among the essential activities in precision agriculture for observing incidence and measuring the severity of variability in crops. 70% to 80% of the variabilities are attributed to diseases caused by pathogens, and 60% to 70% appear on the leaves in comparison to the stem and fruits. This work provides a comparative analysis through the model implementation of the two renowned machine learning models, the support vector machine (SVM) and deep learning (DL), for plant disease detection using leaf image data. Until recently, most of these image processing techniques had been, and some still are, exploiting what some considered as "shallow" machine learning architectures. The DL network is fast becoming the benchmark for research in the field of image recognition and pattern analysis. Regardless, there is a lack of studies concerning its application in plant leaves disease detection. Thus, both models have been implemented in this research on a large plant leaf disease image dataset using standard settings and in consideration of the three crucial factors of architecture, computational power, and amount of training data to compare the duos. Results obtained indicated scenarios by which each model best performs in this context, and within a particular domain of factors suggests improvements and which model would be more preferred. It is also envisaged that this research would provide meaningful insight into the critical current and future role of machine learning in food security

Alzheimer’s disease is a neurodegenerative disease that usually starts slow and progressively worsens. Early detection of Alzheimer’s can help us prevent deterioration of the tissue damage in the brain. In this paper, a detailed analysis on detection of Alzheimer’s disease using various Convolutional Neural Network (CNN) architectures in Machine Learning (ML) is shown. This disease causes people to suffer from memory loss and unusual brain tissue damages. Several ML models have been used by researchers for detection of Alzheimer’s. Analyzing Magnetic Resonance Imaging (MRI) scans is a very useful and common practice for the diagnosis of Alzheimer’s disease. ML is used to help the computers to train on the large and complicated datasets and detect the presence of Alzheimer’s disease in brain MRI. In today’s world Deep Learning techniques have successfully brought change in numerous fields including medical image analysis. We propose a deep CNN for Alzheimer’s disease using MRI scan data sets. Our model identifies Alzheimer’s disease in its different stages using different CNN architectures to obtain efficiency and accuracy for early-stage diagnosis of Alzheimer’s disease.

Accurate diagnosis of pathology from ultrasound images is reliant upon images of a suitable diagnostic quality being acquired. This study aimed to create a novel machine learning model to automatically assess transvaginal ultrasound (TVUS) image quality for gynaecological ultrasound. Six imaging professionals (two sonographers, two gynaecological sonologists and two radiologists) assigned a quality score to 150 TVUS images from 50 cases (50 uterus images and 100 ovary images). Images were given a score of 1-4 (1-reject/image inaccurate, 2-poor quality, 3-suboptimal quality or 4-optimal quality). As variation existed between the scores assigned by the labellers, we framed this problem as a multi-annotator noisy label problem. To address this, a new machine learning architecture was developed, combining a weighted ensemble algorithm to estimate consensus labels and a multi-axis vision transformer (MaxViT) to handle noisy labels, improving model accuracy in predicting image quality. Forty cases (120 images) were used for model training, while the remaining 10 cases (30 images) were reserved as a test set for model evaluation. The novel machine learning architecture we created was able to successfully determine image quality with a validation accuracy of 80% and a macro average recall of 77%. This significantly improved upon the 57% accuracy of the baseline machine learning method (ResNet50). The MaxViTs were able to outperform human performance in most cases, with an accuracy of 80% surpassing four of the six human labellers. This novel machine learning model offers an automated method of assessing the quality of TVUS images. The tool has the potential to provide real-time feedback to those performing TVUS, reduce the need for repeated imaging, and improve the diagnosis of gynaecological pathology.

The energy transition, driven by the global shift toward renewable and electrification, necessitates accurate and efficient prediction of electrical load profiles to quantify energy consumption. Therefore, the systematic literature review (SLR), followed by PRISMA guidelines, synthesizes hybrid architectures for sequential electrical load profiles, aiming to span statistical techniques, machine learning (ML), and deep learning (DL) strategies for optimizing performance and practical viability. The findings reveal a dominant trend towards complex hybrid models leveraging the combined strengths of DL architectures such as long short-term memory (LSTM) and optimization algorithms such as genetic algorithm and Particle Swarm Optimization (PSO) to capture non-linear relationships. Thus, hybrid models achieve superior performance by synergistically integrating components such as Convolutional Neural Network (CNN) for feature extraction and LSTMs for temporal modeling with feature selection algorithms, which collectively capture local trends, cross-correlations, and long-term dependencies in the data. A crucial challenge identified is the lack of an established framework to manage adaptable output lengths in dynamic neural network forecasting. Addressing this, we propose the first explicit idea of decoupling output length predictions from the core signal prediction task. A key finding is that while models, particularly optimization-tuned hybrid architectures, have demonstrated quantitative superiority over conventional shallow methods, their performance assessment relies heavily on statistical measures like Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE). However, for comprehensive performance assessment, there is a crucial need for developing tailored, application-based metrics that integrate system economics and major planning aspects to ensure reliable domain-specific validation.

Muscle behaviour during the delivery of neuromuscular electrical stimulation (NMES) is poorly understood, and the use of a novel signal processing and machine learning approach can automate the detection of pain during therapy delivery. Surface electromyography (sEMG) and mechanomyography (MMG) signals represent the electric current generated in muscles and are correlated with neuromuscular activity. Rectus femoris, vastus lateralis, and tibialis anterior muscle groups were stimulated in a self-administered, virtual setting using custom hardware on 12 healthy participants. Subject-reported Verbal Rating Score pain labels were collected during signal acquisition. sEMG and MMG signals were preprocessed, manually and auto-feature extracted, and trained. Noise reduction, time and frequency domain signal processing, and envelope detection techniques were explored. Machine learning architectures were trained using a Leave-One-Subject-Out cross-validation approach to determine the optimal proposed learning architecture: a regularized Long-Term Short Memory neural network with an F1 score of 84.96% over 10 training epochs. This novel pipeline demonstrates the feasibility of using the sEMG and MMG biosignals to predict pain labels. The results show promise in automating NMES delivery to combat muscle atrophy and can be extended to non-healthy subjects in the future.<p></p>

Training machines to develop intelligence is one of the popular studies of computer science. As long as machines with human intelligence is used for the benefits of humans, this technology will flourish civilization. Digital data capturing may process numerous challenges as documents may contain a combination of multiple unstructured data types. Such documents are easily readable to a human but machines may be unable to interpret the data as humans do. Neural network approaches like multilayer perceptron, CNN, RNN and LSTN are effective recognition methods as it can effectively deal with complex computing problems related to training networks and in the process of automating and producing predictive results. In deep neural networks, machines can detect prominent features of a given object without human interventions. The classification and recognition rate of all present algorithms depend chiefly on input images. This paper presents two approaches of deep learning for document recognition problem, convolutional neural network and the result of CNN is continued to RNN and LSTM approaches for synthesizing document. The study attempts to compare the results of CNN vs CNN-LSTM in terms of precision rate.

We present a comparative analysis of traditional machine learning (ML) models, random forest (RF), support vector machine (SVM), and extreme gradient boosting (XGB), and deep learning (DL) architectures, convolutional neural networks (CNN), and bidirectional long short-term memory (BiLSTM) for high-precision 3D positional forecasting. Conventional approaches often underperform when modeling complex spatiotemporal dependencies, limiting their use in dynamic systems such as robotics and autonomous vehicles. This study highlights BiLSTM's advantage in learning bidirectional temporal features, achieving superior R² scores and stable prediction intervals compared to both classical ML and spatially-focused CNN models. Uncertainty metrics, prediction interval coverage probability (PICP), and mean prediction interval width (MPIW) provide additional insight into model reliability. Experiments on a 22-hour GPS dataset confirm that BiLSTM achieves both high accuracy and predictive confidence, underscoring its suitability for real-world trajectory forecasting.

The paper reviews existing models for organizing information for machine learning systems in heterogeneous computing environments. In this context, we focus on structured knowledge representations as they have played a key role in enabling machine learning at scale. The paper highlights recent case studies where knowledge structures when combined with the knowledge of the distributed computation graph have accelerated machine-learning applications by 10 times or more. We extend these concepts to the design of Cognitive Distributed Learning Systems to resolve critical bottlenecks in real-time machine learning applications such as Predictive Analytics and Recommender Systems.

Chemical heuristics have been fundamental to the advancement of chemistry and materials science. These heuristics are typically established by scientists using knowledge and creativity to extract patterns from limited datasets. Machine learning offers opportunities to perfect this approach using computers and larger datasets. Here, we discuss the relationships between traditional heuristics and machine learning approaches. We show how traditional rules can be challenged by large-scale statistical assessment and how traditional concepts commonly used as features are feeding the machine learning techniques. We stress the waste involved in relearning chemical rules and the challenges in terms of data size requirements for purely data-driven approaches. Our view is that heuristic and machine learning approaches are at their best when they work together.

Normal or abnormal? Machine learning for the leakage detection in carbon sequestration projects using pressure field data

Cement is among the materials considered fundamental to the civil construction industry. Quality control of this resource is paramount to ensuring the safety of buildings, reducing environmental impacts, and safeguarding human life. Compressive strength is the primary criterion for measuring cement quality. After preparing a specimen according to current standards, the compressive strength of cement gradually increases until reaching a stability point on the 28th day of production (CS28), a regulatory parameter required by construction standards. However, the 28-day waiting period creates significant challenges for the cement industry, as it delays the identification of factors causing variations in quality. Without timely corrective actions, the production of out-of-specification material may lead to resource wastage and financial losses, while also exacerbating environmental damage. Within this context, this work develops and evaluates machine learning models for predicting CS28, aiming to mitigate these challenges while advancing current knowledge on the application of machine learning to the specific problem of cement compressive strength prediction. The research explores two distinct approaches: local and global analysis. The local analysis focuses on developing models trained exclusively on data from individual cement plants, while the global analysis employs transfer learning techniques-pre-training models on combined data from multiple plants before fine-tuning them for specific plants. Using data from multiple cement manufacturers across two research projects (Innovandi and EMBRAPII CICS), various machine learning architectures were evaluated, including linear regression, tree-based models, and neural networks. To ensure robust model selection, a systematic framework was developed to evaluate models across multiple performance metrics. The results demonstrate strong predictive performance across both approaches, with local models showing effectiveness for individual plants and global models exhibiting remarkable adaptability across different plants. The global models particularly demonstrated enhanced stability and consistency in predictions while maintaining comparable performance to local models. The findings contribute to both theoretical understanding and practical implementation of machine learning in cement production, offering promising solutions for enhancing production efficiency and supporting sustainable industrial practices.

5G is the fifth generation of cellular networks. It enables billions of connected devices to gather and share information in real time; a key facilitator in Industrial Internet of Things (IoT) applications. It has more capabilities in terms of bandwidth, latency/delay, processing powers and flexibility to utilize either edge or cloud resources. Furthermore, 6G is expected to be equipped with the new capability to converge ubiquitous communication, computation, sensing and controlling for a variety of sectors, which heightens the complexity in a more heterogeneous environment This increased complexity, combined with energy efficiency and Service Level Agreement (SLA) requirements makes application of Machine Learning (ML) and distributed ML necessary. A decentralized approach stemming from distributed learning is a very attractive option compared with a centralized architecture for model learning and inference. Distributed ML exploits recent Artificial Intelligence (AI) technology advancements to allow collaborated ML, whilst safeguarding private data, minimizing both communication and computation overhead along with addressing ultra-low latency requirements. In this paper, we review a number of distributed ML architectures and designs, that focus on optimizing communication, computation and resource distribution. Privacy, information security and compute frameworks, are also analyzed and compared with respect to different distributed ML approaches. We summarize the major contributions and trends in this area and highlight the potential of distributed ML to help researchers and practitioners make informed decisions on selecting the right ML approach for 5G and Beyond related AI applications. To enable distributed ML for 5G and Beyond, communication, security, and computing platform often counter balance each other, thus, consideration and optimization of these aspects at an overall system level is crucial to realize the full potential of AI for 5G and Beyond. These different aspects do not only pertain to 5G, but will also enable careful design of distributed machine learning architectures to circumvent the same hurdles that will inevitably burden 5G and Beyond network generations. This is the first survey paper that brings together all these aspects for distributed ML.

Humans are remarkably efficient at decision making, even in "open-ended" problems where the set of possible actions is too large for exhaustive evaluation. Our success relies, in part, on processes for calling to mind the right candidate actions. When these processes fail, the result is a kind of puzzle in which the value of a solution would be obvious once it is considered, but never gets considered in the first place. Recently, machine learning (ML) architectures have attained or even exceeded human performance on open-ended decision making tasks such as playing chess and Go. We ask whether the broad architectural principles that underlie ML success in these domains generate similar consideration failures to those observed in humans. We demonstrate a case in which they do, illuminating how humans make open-ended decisions, how this relates to ML approaches to similar problems, and how both architectures lead to characteristic patterns of success and failure.

Backgrounds: Accurate early classification of heart failure (HF) in women is challenging due to sex-specific gene expression and disease patterns, which traditional models overlook. We propose a generative artificial intelligence (GenAI)-based model for the classification of HF patients using acute myocardial infarction (AMI) gene expression data. Objectives: This study aims to design and scientifically validate a novel GenAI framework, benchmarking it against transformer, deep learning (DL), and machine learning (ML) architectures for robust HF classification using women's transcriptomic data. Methods: Total 26 models designed: A novel wBio-GenAI model, two transformers (Xmers): token diffusion gene (wTDG-Xmer) and neurotopology (wNT-Xmer), 19 deep learning models which include convolutional neural network (CNN)-, long short-term memory (LSTM)- and extended LSTM (xLSTM)-based models, and four ML. The models applied differential expression analysis (DEA) which identifies differentially expressed genes (DEGs) from the public women's microarray GSE57345 samples. Quality control was conducted. The GenAI system was scientifically validated, benchmarked, and statistically tested for reliability. Results: The wBio-GenAI achieved an accuracy of 98.21% and an area-under-the-curve (AUC) of 0.99. The wBio-GenAI is better than the mean of two Xmers by 4.67%, the mean of 19 DLs by 5.16%, and the mean of four MLs by 15.07%. The proposed model meets the regulatory requirements of having a difference < 10% between seen and unseen paradigms. Conclusions: The wBio-GenAI architecture captures the complex transcriptomic patterns, improving HF classification in women and advancing women-specific precision cardiovascular care.

We propose cooperative edge-assisted dynamic federated learning ( CE-FL). CE-FL introduces a distributed machine learning (ML) architecture, where data collection is carried out at the end devices, while the model training is conducted <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">cooperatively</i> at the end devices and the edge servers, enabled via data offloading from the end devices to the edge servers through base stations. CE-FL also introduces floating aggregation point, where the local models generated at the devices and the servers are aggregated at an edge server, which varies from one model training round to another to cope with the network evolution in terms of data distribution and users’ mobility. CE-FL considers the heterogeneity of network elements in terms of communication/computation models and the proximity to one another. CE-FL further presumes a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">dynamic</i> environment with online variation of data at the network devices which causes a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">drift</i> at the ML model performance. We model the processes taken during CE-FL, and conduct analytical convergence analysis of its ML model training. We then formulate network-aware CE-FL which aims to adaptively optimize all the network elements via tuning their contribution to the learning process, which turns out to be a non-convex mixed integer problem. Motivated by the large scale of the system, we propose a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">distributed optimization solver</i> to break down the computation of the solution across the network elements. We finally demonstrate the effectiveness of our framework with the data collected from a real-world testbed.