Search

Managing the rapid growth of deprived areas (commonly known as slums, informal settlements, etc.) in cities of Low- to Middle-Income Countries (LMICs) demands detailed and consistent information that is often unavailable. Recent Earth Observation (EO) mapping approaches with supervised classification models overlook the diversity of deprived areas and require resource-intensive training sets. In this study, we analyse the potential of unsupervised machine learning (ML) models to capture intra-urban diversity of deprived areas in São Paulo, using solely open geodata. We provide a workflow of characterising deprivation at a city scale with a disaggregated approach, offering scalability and transferability potential. First, we extract a pool of spatial features from open geospatial datasets to characterise the morphological and environmental conditions of the study area. After input preparation, we train and optimise a k-means model, including a coupled feature importance tool. Four cluster types emerged with different deprivation aspects such as higher and lower accessibility to services and infrastructure, sparser and denser occupation; regular and complex morphology; flat and steep terrain. This alternative methodology to capture diversity of deprived areas with open EO-based features can inform locally targeted, thus more efficient, urban policies and interventions.

Deep learning facilitates complex medical data analysis and is increasingly being explored in colorectal cancer diagnostics. However, the training cost of the deep learning model limits its real-world medical utility. In this study, we present a composite network that combines deep learning and unsupervised K-means clustering algorithm (RK-net) for automatic processing of medical images. RK-net was more efficient in image refinement compared with manual screening and annotation. The training of a deep learning model for colorectal cancer diagnosis was accelerated by two times with utilization of RK-net-processed images. Better performance was observed in training loss and accuracy achievement as well. RK-net could be useful to refine medical images of the ever-expanding quantity and assist in subsequent construction of the artificial intelligence model.

Epicardial adipose tissue (EAT) accumulation is associated with the progression of atrial fibrillation. However, the histological features of EATs are poorly defined and their correlation with atrial fibrosis is unclear.The purpose of this study was to identify and characterize EAT subgroups in the persistent atrial fibrillation (PeAF) cohorts.EATs and the corresponding left atrial appendage samples were obtained from patients with PeAF via surgical intervention. Adipocyte markers, that is, Uncoupling Protein 1, Transcription Factor 21, and CD137, were examined. On the basis of expression of adipocyte markers, patients with PeAF were categorized into subgroups by using unsupervised clustering analysis. Clinical characteristics, histological analyses, and outcomes were subsequently compared across the clusters. External validation was performed in a validation cohort.The ranking of feature importance revealed that the 3 adipocyte markers were the most relevant factors for atrial fibrosis compared with other clinical indicators. On the k-medoids analysis, patients with PeAF could be categorized into 3 clusters in the discovery cohort. The histological studies revealed that patients in cluster 1 exhibited statistically larger size of adipocytes in EATs and severe atrial fibrosis in left atrial appendages. Findings were replicated in the validation cohort, where severe atrial fibrosis was noted in cluster 1. Moreover, in the validation cohort, there was a high degree of overlap between the supervised classification results and the unsupervised cluster results from the k-medoids method.Machine learning-based cluster analysis could identify subtypes of patients with PeAF having distinct atrial fibrosis profiles. Additionally, EAT whitening (increased proportion of white adipocytes) may be involved in the process of atrial fibrosis.

In this paper, we propose a framework to generate diverse clustering solutions and conduct solution retrieval to improve performance. Specifically, we first project unlabelled data from multiple domains into a shared space while preserving the respective semantics. This space allows that representations of samples in a hard domain are recovered by a linear combination of those of others in the easy domains. Meanwhile, a clustering algorithm is adopted to provide pseudo labels for a conditional generative adversarial network to synthesize representations that in turn promote the learning of the above space. Second, we conduct the joint learning of feature projection and partition matrices on batches of representations, where the former ones are considered as clustering solutions and input into another generative adversarial network to generate more solutions. Third, we utilize the fusion of diffusion to effectively retrieve and extract the knowledge in multiple solutions to obtain the final clustering. We perform comparative experiments against other methods on multiple benchmark data sets. Experimental results demonstrate the effectiveness and superiority of our proposed method.

This study aimed to study the public's sentiments on the current monkeypox outbreaks via an unsupervised machine learning analysis of social media posts.This was an exploratory analysis of tweets sentiments.We extracted original tweets containing the terms 'monkeypox', 'monkey pox' or 'monkey_pox' and posted them in the English language from 6 May 2022 (first case detected in the United Kingdom) to 23 July 2022 (when World Health Organization declared Monkeypox to be a global health emergency). Retweets and duplicate tweets were excluded from study. Bidirectional Encoder Representations from Transformers (BERT) Named Entity Recognition. This was followed by topic modelling (specifically BERTopic) and manual thematic analysis by the study team, with independent reviews of the topic labels and themes.Based on topic modelling and thematic analysis of a total of 352,182 Twitter posts, we derived five topics clustered into three major themes related to the public discourse on the ongoing outbreaks. These include concerns of safety, stigmatisation of minority communities, and a general lack of faith in public institutions. The public sentiments underscore growing (and existing) partisanship, personal health worries in relation to the evolving situation, as well as concerns of the media's portrayal of lesbian, gay, bisexual, transgender and queer and minority communities, which might further stigmatise these groups.Monkeypox is an emerging infectious disease of public concern. Our study has highlighted important societal issues, including misinformation, political mistrust and anti-gay stigma that should be sensitively considered when designing public health policies to contain the ongoing outbreaks.

Computer tomography is an extensively used method for the detection of the disease in the subjects. Basically, computer-aided tomography depending on the artificial intelligence reveals its significance in smart health care monitoring system. Owing to its security and the private issue, analyzing the computed tomography dataset has become a tedious process. This study puts forward the convolutional autoencrypted deep learning neural network to assist unsupervised learning technique. By carrying out various experiments, our proposed method produces better results comparative to other traditional methods, which efficaciously solves the issues related to the artificial image description. Hence, the convolutional autoencoder is widely used in measuring the lumps in the bronchi. With the unsupervised machine learning, the extracted features are used for various applications.

We previously reported that archetypal analysis (AA), a type of unsupervised machine learning, identified and quantified patterns of visual field (VF) loss in idiopathic intracranial hypertension (IIH), referred to as archetypes (ATs). We assessed whether AT weight changes over time are consistent with changes in conventional global indices, whether visual outcome or treatment effects are associated with select AT, and whether AA reveals residual VF defects in eyes deemed normal after treatment.Analysis of data collected from a randomized controlled trial.Two thousand eight hundred sixty-two VFs obtained from 165 participants during the Idiopathic Intracranial Hypertension Treatment Trial (IIHTT).We applied a 14-AT model derived from IIHTT VFs. We examined changes in individual AT weights over time for all study eyes and evaluated differences between treatment groups. We created an AT change score to assess overall VF change from baseline. We tested threshold baseline AT weights for association with VF outcome and treatment effect at 6 months. We determined the abnormal ATs with meaningful weight at outcome for VFs with a mean deviation (MD) of -2.00 dB or more.Individual AT weighting coefficients and MD.Archetype 1 (a normal VF pattern) showed the greatest weight change for all study eyes, increasing from 11.9% (interquartile range [IQR], 0.44%-24.1%) at baseline to 31.2% (IQR, 16.0%-45.5%) at outcome (P 0.001). Archetype 1 weight change (r = 0.795; P0.001) and a global score of AT change (r = 0.988; P0.001) correlated strongly with MD change. Study eyes with baseline AT2 (a mild diffuse VF loss pattern) weight of 44% or more (≥ 1 standard deviation more than the mean) showed higher AT2 weights at outcome than those with AT2 weight of44% at baseline (P0.001). Only the latter group showed a significant acetazolamide treatment effect. Archetypal analysis revealed residual VF loss patterns, most frequently representing mild diffuse loss and an enlarged blind spot in 64 of 66 study eyes with MD of -2.00 dB or more at outcome.Archetypal analysis provides a quantitative approach to monitoring VF changes in IIH. Baseline AT features may be associated with treatment response and VF outcome. Archetypal analysis uncovers residual VF defects not otherwise revealed by MD.

Modern mobile devices are able to provide context-aware and personalized services to the users, by leveraging on their sensing capabilities to infer the activity and situation in which a person is currently involved. Current solutions for context-recognition rely on annotated data and experts’ knowledge to predict the user context. In addition, their prediction ability is strongly limited to the set of situations considered during the model training or definition. However, in a mobile environment, the user context continuously evolves, and it cannot be merely restricted to a set of predefined classes. To overcome these limitations, we propose COMPASS, a novel unsupervised and online clustering algorithm aimed at identifying the user context in mobile environments based on the stream of high-dimensional data generated by smartphone sensors. COMPASScan distinguish an arbitrary number of user’s contexts from the sensors’ data, without defining a priori the collection of expected situations. This key feature makes it a general-purpose solution to provide context-aware features to mobile devices, supporting a broad set of applications. Experimental results on 18 synthetic and 2 real-world datasets show that COMPASS correctly identifies the user context from the sensors’ data stream, and outperforms the state-of-the-art solutions in terms of both clusters configuration and purity. Eventually, we evaluate its performances in terms of execution time and the results show that COMPASS can process 1000 high-dimensional samples in less than 20 seconds, while the reference solutions require about 60 minutes to evaluate the entire dataset. Keywords: Context-awareness, Unsupervised Machine Learning, Online Clustering, Mobile Computing

Ensuring that medicines are prescribed safely is fundamental to the role of healthcare professionals who need to be vigilant about the risks associated with drugs and their interactions with other medicines (polypharmacy). One aspect of preventative healthcare is to use artificial intelligence to identify patients at risk using big data analytics. This will improve patient outcomes by enabling pre-emptive changes to medication on the identified cohort before symptoms present. This paper presents a mean-shift clustering technique used to identify groups of patients at the highest risk of polypharmacy. A weighted anticholinergic risk score and a weighted drug interaction risk score were calculated for each of 300,000 patient records registered with a major regional UK-based healthcare provider. The two measures were input into the mean-shift clustering algorithm and this grouped patients into clusters reflecting different levels of polypharmaceutical risk. Firstly, the results showed that, for most of the data, the average scores are not correlated and, secondly, the high risk outliers have high scores for one measure but not for both. These suggest that any systematic recognition of high-risk groups should consider both anticholinergic and drug–drug interaction risks to avoid missing high-risk patients. The technique was implemented in a healthcare management system and easily and automatically identifies groups at risk far faster than the manual inspection of patient records. This is much less labour-intensive for healthcare professionals who can focus their assessment only on patients within the high-risk group(s), enabling more timely clinical interventions where necessary.

Mitochondrial RNA editing in trypanosomes represents an attractive target for developing safer and more efficient drugs for treating infections with trypanosomes because this RNA editing pathway is not found in humans. Other workers have targeted several enzymes in this editing system, but not the RNA. Here, we target a universal domain of the RNA editing substrate, which is the U-helix formed between the oligo-U tail of the guide RNA and the target mRNA. We selected a part of the U-helix that is rich in G-U wobble base pairs as the target site for the virtual screening of 262,000 compounds. After chemoinformatic filtering of the top 5000 leads, we subjected 50 representative complexes to 50 nanoseconds of molecular dynamics simulations. We identified 15 compounds that retained stable interactions in the deep groove of the U-helix. The microscale thermophoresis binding experiments on these five compounds show low-micromolar to nanomolar binding affinities. The UV melting studies show an increase in the melting temperatures of the U-helix upon binding by each compound. These five compounds can serve as leads for drug development and as research tools to probe the role of the RNA structure in trypanosomal RNA editing.

Šiame darbe yra pristatoma jutiklių stebėjimo ir anomalinio vandens naudojimo aptikimo sistema, skirta palengvinti jutiklių duomenų monitoringą bei galimo vandens nuotėkio aptikimą. Ši sistema yra sudaryta iš mobiliosios aplikacijos, kuri yra įdiegta naudotojo telefone, serverio bei mikrovaldiklio, kuris yra įrengtas naudotojo vandens tiekimo sistemoje. Serveris naudoja mašininio mokymo algoritmą, kad būtų galima efektyviai aptikti galimą vandens nuotėkį ir greičiau išvengti didelių nuostolių. Serveris taip pat siunčia aliarminius pranešimus naudotojui į mobilųjį įrenginį pranešti apie anomalinį vandens naudojimą su tikslu vartotojui leisti išvengti nepageidaujamų pasekmių efektyviai ir greitai. Darbo tiriamojoje dalyje yra palyginami įvairūs mašininio mokymo algoritmai naudojant įvairius scenarijus bei metrikas, siekiant surasti efektyviausią modelį nustatant anomalinį vandens naudojimą. Tokio tipo sistema gali būti naudojama tiek privačiuose namuose, tiek ir viešosiose įstaigose. Projektas yra naudingas ne tik siekiant išvengti finansinių nuostolių, tačiau ir tausojant vieną iš svarbiausių aplinkos resursų – vandenį., In this paper we present a sensors monitoring and anomalous water use detection system to facilitate sensors data monitoring and detection of potential water leaks. The system consists of a mobile application that is installed on the user’s phone, a server, and a microcontroller, that is installed in the user’s water supply system. The server uses a machine learning algorithm to efficiently detect potential water leaks and prevent major losses. The server also sends alarm messages to the user’s mobile device to report anomalous water use with the aim of allowing the user to avoid unwanted consequences quickly. In the exploratory part of the work different machine learning algorithms are compared using different scenarios and metrics to find the most efficient model for detecting anomalous water usage. This type of system can be used in private homes as well as in public institutions. The project is not only beneficial for avoiding financial losses, but also for conserving one of the most important environmental resource – water.

Fields in the social sciences, such as education research, have started to expand the use of computer-based research methods to supplement traditional research approaches. Natural language processing techniques, such as topic modeling, may support qualitative data analysis by providing early categories that researchers may interpret and refine. This study contributes to this body of research and answers the following research questions: (RQ1) What is the relative coverage of the latent Dirichlet allocation (LDA) topic model and human coding in terms of the breadth of the topics/themes extracted from the text collection? (RQ2) What is the relative depth or level of detail among identified topics using LDA topic models and human coding approaches? A dataset of student reflections was qualitatively analyzed using LDA topic modeling and human coding approaches, and the results were compared. The findings suggest that topic models can provide reliable coverage and depth of themes present in a textual collection comparable to human coding but require manual interpretation of topics. The breadth and depth of human coding output is heavily dependent on the expertise of coders and the size of the collection; these factors are better handled in the topic modeling approach.

The COVID-19 pandemic has impacted daily lives around the globe. Since 2019, the amount of literature focusing on COVID-19 has risen exponentially. However, it is almost impossible for humans to read all of the studies and classify them. This article proposes a method of making an unsupervised model called a zero-shot classification model, based on the pre-trained BERT model. We used the CORD-19 dataset in conjunction with the LitCovid database to construct new vocabulary and prepare the test dataset. For NLI downstream task, we used three corpora: SNLI, MultiNLI, and MedNLI. We significantly reduced the training time by 98.2639% to build a task-specific machine learning model, using only one Nvidia Tesla V100. The final model can run faster and use fewer resources than its comparators. It has an accuracy of 27.84%, which is lower than the best-achieved accuracy by 6.73%, but it is comparable. Finally, we identified that the tokenizer and vocabulary more specific to COVID-19 could not outperform the generalized ones. Additionally, it was found that BART architecture affects the classification results.

Many public health policymaking questions involve data subsets representing application-specific attributes and geographic location. We develop and evaluate standard and tailored techniques for clustering via unsupervised learning (UL) algorithms on such amalgamated (dual-domain) data sets. The aim of the associated algorithms is to identify geographically efficient clusters that also maximize the number of statistically significant differences in disease incidence and demographic variables across top clusters. Two standard UL approaches, k means with k++ initialization (k++) and the standard self-organizing map (SSOM), are considered along with a new, tailored version of the SOM (TSOM). The TSOM algorithm involves optimization of a customized objective function with terms promoting individual geographic cluster cohesion while also maximizing the number of differences across clusters, and two hyper-parameters controlling the relative weighting of geographic and attribute subspaces in a non-Euclidean distance measure within the clustering problem. The performance of these three techniques (k++, SSOM, TSOM) is compared and evaluated in the context of a data set for colorectal cancer incidence in the state of California, at the level of individual counties. Clusters are visualized via chloropleth maps and ordered graphs are also used to illustrate disparities in disease incidence among four identity groups. While all three approaches performed well, the TSOM identified the largest number of disease and demographic disparities while also yielding more geographically efficient top clusters. Techniques presented in this study are relevant to applications including the delivery of health care resources and identifying disparities among identity groups, and to questions involving coordination between county- and state-level policymakers.

While much research has highlighted phenotypic heterogeneity in obsessive compulsive disorder (OCD), less work has focused on heterogeneity in neural activity. Conventional neuroimaging approaches rely on group averages that assume homogenous patient populations. If subgroups are present, these approaches can increase variability and can lead to discrepancies in the literature. They can also obscure differences between various subgroups. To address this issue, we used unsupervised machine learning to identify subgroup clusters of patients with OCD who were assessed by task-based fMRI. We predominantly focused on activation of cognitive control and performance monitoring neurocircuits, including three large-scale brain networks that have been implicated in OCD (the frontoparietal network, cingulo-opercular network, and default mode network). Participants were patients with OCD (n = 128) that included both adults (ages 24-45) and adolescents (ages 12-17), as well as unaffected controls (n = 64). Neural assessments included tests of cognitive interference and error processing. We found three patient clusters, reflecting a "normative" cluster that shared a brain activation pattern with unaffected controls (65.9% of clinical participants), as well as an "interference hyperactivity" cluster (15.2% of clinical participants) and an "error hyperactivity" cluster (18.9% of clinical participants). We also related these clusters to demographic and clinical correlates. After post-hoc correction for false discovery rates, the interference hyperactivity cluster showed significantly longer reaction times than the other patient clusters, but no other between-cluster differences in covariates were detected. These findings increase precision in patient characterization, reframe prior neurobehavioral research in OCD, and provide a starting point for neuroimaging-guided treatment selection.

Hyperspectral microscopy in biology and minerals, unsupervised deep learning neural network denoising SRS photos: hyperspectral resolution enhancement and denoising one hyperspectral picture is enough to teach unsupervised method. An intuitive chemical species map for a lithium ore sample is produced using k -means clustering. Many researchers are now interested in biosignals. Uncertainty limits the algorithms’ capacity to evaluate these signals for further information. Even while AI systems can answer puzzles, they remain limited. Deep learning is used when machine learning is inefficient. Supervised learning needs a lot of data. Deep learning is vital in modern AI. Supervised learning requires a large labeled dataset. The selection of parameters prevents over- or underfitting. Unsupervised learning is used to overcome the challenges outlined above (performed by the clustering algorithm). To accomplish this, two processing processes were used: (1) utilizing nonlinear deep learning networks to turn data into a latent feature space ( Z ). The Kullback–Leibler divergence is used to test the objective function convergence. This article explores a novel research on hyperspectral microscopic picture using deep learning and effective unsupervised learning.

The microscopic description of the local structure of water remains an open challenge. Here, we adopt an agnostic approach to understanding water's hydrogen bond network using data harvested from molecular dynamics simulations of an empirical water model. A battery of state-of-the-art unsupervised data-science techniques are used to characterize the free-energy landscape of water starting from encoding the water environment using local atomic descriptors, through dimensionality reduction and finally the use of advanced clustering techniques. Analysis of the free energy under ambient conditions was found to be consistent with a rough single basin and independent of the choice of the water model. We find that the fluctuations of the water network occur in a high-dimensional space, which we characterize using a combination of both atomic descriptors and chemical-intuition-based coordinates. We demonstrate that a combination of both types of variables is needed in order to adequately capture the complexity of the fluctuations in the hydrogen bond network at different length scales both at room temperature and also close to the critical point of water. Our results provide a general framework for examining fluctuations in water under different conditions.

The security of programmable logic controllers (PLCs) that control industrial systems is becoming increasingly critical due to the ubiquity of the Internet of Things technologies and increasingly nefarious cyber-attack activity. Conventional techniques for safeguarding PLCs are difficult due to their unique architectures. This work proposes a one-class support vector machine, one-class neural network interconnected in a feed-forward manner, and isolation forest approaches for verifying PLC process integrity by monitoring PLC memory addresses. A comprehensive experiment is conducted using an open-source PLC subjected to multiple attack scenarios. A new histogram-based approach is introduced to visualize anomaly detection algorithm performance and prediction confidence. Comparative performance analyses of the proposed algorithms using decision scores and prediction confidence are presented. Results show that isolation forest outperforms one-class neural network, one-class support vector machine, and previous work, in terms of accuracy, precision, recall, and F1-score on seven attack scenarios considered. Statistical hypotheses tests involving analysis of variance and Tukey’s range test were used to validate the presented results.

Purpose This paper contributes to the literature by discussing the impact of machine learning (ML) on management accounting (MA) and the management accountant based on three sources: academic articles, papers and reports from accounting bodies and consulting companies. The purpose of this paper is to identify, discuss and provide suggestions for how ML could be included in research and education in the future for the management accountant. Design/methodology/approach This paper identifies three types of studies on the influence of ML on MA issued between 2015 and 2021 in mainstream accounting journals, by professional accounting bodies and by large consulting companies. Findings First, only very few academic articles actually show examples of using ML or using different algorithms related to MA issues. This is in contrast to other research fields such as finance and logistics. Second, the literature review also indicates that if the management accountants want to keep up with the demand of their qualifications, they must take action now and begin to discuss how big data and other concepts from artificial intelligence and ML can benefit MA and the management accountant in specific ways. Originality/value Even though the paper may be classified as inspirational in nature, the paper documents and discusses the revised environment that surrounds the accountant today. The paper concludes by highlighting specifically the necessity of including exploratory data analysis and unsupervised ML in the field of MA to close the existing gaps in both education and research and thus making the MA profession future-proof.

Deformable medical image registration estimates corresponding deformation to align the regions of interest (ROIs) of two images to a same spatial coordinate system. However, recent unsupervised registration models only have correspondence ability without perception, making misalignment on blurred anatomies and distortion on task-unconcerned backgrounds. Label-constrained (LC) registration models embed the perception ability via labels, but the lack of texture constraints in labels and the expensive labeling costs causes distortion internal ROIs and overfitted perception. We propose the first few-shot deformable medical image registration framework, Perception-Correspondence Registration (PC-Reg), which embeds perception ability to registration models only with few labels, thus greatly improving registration accuracy and reducing distortion. 1) We propose the Perception-Correspondence Decoupling which decouples the perception and correspondence actions of registration to two CNNs. Therefore, independent optimizations and feature representations are available avoiding interference of the correspondence due to the lack of texture constraints. 2) For few-shot learning, we propose Reverse Teaching which aligns labeled and unlabeled images to each other to provide supervision information to the structure and style knowledge in unlabeled images, thus generating additional training data. Therefore, these data will reversely teach our perception CNN more style and structure knowledge, improving its generalization ability. Our experiments on three datasets with only five labels demonstrate that our PC-Reg has competitive registration accuracy and effective distortion-reducing ability. Compared with LC-VoxelMorph(lambda=1), we achieve the 12.5%, 6.3% and 1.0% Reg-DSC improvements on three datasets, revealing our framework with great potential in clinical application.

Funder: Fondation Brain Canada; Id: http://dx.doi.org/10.13039/100009408, Funder: Fonds de Recherche du Québec ‐ Santé; Id: http://dx.doi.org/10.13039/501100000156, Funder: Health Canada; Id: http://dx.doi.org/10.13039/501100000008, Funder: Brain Canada Foundation; Id: http://dx.doi.org/10.13039/100009408, Frontotemporal dementia in genetic forms is highly heterogeneous and begins many years to prior symptom onset, complicating disease understanding and treatment development. Unifying methods to stage the disease during both the presymptomatic and symptomatic phases are needed for the development of clinical trials outcomes. Here we used the contrastive trajectory inference (cTI), an unsupervised machine learning algorithm that analyzes temporal patterns in high-dimensional large-scale population datasets to obtain individual scores of disease stage. We used cross-sectional MRI data (gray matter density, T1/T2 ratio as a proxy for myelin content, resting-state functional amplitude, gray matter fractional anisotropy, and mean diffusivity) from 383 gene carriers (269 presymptomatic and 115 symptomatic) and a control group of 253 noncarriers in the Genetic Frontotemporal Dementia Initiative. We compared the cTI-obtained disease scores to the estimated years to onset (age-mean age of onset in relatives), clinical, and neuropsychological test scores. The cTI based disease scores were correlated with all clinical and neuropsychological tests (measuring behavioral symptoms, attention, memory, language, and executive functions), with the highest contribution coming from mean diffusivity. Mean cTI scores were higher in the presymptomatic carriers than controls, indicating that the method may capture subtle pre-dementia cerebral changes, although this change was not replicated in a subset of subjects with complete data. This study provides a proof of concept that cTI can identify data-driven disease stages in a heterogeneous sample combining different mutations and disease stages of genetic FTD using only MRI metrics.

Patients with continuous flow destination therapy (DT) left ventricular assist devices (LVAD) comprise a heterogeneous population. We hypothesized that phenotypic clustering of individuals with DT LVADs by their implantation characteristics will be associated with different long-term risk profiles. We analyzed 5,999 patients with continuous flow DT LVADs in Interagency Registry for Mechanically Assisted Circulatory Support using 18 continuous variable baseline characteristics. We Z-transformed the variables and applied a Gaussian finite mixture model to perform unsupervised clustering resulting in identification of 4 phenogroups. Survival analyses considered the competing risk for cumulative incidence of transplant or the composite end point of death or heart transplant where appropriate. Phenogroup 1 (n = 1,163, 19%) was older (71 years) and primarily white (81%). Phenogroups 2 (n = 648, 11%) and 3 (n = 3,671, 61%) were of intermediate age (70 and 62 years), weight (85 and 87 kg), and ventricular size. Phenogroup 4 (n = 517, 9%) was younger (40 years), heavier (108 kg), and more racially diverse. The cumulative incidence of death, heart transplant, bleeding, LVAD malfunction, and LVAD thrombosis differed among phenogroups. The highest incidence of death and the lowest rate of heart transplant was seen in phenogroup 1 (p0.001). For adverse outcomes, phenogroup 4 had the lowest incidence of bleeding, whereas LVAD device thrombosis and malfunction were lowest in phenogroup 1 (p0.001 for all). Finally, the incidence of stroke, infection, and renal dysfunction were not statistically different. In conclusion, the present unsupervised machine learning analysis identified 4 phenogroups with different rates of adverse outcomes and these findings underscore the influence of phenotypic heterogeneity on post-LVAD implantation outcomes.

RESUMEN Los métodos de inteligencia artificial utilizando herramientas de aprendizaje no supervisado pueden apoyar la resolución de problemas al establecer patrones de agrupación o clasificación no identificados, que permiten tipificar subgrupos para manejos más individualizados. Existen pocos estudios que permiten conocer la influencia de síntomas digestivos y extradigestivos en la tipificación dispepsia funcional; esta investigación realizó un análisis de aprendizaje no supervisado por conglomerados basándose en dichos síntomas, para discriminar subtipos de dispepsia y comparar con una de las clasificaciones actualmente más aceptadas. Se realizó un análisis exploratorio de conglomerados en adultos con dispepsia funcional según síntomas digestivos, extradigestivos y emocionales. Se conformaron patrones de agrupación de tal manera que dentro de cada grupo existiera homogeneidad en cuanto a los valores adoptados por cada variable. El método de análisis de conglomerados fue bietápico y los resultados del patrón de clasificación se compararon con una de las clasificaciones más aceptadas de dispepsia funcional. De 184 casos, 157 cumplieron con criterios de inclusión. El análisis de conglomerados excluyó 34 casos no clasificables. Los pacientes con dispepsia de tipo 1 (conglomerado uno), presentaron mejoría al tratamiento en el 100% de los casos, solo una minoría presentaron síntomas depresivos. Los pacientes con dispepsia de tipo 2 (conglomerado dos) presentaron una mayor probabilidad de falla al tratamiento con inhibidor de bomba de protones, padecieron con mayor frecuencia trastornos de sueño, ansiedad, depresión, fibromialgia, limitaciones físicas o dolor crónico de naturaleza no digestiva. Esta clasificación de dispepsia por análisis de clúster establece una visión más holística de la dispepsia en la cual características extradigestivas, síntomas afectivos, presencia o no de trastornos de sueño y de dolor crónico permiten discriminar el comportamiento y respuesta al manejo de primera línea. ABSTRACT Artificial intelligence methods using unsupervised learning tools can support problem solving by establishing unidentified grouping or classification patterns that allow typing subgroups for more individualized management. There are few studies that allow us to know the influence of digestive and extra-digestive symptoms in the classification of functional dyspepsia. This research carried out a cluster unsupervised learning analysis based on these symptoms to discriminate subtypes of dyspepsia and compare with one of the currently most accepted classifications. An exploratory cluster analysis was carried out in adults with functional dyspepsia according to digestive, extra-digestive and emotional symptoms. Grouping patterns were formed in such a way that within each group there was homogeneity in terms of the values adopted by each variable. The cluster analysis method was two-stage and the results of the classification pattern were compared with one of the most accepted classifications of functional dyspepsia. Of 184 cases, 157 met the inclusion criteria. The cluster analysis excluded 34 unclassifiable cases. Patients with type 1 dyspepsia (cluster one) presented improvement after treatment in 100% of cases, only a minority presented depressive symptoms. Patients with type 2 dyspepsia (cluster two) presented a higher probability of failure to treatment with proton pump inhibitor, suffered more frequently from sleep disorders, anxiety, depression, fibromyalgia, physical limitations or chronic pain of a non-digestive nature. This classification of dyspepsia by cluster analysis establishes a more holistic vision of dyspepsia in which extradigestive characteristics, affective symptoms, presence or absence of sleep disorders and chronic pain allow discriminating behavior and response to first-line management.

Background and objectives: Chronic low back pain (CLBP) is a leading cause of disability. The management guidelines for the management of CLBP often recommend optimizing physical activity (PA). Among a subsample of patients with CLBP, central sensitization (CS) is present. However, knowledge about the association between PA intensity patterns, CLBP, and CS is limited. The objective PA computed by conventional approaches (e.g. cut-points) may not be sensitive enough to explore this association. This study aimed to investigate PA intensity patterns in patients with CLBP and low or high CS (CLBP-, CLBP+, respectively) by using advanced unsupervised machine learning approach, Hidden semi-Markov model (HSMM).Methods: Forty-two patients were included (23 CLBP-, 19 CLBP+). CS-related symptoms (e.g. fatigue, sensitivity to light, psychological features) were assessed by a CS Inventory. Patients wore a standard 3D-accelerometer for one week and PA was recorded. The conventional cut-points approach was used to compute the time accumulation and distribution of PA intensity levels in a day. For the two groups, two HSMMs were developed to measure the temporal organization of and transition between hidden states (PA intensity levels), based on the accelerometer vector magnitude.Results: Based on the conventional cut-points approach, no significant differences were found between CLBP- and CLBP+ groups (p = 0.87). In contrast, HSMMs revealed significant differences between the two groups. For the 5 identified hidden states (rest, sedentary, light PA, light locomotion, and moderate-vigorous PA), the CLBP- group had a higher transition probability from rest, light PA, and moderate-vigorous PA states to the sedentary state (p < 0.001). In addition, the CBLP- group had a significantly shorter bout duration of the sedentary state (p < 0.001). The CLBP+ group exhibited longer durations of active (p < 0.001) and inactive states (p = 0.037) and had higher transition probabilities between active states (p < 0.001).Conclusions: HSMM discloses the temporal organization and transitions of PA intensity levels based on accelerometer data, yielding valuable and detailed clinical information. The results imply that patients with CLBP- and CLBP+ have different PA intensity patterns. CLBP+ patients may adopt the distress-endurance response pattern with a prolonged bout duration of activity engagement.

In rarefied gas dynamics, gas-surface scattering kernels play an important role in evaluating the gas flow field properties since they include crucial information about the effects of the physical and chemical properties of the gas-surface interface on the gas scattering process. In our previous work, an unsupervised machine learning approach, the Gaussian Mixture (GM) approach, is employed to establish a statistical gas-surface scattering model. In this paper, the GM- scattering kernel is coupled to a Direct Simulation Monte Carlo (DSMC) solver as a boundary model to study the Fourier thermal problem. Two gas-solid pairs are considered here: Ar-Au and H2-Ni. To fully examine the coupling mechanism between the GM scattering kernel and the DSMC approach, a one-to-one correspondence between MD and DSMC particles is considered here. Benchmarked by MD results, the performance of the Gaussian Mixture model coupled to DSMC (GM-DSMC) is assessed against the Cercignani-Lampis-Lord (CLL) kernel incorporated into DSMC simulation (CLL-DSMC). To assess the performance of the applied scattering kernels, the number density and temperature profiles between two walls are utilized. The accuracy of the simulation results is determined by measuring the deviations of the DSMC results from full MD simulation results. For the Ar-Au case, the predicted results by DSMC incorporating both scattering models are consistent with the MD data. Here, the average deviations of the predicted density results are 0.6% for both GM-DSMC and CLL-DSMC approaches. In the case of the temperature profile, the trend predicted by the GM-DSMC in the bulk of the domain matches well the pure MD results. Here, the deviations are 0.2%. On the other hand, in most parts of the simulation domain, the results predicted by the CLL-DSMC method deviate from the MD results, observing the highest deviation close to the top wall, which is 4%. For the H2-Ni system,it is observed that within the bulk of the simulation domain, the GM-DSMC approach the results for the number density match well the MD data. Here, the deviations of the GM-DSMC results are around 0.6% on average. However, a notable discrepancy between the density profiles obtained from the MD and CLL-DSMC is observed: the highest deviation is measured near the top wall, which is 8%. The predicted temperature profiles based on both GM-DSMC and CLL-DSMC approaches deviate from the reference MD results. Nevertheless, the GM-DSMC approach still outperforms the CLL-DSMC approach. Herein, the highest deviations based on the GM-DSMC approach are measured at the first and last bins, which are 2% and 1%, respectively. On the other hand, using the CLL-DSMC approach, the deviations on the same bins are 9% and 10%, respectively. These results confirm the better performance of the GM-DSMC approach compared to the CLL-DSMC approach, especially for the diatomic systems.

In rarefied gas dynamics, gas-surface scattering kernels play an important role in evaluating the gas flow field properties since they include crucial information about the effects of the physical and chemical properties of the gas-surface interface on the gas scattering process. In our previous work, an unsupervised machine learning approach, the Gaussian Mixture (GM) approach, is employed to establish a statistical gas-surface scattering model. In this paper, the GM- scattering kernel is coupled to a Direct Simulation Monte Carlo (DSMC) solver as a boundary model to study the Fourier thermal problem. Two gas-solid pairs are considered here: Ar-Au and H2-Ni. To fully examine the coupling mechanism between the GM scattering kernel and the DSMC approach, a one-to-one correspondence between MD and DSMC particles is considered here. Benchmarked by MD results, the performance of the Gaussian Mixture model coupled to DSMC (GM-DSMC) is assessed against the Cercignani-Lampis-Lord (CLL) kernel incorporated into DSMC simulation (CLL-DSMC). To assess the performance of the applied scattering kernels, the number density and temperature profiles between two walls are utilized. The accuracy of the simulation results is determined by measuring the deviations of the DSMC results from full MD simulation results. For the Ar-Au case, the predicted results by DSMC incorporating both scattering models are consistent with the MD data. Here, the average deviations of the predicted density results are 0.6% for both GM-DSMC and CLL-DSMC approaches. In the case of the temperature profile, the trend predicted by the GM-DSMC in the bulk of the domain matches well the pure MD results. Here, the deviations are 0.2%. On the other hand, in most parts of the simulation domain, the results predicted by the CLL-DSMC method deviate from the MD results, observing the highest deviation close to the top wall, which is 4%. For the H2-Ni system,it is observed that within the bulk of the simulation domain, the GM-DSMC approach the results for the number density match well the MD data. Here, the deviations of the GM-DSMC results are around 0.6% on average. However, a notable discrepancy between the density profiles obtained from the MD and CLL-DSMC is observed: the highest deviation is measured near the top wall, which is 8%. The predicted temperature profiles based on both GM-DSMC and CLL-DSMC approaches deviate from the reference MD results. Nevertheless, the GM-DSMC approach still outperforms the CLL-DSMC approach. Herein, the highest deviations based on the GM-DSMC approach are measured at the first and last bins, which are 2% and 1%, respectively. On the other hand, using the CLL-DSMC approach, the deviations on the same bins are 9% and 10%, respectively. These results confirm the better performance of the GM-DSMC approach compared to the CLL-DSMC approach, especially for the diatomic systems.

Data for "Sonnewald, M., Reeve, K., and Lguensat, R., The Southern Ocean supergyre: a unifying dynamical framework identified by machine learning. In press, Commun Earth Environ." Included: Barotropic vorticity terms, labels arrived at in Sonnewald et al., 2019, additional fields for plotting for the readers convenience. Additional data for the ECCOv4 model can be found here including velocities. Sonnewald, M., Wunsch, C. and Heimbach, P. Unsupervised Learning Reveals Geography of Global Ocean Dynamical Regions, 2019, Journal of Earth and Space Science., {"references":["Sonnewald et al., 2023, The Southern Ocean supergyre: a unifying dynamical framework identified by machine learning. Available: Researchsquare"]}

In rarefied gas dynamics, gas-surface scattering kernels play an important role in evaluating the gas flow field properties since they include crucial information about the effects of the physical and chemical properties of the gas-surface interface on the gas scattering process. In our previous work, an unsupervised machine learning approach, the Gaussian Mixture (GM) approach, is employed to establish a statistical gas-surface scattering model. In this paper, the GM- scattering kernel is coupled to a Direct Simulation Monte Carlo (DSMC) solver as a boundary model to study the Fourier thermal problem. Two gas-solid pairs are considered here: Ar-Au and H2-Ni.To fully examine the coupling mechanism between the GM scattering kernel and the DSMC approach, a one-to-one correspondence between MD and DSMC particles is considered here. Benchmarked by MD results, the performance of the Gaussian Mixture model coupled to DSMC (GM-DSMC) is assessed against the Cercignani-Lampis-Lord (CLL) kernel incorporated into DSMC simulation (CLL-DSMC). To assess the performance of the applied scattering kernels, the number density and temperature profiles between two walls are utilized. The accuracy of the simulation results is determined by measuring the deviations of the DSMC results from full MD simulation results.For the Ar-Au case, the predicted results by DSMC incorporating both scattering models are consistent with the MD data. Here, the average deviations of the predicted density results are 0.6% for both GM-DSMC and CLL-DSMC approaches. In the case of the temperature profile, the trend predicted by the GM-DSMC in the bulk of the domain matches well the pure MD results. Here, the deviations are 0.2%. On the other hand, in most parts of the simulation domain, the results predicted by the CLL-DSMC method deviate from the MD results, observing the highest deviation close to the top wall, which is 4%.For the H2-Ni system,it is observed that within the bulk of the simulation domain, the GM-DSMC approach the results for the number density match well the MD data. Here, the deviations of the GM-DSMC results are around 0.6% on average. However, a notable discrepancy between the density profiles obtained from the MD and CLL-DSMC is observed: the highest deviation is measured near the top wall, which is 8%. The predicted temperature profiles based on both GM-DSMC and CLL-DSMC approaches deviate from the reference MD results. Nevertheless, the GM-DSMC approach still outperforms the CLL-DSMC approach. Herein, the highest deviations based on the GM-DSMC approach are measured at the first and last bins, which are 2% and 1%, respectively. On the other hand, using the CLL-DSMC approach, the deviations on the same bins are 9% and 10%, respectively.These results confirm the better performance of the GM-DSMC approach compared to the CLL-DSMC approach, especially for the diatomic systems.

Objectives To identify and evaluate predictive lung imaging markers and their pathways of change during progression of idiopathic pulmonary fibrosis (IPF) from sequential data of an IPF cohort. To test if these imaging markers predict outcome. Methods We studied radiological disease progression in 76 patients with IPF, including overall 190 computed tomography (CT) examinations of the chest. An algorithm identified candidates for imaging patterns marking progression by computationally clustering visual CT features. A classification algorithm selected clusters associated with radiological disease progression by testing their value for recognizing the temporal sequence of examinations. This resulted in radiological disease progression signatures, and pathways of lung tissue change accompanying progression observed across the cohort. Finally, we tested if the dynamics of marker patterns predict outcome, and performed an external validation study on a cohort from a different center. Results Progression marker patterns were identified and exhibited high stability in a repeatability experiment with 20 random sub-cohorts of the overall cohort. The 4 top-ranked progression markers were consistently selected as most informative for progression across all random sub-cohorts. After spatial image registration, local tracking of lung pattern transitions revealed a network of tissue transition pathways from healthy to a sequence of disease tissues. The progression markers were predictive for outcome, and the model achieved comparable results on a replication cohort. Conclusions Unsupervised learning can identify radiological disease progression markers that predict outcome. Local tracking of pattern transitions reveals pathways of radiological disease progression from healthy lung tissue through a sequence of diseased tissue types. Key Points • Unsupervised learning can identify radiological disease progression markers that predict outcome in patients with idiopathic pulmonary fibrosis. • Local tracking of pattern transitions reveals pathways of radiological disease progression from healthy lung tissue through a sequence of diseased tissue types. • The progression markers achieved comparable results on a replication cohort.

The ability to anticipate team members' actions enables joint action towards a common goal. Task knowledge and mental simulation allow for anticipating other agents' actions and for making inferences about their underlying mental representations. In human-AI teams, providing AI agents with anticipatory mechanisms can facilitate collaboration and successful execution of joint action. This paper presents a computational cognitive model demonstrating mental simulation of operators' mental models of a situation and anticipation of their behavior. The work proposes two successive steps: (1) A hierarchical cluster algorithm is applied to recognize patterns of behavior among pilots. These behavioral clusters are used to derive commonalities in situation models from empirical data (N = 13 pilots). (2) An ACT-R (adaptive control of thought - rational) cognitive model is implemented to mentally simulate different possible outcomes of action decisions and timing of a pilot. model tracing of ACT-R allows following up on operators' individual actions. Two models are implemented using the symbolic representations of ACT-R: one simulating normative behavior and the other by simulating individual differences and using subsymbolic learning. Model performance is analyzed by a comparison of both models. Results indicate the improved performance of the individual differences over the normative model and are discussed regarding implications for cognitive assistance capable of anticipating operator behavior.

To treat large intracranial aneurysms, pipeline embolization device (PED) stent with unsupervised learning algorithms was utilized. Unsupervised learning model algorithm was used to screen aneurysm health big data, find aneurysm blood flow and PED stent positioning characteristic parameters, and guide PED stent treatment of intracranial aneurysms. The research objects were 100 patients with intracranial large aneurysm admitted to X Hospital of X Province from June 2020 to June 2021, who were enrolled into two groups. One group used the prototype transfer generative adversarial network (PTGAN) model to measure mean blood flow and mean vascular pressure and guide the placement of PED stents (PTGAN group). The other group did not use the model to place PED (control group). The PTGAN model can learn feature information from horizontal and vertical directions, with smooth edges and prominent features, which can effectively extract the main morphological and texture features of aneurysms. Compared with the convolutional neural network (CNN) model, the accuracy of the PTGAN model increased by 8.449% (87.452%–79.003%), and the precision increased by 8.347% (91.23%–82.883%). The recall rate increased by 7.011% (87.231%–80.22%), and the F1 score increased by 8.09% (89.73%–81.64%). After the adoption of the PTGAN model, the average blood flow inside the aneurysm body was 0.22 (m/s). After the adoption of the CNN model, the average blood flow inside the aneurysm body was 0.21 (m/s), and the difference was 0.01 (m/s), which was considerable ( p < 0.05 ). Through this research, it was found that the PTGAN model was better than the CNN model in terms of accuracy, precision, recall, and F1 score values. The PTGAN model was better than the CNN model in detecting the average blood flow rate and average blood pressure after treatment, and the blood flowed smoothly. Postoperative complications and postoperative relief were also better than those of the control group. In summary, based on the unsupervised learning algorithm, the PED stent had a good adoption effect in the treatment of intracranial aneurysms and was suitable for subsequent treatment.

Although the availability of various sequencing technologies allows us to capture different genome properties at single-cell resolution, with the exception of a few co-assaying technologies, applying different sequencing assays on the same single cell is impossible. Single-cell alignment using optimal transport (SCOT) is an unsupervised algorithm that addresses this limitation by using optimal transport to align single-cell multiomics data. First, it preserves the local geometry by constructing a k-nearest neighbor (k-NN) graph for each data set (or domain) to capture the intra-domain distances. SCOT then finds a probabilistic coupling matrix that minimizes the discrepancy between the intra-domain distance matrices. Finally, it uses the coupling matrix to project one single-cell data set onto another through barycentric projection, thus aligning them. SCOT requires tuning only two hyperparameters and is robust to the choice of one. Furthermore, the Gromov-Wasserstein distance in the algorithm can guide SCOT's hyperparameter tuning in a fully unsupervised setting when no orthogonal alignment information is available. Thus, SCOT is a fast and accurate alignment method that provides a heuristic for hyperparameter selection in a real-world unsupervised single-cell data alignment scenario. We provide a tutorial for SCOT and make its source code publicly available on GitHub.

Additive manufacturing (AM) of high-strength metals, which is typically based on laser powder bed fusion (LPBF), can introduce microscopic pores in the AM metal. Pulsed Infrared Thermography (PIT) offers several advantages for nondestructive imaging of subsurface defects in AM structures because the method is one-sided, non-contact and scalable to structures of arbitrary size. However, high-resolution PIT imaging results in the generation of a large volume of thermography data (~TB), which creates challenges for the storage and transmission of data. Compression of thermography data requires an approach that achieves high data compression ratio while preserving weak thermal features corresponding to microscopic material defects. We investigate thermography data compression using several unsupervised learning (UL) algorithms, which include Principal Component Analysis (PCA), Independent Component Analysis (ICA), Exploratory Factor Analysis (EFA), Sparse Dictionary Learning (SDL), and a novel lightweight Thermography Compressive Sparse Autoencoder (TCSA). Algorithms are benchmarked using PIT experimental data obtained from imaging of a stainless steel plate with calibrated porosity defects imprinted with AM process. For all algorithms, we obtain compression ratio >30 (highest compression of 46 is achieved with TCSA), and peak signal-to-noise ratio for reconstruction accuracy >73dB. Compared to existing methods, advantages of UL algorithms include achieving high compression ratio while preserving weak features to allow extraction of microscopic material defects from images. UL-based methods have general applicability because they are adaptable to compression of different data types, and allow for memory-efficient training and rapid on-line augmentation of the model.

Background: Mutation processes leave different signatures in genes. For single-base substitutions, previous studies have suggested that mutation signatures are not only reflected in mutation bases but also in neighboring bases. However, because of the lack of a method to identify features of long sequences next to mutation bases, the understanding of how flanking sequences influence mutation signatures is limited.Methods: We constructed a long short-term memory – self organizing map (LSTM-SOM) unsupervised neural network. By extracting mutated sequence features via LSTM and clustering similar features with the SOM, single-base substitutions in The Cancer Genome Atlas database were clustered according to both their mutation site and flanking sequences. The relationship between mutation sequence signatures and clinical features was then analyzed. Finally, we clustered patients into different classes according to the composition of the mutation sequence signatures by the K-means method and then studied the differences in clinical features and survival between classes.Results: Ten classes of mutant sequence signatures (mutation blots, MBs) were obtained from 2,141,527 single-base substitutions via LSTM-SOM machine learning approach. Different features in mutation bases and flanking sequences were revealed among MBs. MBs reflect both the site and pathological features of cancers. MBs were related to clinical features, including age, gender, and cancer stage. The class of an MB in a given gene was associated with survival. Finally, patients were clustered into 7 classes according to the MB composition. Significant differences in survival and clinical features were observed among different patient classes.Conclusions: We provided a method for analyzing the characteristics of mutant sequences. Result of this study showed that flanking sequences, together with mutation bases, shape the signatures of SBSs. MBs were shown related to clinical features and survival of cancer patients. Composition of MBs is a feasible predictive factor of clinical prognosis. Further study of the mechanism of MBs related to cancer characteristics is suggested.

Objectives: This study describes an unsupervised machine learning approach used to estimate the homeostatic model assessment-insulin resistance (HOMA-IR) cut-off for identifying subjects at risk of IR in a given ethnic group based on the clinical data of a representative sample. Methods: The approach was applied to analyse the clinical data of individuals with Arab ancestors, which was obtained from a family study conducted in Nizwa, Oman, between January 2000 and December 2004. First, HOMA-IR-correlated variables were identified to which a clustering algorithm was applied. Two clusters having the smallest overlap in their HOMA-IR values were retrieved. These clusters represented the samples of two populations, which are insulin-sensitive subjects and individuals at risk of IR. The cut-off value was estimated from intersections of the Gaussian functions, thereby modelling the HOMA-IR distributions of these populations. Results: A HOMA-IR cut-off value of 1.62 ± 0.06 was identified. The validity of this cut-off was demonstrated by showing the following: 1) that the clinical characteristics of the identified groups matched the published research findings regarding IR; 2) that a strong relationship exists between the segmentations resulting from the proposed cut-off and those resulting from the two-hour glucose cut-off recommended by the World Health Organization for detecting prediabetes. Finally, the method was also able to identify the cut-off values for similar problems (e.g. fasting sugar cut-off for prediabetes). Conclusion: The proposed method defines a HOMA-IR cut-off value for detecting individuals at risk of IR. Such methods can identify high-risk individuals at an early stage, which may prevent or delay the onset of chronic diseases such as type 2 diabetes. Keywords: Unsupervised Machine Learning; Cluster Analysis; Insulin Resistance; Diabetes Mellitus, Type II.

Previous studies on enhancers and their target genes were largely based on bulk samples that represent ‘average’ regulatory activities from a large population of millions of cells, masking the heterogeneity and important effects from the sub-populations. In recent years, single-cell sequencing technology has enabled the profiling of open chromatin accessibility at the single-cell level (scATAC-seq), which can be used to annotate the enhancers and promoters in specific cell types. A comprehensive resource is highly desirable for exploring how the enhancers regulate the target genes at the single-cell level. Hence, we designed a single-cell database scEnhancer (http://enhanceratlas.net/scenhancer/), covering 14 527 776 enhancers and 63 658 600 enhancer-gene interactions from 1 196 906 single cells across 775 tissue/cell types in three species. An unsupervised learning method was employed to sort and combine tens or hundreds of single cells in each tissue/cell type to obtain the consensus enhancers. In addition, we utilized a cis-regulatory network algorithm to identify the enhancer-gene connections. Finally, we provided a user-friendly platform with seven useful modules to search, visualize, and browse the enhancers/genes. This database will facilitate the research community towards a functional analysis of enhancers at the single-cell level.

Current technologies allow the sequencing of microbial communities directly from the environment without prior culturing. One of the major problems when analyzing a microbial sample is to taxonomically annotate its reads to identify the species it contains. The major difficulties of taxonomic analysis are the lack of taxonomically related genomes in existing reference databases, the uneven abundance ratio of species, and sequencing errors. Microbial communities can be studied with reads clustering, a process referred to as genome binning. In this study, we present MetaProb 2 an unsupervised genome binning method based on reads assembly and probabilistic k-mers statistics. The novelties of MetaProb 2 are the use of minimizers to efficiently assemble reads into unitigs and a community detection algorithm based on graph modularity to cluster unitigs and to detect representative unitigs. The effectiveness of MetaProb 2 is demonstrated in both simulated and real datasets in comparison with state-of-art binning tools such as MetaProb, AbundanceBin, Bimeta, and MetaCluster. On real datasets, it is the only one capable of producing promising results while being parsimonious with computational resources.

Clustering tumor metastasis samples from gene expression data at the whole genome level remains an arduous challenge, in particular, when the number of experimental samples is small and the number of genes is huge. We focus on the prediction of the epithelial-mesenchymal transition (EMT), which is an underlying mechanism of tumor metastasis, here, rather than tumor metastasis itself, to avoid confounding effects of uncertainties derived from various factors. In this paper, we propose a novel model in predicting EMT based on multidimensional scaling (MDS) strategies and integrating entropy and random matrix detection strategies to determine the optimal reduced number of dimension in low dimensional space. We verified our proposed model with the gene expression data for EMT samples of breast cancer and the experimental results demonstrated the superiority over state-of-the-art clustering methods. Furthermore, we developed a novel feature extraction method for selecting the significant genes and predicting the tumor metastasis. The source code is available at "https://github.com/yushanqiu/yushan.qiu-szu.edu.cn".

Background: The diagnosis and therapy of heart failure are guided mainly by a single imaging parameter, the left ventricular ejection fraction (LVEF). Recent studies have reported on the value of machine learning in characterising the various phenotypes of heart failure patients. Therefore, this study aims to use unsupervised machine learning algorithms to phenotype heart failure patients into different clusters using multiple clinical parameters. Methods: Seven unsupervised machine learning clustering algorithms were used to cluster heart failure patients hospitalised with acute and chronic heart failure. Results: The agglomerative clustering algorithm identified three clusters with a silhouette score of 0.72. Cluster 1 (uraemic cluster) comprised 229 (36.0%) patients with a mean age of 56.2 ± 17.2 years and a serum urea of 14.5 ± 31.3 mmol/L. Cluster 2 (hypotensive cluster) comprised 117 (18.4%) patients with a minimum systolic and diastolic blood pressure of 91 and 60 mmHg, respectively. In cluster 3 (congestive cluster), patients predominantly had symptoms of fluid overload, and 93 (64.6%) patients had ascites. Among the 636 heart failure patients studied, the median LVEF was 32% (interquartile range: 25–45), and the rate of in-hospital all-cause mortality was 14.5%. Systolic and diastolic blood pressure, age, and the LVEF had the most substantial impact on discriminating between the three clusters. Conclusions: Clinicians without access to echocardiography could potentially rely on blood pressure measurements and age to risk stratify heart failure patients. However, larger prospective studies are mandatory for the validation of these clinical parameters.

This paper uses agglomerative hierarchical cluster analysis to group 2778 farming-defined counties into six clusters, revealing farm patterns across the contiguous 48 states of the United States. To understand the differences in economic performance and improve farm households’ well-being, economists have endeavored to identify patterns in US farming. The US is a leading global producer and exporter of many agricultural and food products. Our primary objective is to construct a policy-relevant farm clustering to characterize agricultural homogeneity in US farms’ production potential. We identify six clusters that are relatively homogeneous in five dimensions: farm size, farm assets, farm labor, farm output, degree of mechanization, and government programs. Minimizing diversity within a cluster allows for analysis of public policy changes on specific clusters and comparison of differential effects of the change across clusters.

The use of machine learning in climate science is expanding rapidly after its success in other fields. The more powerful data driven models of machine learning show promise to give us more accurate predictions of and potentially more insights into the physical climate system than conventional statistical models. Many new applications for machine learning in climate science have been proposed in recent years, although it is uncertain which of them will prove to be fruitful lines of research. In this thesis, we explore two well-established problems in statistical climate science and approach them using unsupervised machine learning techniques, namely climate mode extraction and the bias correction of simulations. In the first part of this thesis, we develop a framework to test methods of mode extraction on climate-like data. We generate imitation global climate fields, which include as much of the expected complexity of real data as possible. We find that the newer mode extraction methods, which brand themselves as machine learning rather than conventional statistics, outperform the conventional approaches by more accurately extracting the known modes in the data. When applied to reanalysis and model surface temperature data, the newer methods extract a well-constrained ENSO signal and warming trend than the classical methods. However, no method can safeguard against generating false modes in all circumstances. We show the consequences of false mode extraction with examples and suggest how incorrectly extracted modes could lead to false proposed modal mechanisms. In the second part, we explore the use of unsupervised deep neural networks for bias correcting large, simulated climate fields of multiple variables. We show that the structure of neural networks allows for more faithful bias correction of cross-variable and spatial correlations when used to map between simulations and reanalysis data. A particular advantage of these methods is that they can shift misplaced climate features. We also show the limitation of these networks and how the lack of constraints in optimising them can lead to unexpected bias correction results. By exploring these uses of unsupervised machine learning in climate science we hope that can be developed further and be utilised more by researchers.

Hovedbudskap Maskinlæring (ML) er et satsingsområde for klynge for vurdering av tiltak, Område for helsetjenester, FHI. Høsten 2021 overtok ML 2.0 arbeidet etter ML 1.0. Denne rapporten beskriver ML 2.0 teamet sitt arbeid, resultater og erfaringer. ML team 2.0s nøkkelprestasjoner inkluderer: • gjennomføring av en intens maskinlæringsuke med mål om kompetanseheving i hele klyngen, • oppstart av en studie som vil anslå ressursbesparelsene ved bruk av ML i kunnskapsoppsummeringer, • formidling av vårt ML-arbeid i internasjonale fora, som har tydeliggjort det unike arbeidet ML laget har bidratt med inn i kunnskapsoppsummeringsarbeidet i klyngen, • støtte innovative bruksområder for ML innen metodevurderinger og oppdateringer, • utarbeidelse av to finansieringssøknader, • bidrag inn i andre nasjonale og europeiske finansieringssøknader. ML-teamet har bidratt til en gradvis tilpassing av klyngens metoder til mer effektive arbeidsflytprosesser, noe som nå merkes i ressursbesparelser i klyngens prosjekter. Ressursbesparelsene gjorde at vi kunne levere flere rapporter enn mulig ved bruk av kun tradisjonelle metoder eller bruke mer tid på andre deler av prosessen. ML 2.0 fremstår fortsatt som et unikt og innovativt team som leder og tilrettelegger for implementering av ML innenfor kunnskapsoppsummering. Nye ML-aktiviteter knyttet til utforskning og evaluering av nye funksjoner, verktøy eller emner vil gi et mer åpent og flytende ML-miljø der enhver ansatt kan føle mestring og eierskap over ML-funksjoner eller -verktøy. Disse faktorene er avgjørende for Folkehelseinstituttets evne til å tilpasse seg og fortsette å utmerke seg i kunnskapsoppsummeringsfeltet.

IT-Sicherheitsbedrohungen können für Unternehmen schwerwiegende Folgen mitsich bringen, wie beispielsweise Einnahmeverluste, Rufschädigung, Diebstahl von geschützten Daten und Kundeninformationen. Die meisten Computersysteme zeichnenEreignisse in Logdateien auf, die in vielen Fällen eine zeitliche Komponente enthalten.Der Typ und die Struktur der Protokolldateien variieren stark je nach System undPlattform. Die Erkennung von IT-Sicherheitsbedrohungen aus diesen Logdateienstellt sich oft als schwierig heraus, da die Aufzeichnungen meist große Datenmengenumfassen. Die Verarbeitung einer solchen großen Datenmenge in Echtzeit ist durchrein menschliche Analyse nur begrenzt möglich, was zur Folge hat, dass aktuelle Vorfälle in Bezug auf Verfügbarkeit und Sicherheit oft spät erkannt werden. Mit den Fortschritten im Bereich des maschinellen Lernens ergeben sich neue Toolsets und Möglichkeiten zur Überwachung von Fehlern und Anomalien in Logdaten.Das Ziel der vorliegenden Arbeit ist es, zu untersuchen, ob sich Methoden des unüberwachten maschinellen Lernens zur Identifikation von Sicherheitsrisiken durchdie Analyse von Logdateien eignen. Zudem wird ein Vergleich zwischen den Implementierungsmöglichkeiten für Logfileanalyse mittels Splunk und in Python verwendeten Methoden vorgenommen.Um die Forschungsfragen zu beantworten, wurden neben einer Literaturrechercheunterschiedliche Methoden zur Anomalie-Erkennung in Zeitreihen am Beispiel desARCS Log Data Sets getestet. Spezifisch kamen die Algorithmen Isolation Forest, Autoencoder, Seasonal and Trend Decomposition sowie Forecasting in einer PythonUmgebung und die Methode des Kalman-Filters in einer Splunk-Umgebung zum Einsatz. Es wurde untersucht, ob sich anhand der Algorithmen Anomalien in den Logdaten des ASCR Datensatzes identifizieren lassen. Des Weiteren wurde die Identifizierung einer synthetisch injizierten Password-Spraying-Attacke mittels Autoencoderbehandelt.Aus den Ergebnissen dieser Arbeit lässt sich schlussfolgern, dass Datenlecks zu dengrößten IT-Sicherheitsrisiken für IT-gestützte Unternehmen zählen. Zu den Ursachenvon Datenlecks gehören unter anderem Phishing- und Password-Spraying-Angriffe.Die evaluierten Methoden sind im Allgemeinen in der Lage, Anomalien in den ausLogdateien abgeleiteten Zeitreihen zu erkennen. Auch die Identifizierung von IT-Sicherheitsrisiken mithilfe eines unüberwachten maschinellen Lernansatzes wird in derArbeit praktisch umgesetzt und literarisch belegt, wenngleich die Effizienz der praktischen Umsetzung nicht tiefergehend untersucht wird.Weiterhin wurde festgestellt, dass die Umsetzung von Analysemethoden mittels Python eine größere Vielfalt an Möglichkeiten und einen vergleichsweise einfachen Datenimportprozess bietet. Im Gegensatz dazu bieten sich für die Analyse mittels Splunkweniger Möglichkeiten zur Analyse der Logdateien, jedoch sind die implementiertenTools der Plattform weniger komplex in der Handhabung. IT security threats pose serious consequences for companies, such as loss of revenue,damage to reputation, theft of protected data, and customer information. Most computer systems record events in log files, which typically include a temporal component. The type and structure of log files vary widely depending on the system andplatform. Detecting IT security threats from these log files often proves difficult because the records usually consist of large volumes of data. Processing such vastamounts of data in realtime through purely human analysis is only possible to a limited extent, meaning that current incidents related to availability and security are often detected late. Advances in the field of machine learning introduce new toolsetsand possibilities for monitoring errors and anomalies in log data.The goal of this paper is to determine whether unsupervised machine learning methods are suitable for identifying security risks by analyzing log files. Additionally, acomparison between the implementation options for log file analysis is made, focusing on Splunk and the methods used in Python.To answer the research questions, in addition to a literature review, the feasibility ofdifferent methods for anomaly detection in time series was tested using the ARCSLog Data Set as an example. Specifically, the algorithms Isolation Forest, Autoencoders, Seasonal and Trend Decomposition, and Forecasting were employed in a Pythonenvironment, as well as the Kalman Filter method in a Splunk environment. It wasinvestigated whether the algorithms could identify anomalies in the log data of theASCR dataset. Moreover, the identification of a synthetically injected passwordspraying attack using Autoencoder was addressed.From the results of this work, it can be concluded that data leakage is one of the mostsignificant IT security risks for IT-based organizations. Among the causes of data leakage are phishing and password spraying attacks.The evaluated methods are generally capable of detecting anomalies in the time series derived from log files. Additionally, the identification of IT security risks using anunsupervised machine learning approach is practically implemented and supportedby literature in the thesis, although the efficacy of the practical implementation is notinvestigated in depth.Furthermore, it was found that the implementation of analysis methods using Pythonoffers a greater variety of possibilities and a comparatively simple data import process. Conversely, there were fewer options for analyzing log files using Splunk, butthe tools implemented on the platform are less complex to use. Abweichender Titel laut Übersetzung von der:dem Verfasser:in Bachelorarbeit FH JOANNEUM 2023

The study examines language dimensions of mathematical word problems and the classification of mathematical word problems according to these dimensions with unsupervised machine learning (ML) techniques. Previous research suggests that the language dimensions are important for mathematical word problems because it has an influence on the linguistic complexity of word problems. Depending on the linguistic complexity students can have language obstacles to solve mathematical word problems. A lot of research in mathematics education research focus on the analysis on the linguistic complexity based on theoretical build language dimensions. To date, however it has been unclear what empirical relationship between the linguistic features exist for mathematical word problems. To address this issue, we used unsupervised ML techniques to reveal latent linguistic structures of 17 linguistic features for 342 mathematical word problems and classify them. The models showed that three- and five-dimensional linguistic structures have the highest explanatory power. Additionally, the authors consider a four-dimensional solution. Mathematical word problem from the three-dimensional solution can be classify in two groups, three- and five-dimensional solutions in three groups. The findings revealed latent linguistic structures and groups that could have an implication of the linguistic complexity of mathematical word problems and differ from language dimensions, which are considered theoretically. Therefore, the results indicate for new design principles for interventions and materials for language education in mathematics learning and teaching.

International audience; The objective of this study is to improve the inter-pretability of a previous unsupervised clustering analysis of the CRT response through a physiological model-based approach. The developed clustering approach was applied on 250 CRT candidates based on clinical, original and classical echocardiographic features. Patient-specific computational models were proposed for patients associated of each cluster barycenter in order to provide an ex-plainable analysis in relation with physiological mecha-nisms. Five phenogroups were identified from the clustering approach with response rates ranging from 50% to 92.7%. Concerning the model-based approach, a match was observed between the 16 experimental and simulated myocardial strain curves pattern with a mean RMSE of 3.97%(± 1.74) on the five patients. Moreover, the identified model parameters provide us information about the mecano-electrical coupling and tissue properties. The gain of information provides by the parameters model identification, added to the clinical and classical echocar-diographic features is promising for an understanding of LV mechanical dyssynchrony and the identification of patients suitable for CRT. © 2022 Creative Commons.

Electronic Commerce is interaction of carrying on with work through PC organizations. An individual sitting on his seat before PC can get to every one of the offices of the web to trade the items. Not at all like conventional business that is completed genuinely with exertion of an individual to proceed to get items, web-based business has made it more straightforward for human to diminish actual work and to save time. In this undertaking we look to carry out an internet-based clothing store. As we know that each business these days is moving towards web-based business model, so remembering that we look to carry out an internet-based store utilizing web advancement tech stack and innovations. The fundamental target of this application is to make it intelligent and its convenience. We trust that this undertaking, whenever carried out, an association can keep up with mechanized records in secure information bases without repetitive sections.

Robust, fault tolerant, and available systems are fundamental for the adoption of Internet of Things (IoT) in critical domains, such as finance, health, and safety. The IoT infrastructure is often used to collect a large amount of data to meet the business demands of Smart Cities, Industry 4.0, and Smart Home, but there is a opportunity to use these data to intrinsically monitor an IoT system in an autonomous way. A Test Driven Development (TDD) approach for automatic module assessment for ESP32 and ESP8266 IoT development devices based on unsupervised Machine Learning (ML) is proposed to monitor IoT device status. A framework consisting of business drivers, non-functional requirements, engineering view, dynamic system evaluation, and recommendations phases is proposed to be used with the TDD development tool. The proposal is evaluated in academic and smart home study cases with 25 devices, consisting of 15 different firmware versions collected in one week, with a total of over 550,000 IoT status readings. The K-Means algorithm was applied to free memory available, internal temperature, and Wi-Fi level metrics to automatically monitor the IoT devices under development to identify device constraints violation and provide insights for monitoring frequency configuration of different firmware versions. To the best of the authors’ knowledge, it is the first TDD approach for IoT module automatic assessment which uses machine learning based on the real testbed data. The IoT status monitoring and the Python scripts for model training and inference with K-Means algorithm are available under a Creative Commons license.

Mestrado Bolonha em Finanças The COVID-19 pandemic was the infectious disease outbreak that has had the strongest impact on the U.S. stock market. In this dissertation, we study how this impact affected some of the conditional volatilities of S&P 500 industries, using a new model feature-based clustering method on a fitted threshold generalised autoregressive conditional heteroscedasticity (TGARCH) specification. Rather than using the estimated model parameters to compute a distance matrix for the stock indices, which cannot capture all the information about the dependence of the time-varying variance, we suggest using a distance based on the autocorrelations of the estimated conditional volatilities. Both hierarchical (complete linkage) and non-hierarchical (k-means) unsupervised machine learning algorithms are used to assign the set of industries into clusters. The results show a clear change in the composition of each cluster between the period before the first U.S. COVID-19 case and the period during the pandemic, leading to the conclusion that the similarities or distances between industries underwent a significant change, with the industries most affected by the pandemic being Hotels, Consumer Durable & Apparel, Automobile, and Airlines. It was also made an analysis regarding the forecast accuracy of simple asymmetric GARCH models applied to S&P 500 industries and use the model forecast errors for different horizons to calculate a distance matrix for the stock indices. A hierarchical clustering algorithm is used to assign the set of industries into clusters. We found homogeneous clusters of industries in terms of the impact of COVID-19 on US stock market volatility. The industries most affected by the pandemic and with less accurate stock market prediction (Hotels, Resorts & Cruise Lines, Airline, Apparel, Accessories & Luxury Goods, and Automobile) are separated in Euclidean distance from those industries that were less impacted by COVID-19 and which had more accurate forecasting (Pharmaceuticals, Internet & Direct Marketing Retail, Data Processing & Outsourcing Services, and Movies & Entertainment). A pandemia do COVID-19 foi o surto com maior impacto de sempre no mercado de ações dos EUA. Nesta dissertação, estudamos como este impacto afetou algumas das volatilidades condicionadas das indústrias do S&P 500, usando métodos de clustering aplicados numa especificação de um modelo derivado (threshold) do generalizado autorregressivo condicional e heterocedástico (TGARCH). OS parâmetros do modelo estimado não capturam toda a informação acerca da dependência da variância não constante ao longo do tempo pelo que usamos distância baseada nas autocorrelações das volatilidades condicionais, para calcular a matriz de distância para os índices de ações. Ambos os algorítmos hierárquicos (complete linkage) e não-hierárquicos (k-means), técnicas de unsupervised machine learning, são usados para agrupar conjuntos de indústrias em clusters. Os resultados mostram uma clara mudança na composição de cada cluster entre o período anterior ao primeiro caso de COVID-19 nos EUA e o período durante a pandemia. As semelhanças ou distâncias entre as indústrias sofreram uma mudança significativa, sendo as indústrias mais afetadas pela pandemia, os Hotéis, Bens de Consumo Duráveis e Vestuário, Automóveis e Companhias Aéreas. Foi feita uma análise quanto à precisão da previsão de modelos GARCH assimétricos simples aplicados a indústrias do S&P 500. Esses erros de previsão calculados em diferentes intervalos de tempo, durante a pandemia, foram usados para calcular a matriz de distâncias entre os indices de mercado. Um algoritmo de hierarchical clustering foi usado para agrupar um grupo de indústrias em clusters. Obtivemos clusters homogéneos em termos do impacto do COVID-19 na volatilidade do mercado de ações nos EUA. As indústrias mais afectadas pela pandemia, nas quais os modelos preditivos se mostraram menos precisos em termos preditivos (Hotels, Resorts & Cruise Lines, Airline, Apparel, Acessories & Luxury Goods, and Automobile) estão separadas por uma distância Euclideana das indústrias que foram menos impactadas pelo COVID-19 e as previsões mais precisas (Pharmaceuticals, Internet, Data Processing, e Movies & Entertainment). info:eu-repo/semantics/publishedVersion

Never before such a vast amount of data, including genome sequencing, has been collected for any viral pandemic than for the current case of COVID-19. This offers the possibility to trace the virus evolution and to assess the role mutations play in its spread within the population, in real time. To this end, we focused on the Spike protein for its central role in mediating viral outbreak and replication in host cells. Employing the Levenshtein distance on the Spike protein sequences, we designed a machine learning algorithm yielding a temporal clustering of the available dataset. From this, we were able to identify and define emerging persistent variants that are in agreement with known evidences. Our novel algorithm allowed us to define persistent variants as chains that remain stable over time and to highlight emerging variants of epidemiological interest as branching events that occur over time. Hence, we determined the relationship and temporal connection between variants of interest and the ensuing passage to dominance of the current variants of concern. Remarkably, the analysis and the relevant tools introduced in our work serve as an early warning for the emergence of new persistent variants once the associated cluster reaches 1% of the time-binned sequence data. We validated our approach and its effectiveness on the onset of the Alpha variant of concern. We further predict that the recently identified lineage AY.4.2 (‘Delta plus’) is causing a new emerging variant. Comparing our findings with the epidemiological data we demonstrated that each new wave is dominated by a new emerging variant, thus confirming the hypothesis of the existence of a strong correlation between the birth of variants and the pandemic multi-wave temporal pattern. The above allows us to introduce the epidemiology of variants that we described via the Mutation epidemiological Renormalisation Group framework.