Unsupervised Learning Techniques Research Articles

Recovering 3D Salt Dome by Gravity Data Inversion Using ResU-Net++

In geophysical research, gravity-based inversion is essential for identifying geological anomalies, mapping rock structures, and extracting resources such as oil and minerals. Traditional gravity inversion methods, however, face challenges such as the volumetric effects of gravity fields and the management of large, complex matrices. Unsupervised learning techniques often struggle with overfitting and interpreting gravity data. This study explores the application of various U-Net-based network architectures in gravity inversion, each offering distinct challenges and advantages. Nested U-Net, although effective, requires a high parameter count, leading to extended training periods. Recurrent Residual U-Net's implicit attention mechanism restricts its dynamic adaptability, while Attention U-Net's lack of residual connections raises concerns about gradient issues. This research comprehensively analyzes the training processes, core functionalities, and module distribution of these networks, including Residual U-Net++. Our synthetic studies compare these networks with traditional focused regularized gravity inversion for reconstructing density anomalies. The results demonstrate that Nested U-Net closely approximates the actual model, despite some redundancy. Recurrent Residual U-Net shows improved alignment with minimal redundancies, and Attention U-Net is effective in density prediction but encounters difficulties in areas of low density. Notably, Residual U-Net++ excels in inversion modeling, achieving the lowest misfit percentage and accurately replicating density values. In practical applications, Residual U-Net++ impressively reconstructed the F2 salt diapir in the Nordkapp Basin with well-defined boundaries that closely match seismic data interpretations. These results underscore the capabilities of Residual U-Net++ in geophysical data analysis, structural reconstruction, and inversion, demonstrating its effectiveness in both simulated settings and real-world scenarios.

Evaluation of Density Functional Theory-Generated Data for Infrared Spectroscopy of Novel Psychoactive Substances Using Unsupervised Learning

Novel psychoactive substances (NPSs) are compounds plotted to modify the chemical structures of prohibited substances, offering alternatives for consumption and evading legislation. The prompt emergence of these substances presents challenges in health concerns and forensic assessment because of the lack of analytical standards. A viable alternative for establishing these standards involves leveraging in silico methods to acquire spectroscopic data. This study assesses the efficacy of utilizing infrared spectroscopy (IRS) data derived from density functional theory (DFT) for analyzing NPSs. Various functionals were employed to generate infrared spectra for five distinct NPS categories including the following: amphetamines, benzodiazepines, synthetic cannabinoids, cathinones, and fentanyls. PRISMA software was conceived to rationalize data management. Unsupervised learning techniques, including Hierarchical Cluster Analysis (HCA), Principal Component Analysis (PCA), and t-distributed stochastic neighbor embedding (t-SNE), were utilized to refine the assessment process. Our findings reveal no significant disparities among the different functionals used to generate infrared spectra data. Additionally, the application of unsupervised learning demonstrated adequate segregation of NPSs within their respective groups. In conclusion, integrating theoretical data and dimension reduction techniques proves to be a powerful strategy for evaluating the spectroscopic characteristics of NPSs. This underscores the potential of this combined methodology as a diagnostic tool for distinguishing IR spectra across various NPS groups, facilitating the evaluation of newly unknown compounds.

Probabilistic modeling of renewable energy sources in smart grids: A stochastic optimization perspective

Machine learning (ML) techniques have emerged as powerful tools for optimizing renewable energy management in smart grids. This paper focuses on the application of ML algorithms to enhance the efficiency and reliability of renewable energy integration within smart grid systems. By leveraging predictive analytics, ML models can forecast energy production and consumption patterns, facilitating proactive decision-making for grid operators and energy stakeholders. The abstract explores various ML methodologies such as supervised learning, unsupervised learning, and reinforcement learning, tailored to address specific challenges in renewable energy management. Supervised learning algorithms enable accurate prediction of renewable energy generation, aiding in resource allocation and demand-response strategies. Unsupervised learning techniques facilitate anomaly detection and clustering of energy consumption patterns, contributing to grid stability and optimization. Reinforcement learning algorithms optimize control strategies, enabling autonomous decision-making in dynamic grid environments. The abstract highlights case studies and real-world implementations where ML techniques have demonstrated significant improvements in renewable energy integration, grid reliability, and cost-effectiveness. Through a synthesis of research findings and practical insights, this abstract elucidates the transformative potential of machine learning in shaping the future of smart grids and renewable energy management.

MACHINE LEARNING AND THE STUDY OF LANGUAGE CHANGE: A REVIEW OF METHODOLOGIES AND APPLICATION

Language change, a fundamental aspect of human communication, has long been a central focus in linguistic research. Traditional methods of analysis, while valuable, have been limited by the scale and complexity of linguistic data. The advent of machine learning (ML) offers transformative potential in this field, enabling the analysis of vast datasets and the discovery of subtle patterns that may elude manual scrutiny. This review paper comprehensively examines the current state of ML methodologies in the study of language change, synthesizing findings from 67 peer-reviewed articles. We delve into diverse ML approaches, including supervised, unsupervised, and deep learning techniques, and critically evaluate their applications across various linguistic domains, such as historical linguistics, sociolinguistics, and language contact. We address challenges related to data availability, bias, and model interpretability, emphasizing the need for transparent and rigorous methodologies. By summarizing key findings and outlining future directions, this review aims to foster interdisciplinary collaboration between linguists and computer scientists, advancing our understanding of the complex dynamics of language evolution.

Unsupervised machine learning for clustering forward head posture, protraction and retraction movement patterns based on craniocervical angle data in individuals with nonspecific neck pain

ObjectivesThe traditional understanding of craniocervical alignment emphasizes specific anatomical landmarks. However, recent research has challenged the reliance on forward head posture as the primary diagnostic criterion for neck pain. An advanced relationship exists between neck pain and craniocervical alignment, which requires a deeper exploration of diverse postures and movement patterns using advanced techniques, such as clustering analysis. We aimed to explore the complex relationship between craniocervical alignment, and neck pain and to categorize alignment patterns in individuals with nonspecific neck pain using the K-means algorithm.MethodsThis study included 229 office workers with nonspecific neck pain who applied unsupervised machine learning techniques. The craniocervical angles (CCA) during rest, protraction, and retraction were measured using two-dimensional video analysis, and neck pain severity was assessed using the Northwick Park Neck Pain Questionnaire (NPQ). CCA during sitting upright in a comfortable position was assessed to evaluate the resting CCA. The average of midpoints between repeated protraction and retraction measures was considered as the midpoint CCA. The K-means algorithm helped categorize participants into alignment clusters based on age, sex and CCA data.ResultsWe found no significant correlation between NPQ scores and CCA data, challenging the traditional understanding of neck pain and alignment. We observed a significant difference in age (F = 140.14, p < 0.001), NPQ total score (F = 115.83, p < 0.001), resting CCA (F = 79.22, p < 0.001), CCA during protraction (F = 33.98, p < 0.001), CCA during retraction (F = 40.40, p < 0.001), and midpoint CCA (F = 66.92, p < 0.001) among the three clusters and healthy controls. Cluster 1 was characterized by the lowest resting and midpoint CCA, and CCA during pro- and -retraction, indicating a significant forward head posture and a pattern of retraction restriction. Cluster 2, the oldest group, showed CCA measurements similar to healthy controls, yet reported the highest NPQ scores. Cluster 3 exhibited the highest CCA during protraction and retraction, suggesting a limitation in protraction movement.DiscussionAnalyzing 229 office workers, three distinct alignment patterns were identified, each with unique postural characteristics; therefore, treatments addressing posture should be individualized and not generalized across the population.

The impact of Covid‐19 on banking groups’ balance sheets in the euro area

AbstractMost studies have found that the Covid‐19 pandemic did not negatively impact the euro area banking industry's performance at an aggregate level. This study explores whether this finding still holds using the return on assets of 16 banking groups operating in the euro area and considering bank‐specific factors, idiosyncrasies related to different exposures of their portfolios to the business cycle and weaknesses stemming from underlying structural vulnerabilities. The banking groups are classified into clusters using unsupervised learning techniques. This research contributes to the empirical literature on the determinants of banks’ performance by highlighting the importance of banks’ heterogeneity, notably controlling for differential performance due to asset quality, solvency and business model. In addition, this paper shows that the magnitude of return on assets’ exposure to the business cycle varies across banks and that inflation in the euro area matters only for a subset of them. Importantly, the study sheds some light on the possible reasons for the mixed results in the literature regarding the role of non‐interest income, the T1 capital ratio and inflation. Finally, no significant effects of the Covid‐19 variables on banks’ return on assets are found during the first seven quarters of the pandemic.

Hype or hope? Ketamine for the treatment of depression: results from the application of deep learning to Twitter posts from 2010 to 2023.

To investigate societal perceptions of ketamine's use in depression therapy by analysing Twitter posts from January 1, 2010 to April 1, 2023. Using Twitter as the social media platform of choice, and employing search terms based on (depression OR depressed OR depressive) AND (ketamine OR esketamine OR Spravato), we collected English-language tweets from January 1, 2010, to April 1, 2023. Using unsupervised machine learning and natural language processing (NLP) techniques, including Bidirectional Encoder Representations from Transformers (BERT) and BERTopic, the study identified prevalent topics surrounding public chatter around the use of ketamine in depression treatment. Manual thematic analyses further refined these topics into themes. Out of an initial dataset of 99,405 tweets, after removing duplicate tweets, re-tweets and tweets posted by organizations over Twitter, 18,899 unique tweets from presumably individual users were analysed. Analysis of temporal trends revealed a shift in public attitudes, particularly after the United States Food and Drug Administration (FDA)'s 2019 approval of ketamine for depression. Three major themes emerged: a changing regulatory landscape, cautious optimism, and personal experiences with the drug. There was an initial spike in discussions post-FDA approval in 2019. Thereafter, cautious optimism (Theme 2) decreased among the general public, with more personal accounts (Theme 3) highlighting the potential benefits for some treatment-resistant patients. Limitations of the study include Twitter's inherent biases towards younger, English-speaking demographics. In summary, the public's multifaceted perception leans towards a hopeful stance on ketamine's therapeutic potential for depression.

Machine Learning Library in C++

This project aims to develop a comprehensive and versatile machine learning library in C++ tailored to address the diverse needs of developers and researchers in the field. The library encompasses a robust set of core machine learning algorithms, encompassing supervised, unsupervised, and reinforcement learning techniques. Additionally, it incorporates essential data preprocessing tools to streamline data manipulation and feature engineering tasks, along with model evaluation capabilities crucial for assessing algorithm performance. The library's primary focus is on providing a rich suite of machine learning algorithms, This empowers users to effectively prepare data for training machine learning models. Additionally, the library provides tools for data splitting, and model evaluation to ensure reliable and robust model performance assessment.

Clustering techniques and innovation-based comparison in Londrina and Region companies

Innovation is often considered a cornerstone for success across various companies. However, research focused on measuring and describing innovation frequently relies on classical statistical techniques. In this context, this study applied unsupervised machine learning techniques to cluster companies in the Londrina region, investigating how variables related to innovation differ among the identified clusters. Data were collected through a survey instrument adapted from CIS 4 and PINTEC, encompassing 26 responding companies, although 23 were analyzed in this study. Four clustering algorithms were employed: k-means, k-means+PCA, hierarchical, and hierarchical+PCA. Regarding the results, the hierarchical+PCA algorithm showed improved separation between service and manufacturing companies. Clusters identified with the value ``"1" indicated concerns related to regular investment in R&amp;D, collaborations for innovation, and requests/registrations of patents in the last three years. Analyzing demographic characteristics revealed that clusters identified by hierarchical+PCA exhibited a higher presence of service sector companies, while cluster 1 showed a prevalence of industries, suggesting that these possess more innovative characteristics in the Londrina region.

Carcass traits and morphometry, typification of the Longissimus dorsi muscle and non-carcass components of hair lambs: can biscuit bran completely replace corn? A machine learning approach.

Biscuit bran (BB) is a co-product with worldwide distribution, with Brazil as the second largest cookie producer in the world with 1,157,051 tons. We evaluate the impact of completely replacing corn with BB on the characteristics and morphometry of carcass of purebred and crossbred Morada Nova lambs using machine learning techniques as an auxiliary method. Twenty male lambs from two genetic groups (GG) were used: purebred red-coated Morada Nova (MNR) and crossbred MNR × white-coated Morada Nova (MNF1). Supervised and unsupervised machine learning techniques were used. No interaction (P > 0.05) was observed between diets (D) and genetic groups (GG) and no simple isolated effect was observed for carcass characteristics, qualitative-quantitative typification of the Longissimus dorsi muscle, weight of non-carcass components, weight and yield of commercial cuts and carcass morphometric measurements. The formation of two horizontal clusters was verified: (i) crossed lambs with corn and BB and (ii) purebred lambs fed corn and BB. Vertically, three clusters were formed based on carcass and meat characteristics of native lambs: (i) thermal insulation, body capacity, true yield, and commercial cuts; (ii) choice, performance, physical carcass traits, and palatability; and (iii) yield cuts and non-carcass components. The heatmap also allowed us to observe that pure MN lambs had a greater body capacity when fed BB, while those fed corn showed superiority in commercial cuts, true yields, and non-carcass components. Crossbred lambs, regardless of diet, showed a greater association of physical characteristics of the carcass, performance, palatability, and less noble cuts. Crossbred lambs, regardless of diet, showed a greater association of physical characteristics of the carcass, performance, palatability, and less noble cuts. BB can be considered an alternative energy source in total replacement of corn. Integrating of machine learning techniques is a useful statistical tool for studies with large numbers of variables, especially when it comes to analyzing complex data with multiple effects in the search for data patterns and insights in decision-making on the farm.

A Comprehensive Review of the Three Main Topic Modeling Algorithms and Challenges in Albanian Employability Skills

Today’s jobseekers face many obstacles while trying to find a career that aligns with their interests, employability soft skills, and professional experience. In Albania, jobseekers frequently initiate their job search by actively exploring job vacancies listed on various online job portals. The analysis of job vacancies posted online provides an added advantage to the labour market actors compared to traditional survey-based analyses. This is because it enables a faster analytical process, promotes decision-making based on accurate data, and should be carefully considered by every country when formulating their Labor Market Policies. Since the data posted online are unlabelled, it has been proven that the potential of unsupervised learning techniques, more precisely the Topic Modelling algorithms, is outstanding when applied to analysing job vacancies, mainly with regard to assessing employability soft skills. Algorithms in topic modelling are essential for uncovering hidden patterns in texts, facilitating the extraction of important data, generating document summaries, and enhancing content comprehension. This paper analyses and compares the three primary methodologies and algorithms used in topic modelling, which can be applied to analyse employability soft-skills: Latent Semantic Analysis (LSA), Latent Dirichlet Allocation (LDA), and BERTopic. At the end of the paper, conclusions are drawn regarding superior performance and optimal algorithm applicability, challenges, and limitations through a review of studies conducted in the Albanian job market.

Predictive Analytics-Enabled Cyber Attack Detection

Cyber-attacks are becoming increasingly sophisticated and difficult to detect using traditional security measures. To address this challenge, we propose a predictive analytics- enabled cyber-attack detection system that utilizes machine learning algorithms to analyze network traffic and identify potential security threats in real time. Our system uses a combination of supervised and unsupervised learning techniques to identify patterns and anomalies in network data, and to generate anomaly and normal alert. The system is trained using historical data from known cyber-attacks and anomalies and we visualize the accuracy of various algorithms.

Unveiling the Power of Data: A Journey Through Machine Learning Techniques

Machine Learning Techniques: Unveiling the Power of Data-Driven Solutions Machine learning (ML) has become a transformative force across disciplines, revolutionizing how computers extract knowledge from data and tackle complex tasks without explicit programming. This research paper delves into the core principles of machine learning, exploring its fundamental techniques and their far-reaching applications. We embark on a journey through the two main paradigms that govern learning: supervised and unsupervised. Supervised learning equips algorithms to make informed predictions based on labeled data, encompassing tasks like spam filtering, disease diagnosis, and stock market forecasting. The algorithm ingests this data, which includes both the input features and the corresponding desired output, and learns the intricate relationships between them. This empowers it to generalize this knowledge to unseen data, enabling accurate predictions. Unsupervised learning, in contrast, thrives on unlabeled data, where the data points lack predefined categories or target variables. Here, the objective is to uncover the inherent structures or patterns within the data itself. Unsupervised learning techniques are invaluable for tasks like data clustering, where similar data points are grouped together to reveal underlying relationships. Additionally, they play a crucial role in dimensionality reduction, a process of compressing high-dimensional data into a lower-dimensional space for efficient processing and analysis. This is particularly beneficial when dealing with massive datasets that may overwhelm traditional computational methods. Beyond these fundamental paradigms, the landscape of machine learning extends further. Deep learning, a subfield inspired by the structure and function of the human brain, utilizes artificial neural networks with multiple layers. These complex networks are adept at learning intricate patterns from vast amounts of data, achieving remarkable success in areas like computer vision, natural language processing, and speech recognition. Deep learning models, for instance, power facial recognition technology, enabling our devices to identify individuals with exceptional accuracy. In the realm of natural language processing, deep learning algorithms fuel machine translation systems, allowing seamless communication across languages. Reinforcement learning presents another powerful approach. Here, an agent interacts with an environment, learning through trial and error. The agent receives rewards for desired actions and penalties for undesired ones, constantly refining its decision-making process. This approach has applications in robotics, where robots can learn to navigate complex environments and perform tasks with increasing autonomy. Reinforcement learning is also used in game playing, where AI agents can learn optimal strategies through self-play and achieve superhuman levels of performance. As machine learning continues its exponential growth, ethical considerations become paramount. Potential biases within the training data can lead to discriminatory outcomes in areas like loan approvals or facial recognition software. Additionally, the explainability and transparency of complex models, particularly deep learning models, raise concerns. To ensure responsible development and deployment of ML, addressing these issues is crucial. The field requires robust methodologies to mitigate bias in training data and develop techniques that make complex models interpretable, allowing us to understand the rationale behind their predictions. Looking towards the horizon, the future of machine learning is brimming with potential. From personalized healthcare with tailored treatment plans to self-driving cars that navigate roads with exceptional precision, ML promises to revolutionize various aspects of our lives. Continuous research and development will enhance the capabilities and robustness of machine learning models, further expanding their reach and impact. By harnessing the power of data and fostering responsible development practices, machine learning will undoubtedly play a pivotal role in shaping a more intelligent and data-driven future

Healthcare insurance fraud detection using data mining

BackgroundHealthcare programs and insurance initiatives play a crucial role in ensuring that people have access to medical care. There are many benefits of healthcare insurance programs but fraud in healthcare continues to be a significant challenge in the insurance industry. Healthcare insurance fraud detection faces challenges from evolving and sophisticated fraud schemes that adapt to detection methods. Analyzing extensive healthcare data is hindered by complexity, data quality issues, and the need for real-time detection, while privacy concerns and false positives pose additional hurdles. The lack of standardization in coding and limited resources further complicate efforts to address fraudulent activities effectively.MethodolgyIn this study, a fraud detection methodology is presented that utilizes association rule mining augmented with unsupervised learning techniques to detect healthcare insurance fraud. Dataset from the Centres for Medicare and Medicaid Services (CMS) 2008-2010 DE-SynPUF is used for analysis. The proposed methodology works in two stages. First, association rule mining is used to extract frequent rules from the transactions based on patient, service and service provider features. Second, the extracted rules are passed to unsupervised classifiers, such as IF, CBLOF, ECOD, and OCSVM, to identify fraudulent activity.ResultsDescriptive analysis shows patterns and trends in the data revealing interesting relationship among diagnosis codes, procedure codes and the physicians. The baseline anomaly detection algorithms generated results in 902.24 seconds. Another experiment retrieved frequent rules using association rule mining with apriori algorithm combined with unsupervised techniques in 868.18 seconds. The silhouette scoring method calculated the efficacy of four different anomaly detection techniques showing CBLOF with highest score of 0.114 followed by isolation forest with the score of 0.103. The ECOD and OCSVM techniques have lower scores of 0.063 and 0.060, respectively.ConclusionThe proposed methodology enhances healthcare insurance fraud detection by using association rule mining for pattern discovery and unsupervised classifiers for effective anomaly detection.

Characterization of Relativistic Electron Precipitation Events Observed by the CALET Experiment Using Self‐Organizing‐Maps

AbstractRelativistic electron precipitation (REP) is a relatively high‐latitude phenomenon where high‐energy electrons trapped in the outer radiation belt are lost into the Earth’s atmosphere. REP events observed at low Earth orbit show varying temporal profiles and global distributions. While the precipitation origin has been attributed to specific wave modes or scattering sources, the sorting of REP events by type or driver remains an unsolved challenge. In this study, we analyze the temporal profile of relativistic electron precipitation events observed by the CALorimetric Electron Telescope (CALET) experiment on board the International Space Station. We use an unsupervised machine learning technique called Self‐Organizing‐Maps (SOM) to automatically detect and then classify relativistic electron events observed by the two scintillator layers at the top of the apparatus, sensitive to electrons with energies &gt;1.5 MeV and &gt;3.4 MeV, respectively. We calculate the power spectral density (PSD) of the count rates observed by both sensors and use them as an input for the SOM. The SOM technique groups the PSDs by their similarity, resulting in a classification of relativistic electron events by the periodicity of the observed precipitation. We investigate the L‐shell and magnetic local time distribution of the resulting classification, and energy spectral index associated with the observations. Clear precipitation patterns are observed and compared to past precipitation categorization attempts as well as known distributions of various scattering mechanisms. The classification reveals features through the sorting of the variability of the rapid precipitation, allowing the identification of different precipitation populations with varying properties.

UNCERTAINTY ASSESSMENT IN UNSUPERVISED MACHINE LEARNING METHODS FOR DEEPWATER CHANNEL SEISMIC FACIES USING OUTCROP-DERIVED 3D MODELS AND SYNTHETIC SEISMIC DATA

Unsupervised machine learning (ML) techniques have been widely applied to analyze seismic reflection data, including the identification of seismic facies and structural features. However, interpreting the resulting clusters often relies on geoscientists’ expertise, necessitating a robustness assessment of these methods. To evaluate their reliability, synthetic data generated from an actual outcrop model were employed to demonstrate how two unsupervised methods, Self-Organizing Maps (SOM) and Generative Topographic Maps (GTM), cluster deepwater channel-related seismic facies and then measure the associated error. Six seismic attributes, comprising RMS amplitude, instantaneous envelope, peak magnitude, and spectral decomposition frequencies at 20, 40, and 55 Hz, served as input variables. Geobodies were assigned to each cluster formed, and error in facies clustering was quantified by comparing the actual 3D model with the facies grouped by machine learning methods on a voxel-by-voxel basis. This allowed for error quantification and the computation of metrics such as F1 score and accuracy through correlation matrices. Key findings revealed that (1) GTM and SOM exhibited similar performance, with a clustering configuration of 81 for GTM slightly outperforming others. (2) Error rates were approximately 2% for the predominant facies (background shale) but significantly higher for individual channel-related facies, suggesting that channel clusters might represent multiple facies. (3) Resolution and imbalanced data distribution impacted seismic facies predictability, resulting in nonuniqueness in cluster generation. (4) Using synthetic seismic data proved valuable for experimenting with different unsupervised ML, highlighting the need for assessing uncertainty in these methods, given their implications for crucial economic decisions reliant on reservoir interpretation, modeling, and volumetric estimations.

A Critical Analysis of Deep Semi-Supervised Learning Approaches for Enhanced Medical Image Classification

Deep semi-supervised learning (DSSL) is a machine learning paradigm that blends supervised and unsupervised learning techniques to improve the performance of various models in computer vision tasks. Medical image classification plays a crucial role in disease diagnosis, treatment planning, and patient care. However, obtaining labeled medical image data is often expensive and time-consuming for medical practitioners, leading to limited labeled datasets. DSSL techniques aim to address this challenge, particularly in various medical image tasks, to improve model generalization and performance. DSSL models leverage both the labeled information, which provides explicit supervision, and the unlabeled data, which can provide additional information about the underlying data distribution. That offers a practical solution to resource-intensive demands of data annotation, and enhances the model’s ability to generalize across diverse and previously unseen data landscapes. The present study provides a critical review of various DSSL approaches and their effectiveness and challenges in enhancing medical image classification tasks. The study categorized DSSL techniques into six classes: consistency regularization method, deep adversarial method, pseudo-learning method, graph-based method, multi-label method, and hybrid method. Further, a comparative analysis of performance for six considered methods is conducted using existing studies. The referenced studies have employed metrics such as accuracy, sensitivity, specificity, AUC-ROC, and F1 score to evaluate the performance of DSSL methods on different medical image datasets. Additionally, challenges of the datasets, such as heterogeneity, limited labeled data, and model interpretability, were discussed and highlighted in the context of DSSL for medical image classification. The current review provides future directions and considerations to researchers to further address the challenges and take full advantage of these methods in clinical practices.

Optimization of Obstructive Sleep Apnea Management: Novel Decision Support via Unsupervised Machine Learning

This study addresses Obstructive Sleep Apnea (OSA), which impacts around 936 million adults globally. The research introduces a novel decision support method named Communalities on Ranking and Objective Weights Method (CROWM), which employs principal component analysis (PCA), unsupervised Machine Learning technique, and Multicriteria Decision Analysis (MCDA) to calculate performance criteria weights of Continuous Positive Airway Pressure (CPAP—key in managing OSA) and to evaluate these devices. Uniquely, the CROWM incorporates non-beneficial criteria in PCA and employs communalities to accurately represent the performance evaluation of alternatives within each resulting principal factor, allowing for a more accurate and robust analysis of alternatives and variables. This article aims to employ CROWM to evaluate CPAP for effectiveness in combating OSA, considering six performance criteria: resources, warranty, noise, weight, cost, and maintenance. Validated by established tests and sensitivity analysis against traditional methods, CROWM proves its consistency, efficiency, and superiority in decision-making support. This method is poised to influence assertive decision-making significantly, aiding healthcare professionals, researchers, and patients in selecting optimal CPAP solutions, thereby advancing patient care in an interdisciplinary research context.

Recognising small colour changes with unsupervised learning, comparison of methods

Colour differentiation is crucial in machine learning and computer vision. It is often used when identifying items and objects based on distinct colours. While common colours like blue, red, green, and yellow are easily distinguishable, some applications require recognising subtle colour variations. Such demands arise in sectors like agriculture, printing, healthcare, and packaging. This research employs prevalent unsupervised learning techniques to detect printed colours on paper, focusing on CMYK ink (saturation) levels necessary for recognition against a white background. The aim is to assess whether unsupervised clustering can identify colours within QR-Codes. One use-case for this research is usage of functional inks, ones that change colour based on environmental factors. Within QR-Codes they serve as low-cost IoT sensors. Results of this research indicate that K-means, C-means, Gaussian Mixture Model (GMM), Hierarchical clustering, and Spectral clustering perform well in recognising colour differences when CMYK saturation is 20% or higher in at least one channel. K-means stands out when saturation drops below 10%, although its accuracy diminishes significantly, especially for yellow or magenta channels. A saturation of at least 10% in one CMYK channel is needed for reliable colour detection using unsupervised learning. To handle ink densities below 5%, further research or alternative unsupervised methods may be necessary.

Vision transformers are active learners for image copy detection

Image Copy Detection (ICD) is developed to identify and track duplicated or manipulated images. The majority of existing methods rely on Convolutional Neural Networks (CNNs) and are trained using unsupervised learning techniques, which leads to subpar performance. We discover that by carefully designing the training process, Vision Transformer (ViT) backbones yield superior results. Specifically, directly training a ViT for ICD often leads to overfitting on the training images, which in turn results in poor generalization to unseen (test) images. Consequently, we initially train a CNN (such as ResNet-50), and during the ViT training, the distances between the features of CNN and ViT are regularized. We also incorporate an active learning method to further enhance performance. Notably, due to the visual discrepancy between auto-generated transformations and those used in the query set, we incorporate a small number (approximately 0.5% of unlabeled training images) of manually produced and labeled positive pairs. Training models on these pairs results in a significant performance boost though with little cost. Experimental findings demonstrate the effectiveness of our approach, and our method achieves state-of-the-art performance. Our code is available at: https://github.com/WangWenhao0716/ViT4ICD.