Original Dataset Research Articles

Results Obtained from a Pivotal Validation Trial of a Microsatellite Analysis (MSA) Assay for Bladder Cancer Detection through a Statistical Approach Using a Four-Stage Pipeline of Modern Machine Learning Techniques.

Several studies have shown that microsatellite changes can be profiled in urine for the detection of bladder cancer. The use of microsatellite analysis (MSA) for bladder cancer detection requires a comprehensive analysis of as many as 15 to 20 markers, based on the amplification and interpretations of many individual MSA markers, and it can be technically challenging. Here, to develop fast, more efficient, standardized, and less costly MSA for the detection of bladder cancer, we developed three multiplex-polymerase-chain-reaction-(PCR)-based MSA assays, all of which were analyzed via a genetic analyzer. First, we selected 16 MSA markers based on 9 selected publications. Based on samples from Johns Hopkins University (the JHU sample, the first set sample), we developed an MSA based on triplet, three-tube-based multiplex PCR (a Triplet MSA assay). The discovery, validation, and translation of biomarkers for the early detection of cancer are the primary focuses of the Early Detection Research Network (EDRN), an initiative of the National Cancer Institute (NCI). A prospective study sponsored by the EDRN was undertaken to determine the efficacy of a novel set of MSA markers for the early detection of bladder cancer. This work and data analysis were performed through a collaboration between academics and industry partners. In the current study, we undertook a re-analysis of the primary data from the Compass study to enhance the predictive power of the dataset in bladder cancer diagnosis. Using a four-stage pipeline of modern machine learning techniques, including outlier removal with a nonlinear model, correcting for majority/minority class imbalance, feature engineering, and the use of a model-derived variable importance measure to select predictors, we were able to increase the utility of the original dataset to predict the occurrence of bladder cancer. The results of this analysis showed an increase in accuracy (85%), sensitivity (82%), and specificity (83%) compared to the original analysis. The re-analysis of the EDRN study results using machine learning statistical analysis proved to achieve an appropriate level of accuracy, sensitivity, and specificity to support the use of the MSA for bladder cancer detection and monitoring. This assay can be a significant addition to the tools urologists use to both detect primary bladder cancers and monitor recurrent bladder cancer.

Improved bioimpedance spectroscopy tissue classification through data augmentation from generative adversarial networks.

Bioimpedance spectroscopy is a tissue classification technique with many clinical applications. Similarly to other data-driven methods, it requires large amounts of data to accurately distinguish similar classes of tissue. Classifiers trained on small datasets typically suffer from over-fitting and lack the ability to generalise to previously unseen data. However, a large in or ex vivo spectral database is difficult to attain. Data collection is usually limited to studies that occur infrequently, and publicly available data is often not available. A solution to this problem is to artificially increase the training dataset by creating modified, yet accurate, copies of the original dataset. The most common techniques in spectral classification are to add noise to copies of the original data, over-sample it, or randomly interpolate pairs of the original data. However, simply perturbing or interpolating the data does not guarantee that the new dataset captures the key features of the original data needed for accurate classification. This study proposes a novel way to augment bioimpedance spectral data. It uses generative adversarial networks (GAN)-a model in which two neural networks (NN) compete with each other: while one NN artificially manufactures data that could be mistaken for real data, the role of the second NN is to identify which data it receives has been artificially created. The first NN then interactively adapts its output until the second NN can no longer flag artificially created data. The result is a new dataset that truly represents the features of the original data. In this study, three GAN architectures are used, i.e., the vanilla GAN, the deep convolutional GAN, and the Wasserstein GAN. Then, the generated data is used to train five classification methods, and their results are compared to a baseline that only uses the original data. The results from a dataset of 13 different tissue classes show that the deep convolutional GAN is most statistically similar to the original data and improves classification accuracy by 15% when compared to the same model trained only on the original data. The Wasserstein-GAN architecture also provides significant improvements of up to 24% better accuracy.

Blinded by sunspots: Revealing the multidimensional and intersectional inequities of solar energy in India

Studies of energy transitions have historically lacked a holistic, multi-scalar and multi-site accounting of social and environmental impacts of projects. Scholars increasingly point to the need for integrated studies that highlight impacts at various stages of lifecycles and scales of governance in the normative pursuit of energy justice. In this study, we examine the social and environmental inequities within the solar PV value chain in India from multiple scales using an original dataset and comparative case study approach, including multiple site visits and embedded ethnography. We utilized a mixed methodological approach to data collection that included household surveys (n = 120), semi-structured interviews (n = 43), focus group discussions (n = 6) and naturalistic observation (n = 9 site visits over 24 days) across the states of Uttar Pradesh, Karnataka, Rajasthan, Delhi and Tamil Nadu at six different locations. We buttress recent work examining solar injustices from a ‘whole systems’ perspective, taking a multidimensional approach through an intersectional lens. Drawing from fieldwork and the literatures of energy justice and political ecology, we ask : What are the social and ecological impacts or equity concerns of each stage of the solar photovoltaic value chain in India? At each solar node, there are demographic, spatial, interspecies and temporal inequities occurring. Demographic inequities include resource dispossession, loss of livelihoods and hazards exposure for marginalized groups. Spatial inequities include distributional injustices for capital and electricity, rural land enclosures and substandard infrastructure. Interspecies inequities include ecosystem contamination, habitat destruction and imperiling wildlife. Temporal inequities include time burdens placed on marginalized groups, multi-decadal time horizons for inequitable projects to pay off and the long-term degradation of the environment. We conclude by reflecting on the stakes of solar development in India and offering policy recommendations that could engender a more equitable solar sector and chart future research agendas.

Assessment of a deep learning model for COVID-19 classification on chest radiographs: a comparison across image acquisition techniques and clinical factors.

The purpose is to assess the performance of a pre-trained deep learning model in the task of classifying between coronavirus disease (COVID)-positive and COVID-negative patients from chest radiographs (CXRs) while considering various image acquisition parameters, clinical factors, and patient demographics. Standard and soft-tissue CXRs of 9860 patients comprised the "original dataset," consisting of training and test sets and were used to train a DenseNet-121 architecture model to classify COVID-19 using three classification algorithms: standard, soft tissue, and a combination of both types of images via feature fusion. A larger more-current test set of 5893 patients (the "current test set") was used to assess the performance of the pretrained model. The current test set contained a larger span of dates, incorporated different variants of the virus and included different immunization statuses. Model performance between the original and current test sets was evaluated using area under the receiver operating characteristic curve (ROC AUC) [95% CI]. The model achieved AUC values of 0.67 [0.65, 0.70] for cropped standard images, 0.65 [0.63, 0.67] for cropped soft-tissue images, and 0.67 [0.65, 0.69] for both types of cropped images. These were all significantly lower than the performance of the model on the original test set. Investigations regarding matching the acquisition dates between the test sets (i.e., controlling for virus variants), immunization status, disease severity, and age and sex distributions did not fully explain the discrepancy in performance. Several relevant factors were considered to determine whether differences existed in the test sets, including time period of image acquisition, vaccination status, and disease severity. The lower performance on the current test set may have occurred due to model overfitting and a lack of generalizability.

Decarbonisation through modernisation: The UK’s EV transition as a vehicle for industrial change

As the transport sector remains the only sector in the United Kingdom (UK) in which emissions continue to rise, electric vehicles (EVs) have become an increasingly prominent political subject. This article finds that far from simply a low carbon transition, the EV transition in the UK is a means to address the attendant frailties of its economic model. Drawing upon an original empirical dataset of 33 interviews alongside a document analysis, I show that decarbonising the UK automobile sector is tied to a process of industrial and economic modernisation. These findings reveal the contested features of the transition, as the desire to capitalise on broader shifts in the global automobile market is brought into opposition with Britain’s financialised political economy. Finally, it shows that the modernisation tenet of Ecological Modernisation, which has come to be routinely neglected in the literature, should be revisited in the context of Brexit, COVID-19 and economic crises.

Bridging the Camera Domain Gap With Image-to-Image Translation Improves Glaucoma Diagnosis.

The purpose of this study was to improve the automated diagnosis of glaucomatous optic neuropathy (GON), we propose a generative adversarial network (GAN) model that translates Optain images to Topcon images. We trained the GAN model on 725 paired images from Topcon and Optain cameras and externally validated it using an additional 843 paired images collected from the Aravind Eye Hospital in India. An optic disc segmentation model was used to assess the disparities in disc parameters across cameras. The performance of the translated images was evaluated using root mean square error (RMSE), peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), 95% limits of agreement (LOA), Pearson's correlations, and Cohen's Kappa coefficient. The evaluation compared the performance of the GON model on Topcon photographs as a reference to that of Optain photographs and GAN-translated photographs. The GAN model significantly reduced Optain false positive results for GON diagnosis, with RMSE, PSNR, and SSIM of GAN images being 0.067, 14.31, and 0.64, respectively, the mean difference of VCDR and cup-to-disc area ratio between Topcon and GAN images being 0.03, 95% LOA ranging from -0.09 to 0.15 and -0.05 to 0.10. Pearson correlation coefficients increased from 0.61 to 0.85 in VCDR and 0.70 to 0.89 in cup-to-disc area ratio, whereas Cohen's Kappa improved from 0.32 to 0.60 after GAN translation. Image-to-image translation across cameras can be achieved by using GAN to solve the problem of disc overexposure in Optain cameras. Our approach enhances the generalizability of deep learning diagnostic models, ensuring their performance on cameras that are outside of the original training data set.

The Global Maritime Fulcrum as A Strategic Narrative: A Critical Examination of Its Emergence, Conveyance, and Decline in Indonesian Strategic Discourse

In 2014, President Joko ‘Jokowi’ Widodo declared he would lead Indonesia, the most populous Muslim nation on earth, to become a ‘global maritime fulcrum’ (GMF). His announcement was taken as a sign of Indonesia’s commitment to be a more active participant in regional maritime security. This nascent maritime vision, however, suffers from endogenous flaws, ranging from personal leadership to widespread bureaucratic deficiencies, culminating in its recession from Indonesian political discourse. This paper seeks to understand how and why the GMF receded from political discourse through the lens of strategic narrative. It finds that technocratic appeal serves as a stronger legitimating method of the GMF compared to other forms, which is due to agential preferences aligning with domestic structural conditions, resulting in development narratives being favoured over maritime-related narratives. The relationship results in government priorities being skewed towards the infrastructure aspect of the GMF at the cost of the other aspects. To demonstrate this argument, this paper analyses an original dataset of 267 speeches and statements, both in English and Indonesian, delivered by key government officials related to formulating and implementing the GMF concept from 2014 to 2021, in addition to relevant policy documents.

HubHSP graph: Capturing local geometrical and statistical data properties via spanning graphs

The computation of a continuous generative model to describe a finite sample of an infinite metric space can prove challenging and lead to erroneous hypothesis, particularly in high-dimensional spaces. In this paper, we follow a different route and define the Hubness Half Space Partitioning graph (HubHSP graph). By constructing this spanning graph over the dataset, we can capture both the geometrical and statistical properties of the data without resorting to any continuity assumption. Leveraging the classical graph-theoretic apparatus, the HubHSP graph facilitates critical operations, including the creation of a representative sample of the original dataset, without relying on density estimation. This representative subsample is essential for a range of operations, including indexing, visualization, and machine learning tasks such as clustering or inductive learning. With the HubHSP graph, we can bypass the limitations of traditional methods and obtain a holistic understanding of our dataset’s properties, enabling us to unlock its full potential.

Ship-Go: SAR Ship Images Inpainting via instance-to-image Generative Diffusion Models

We present Ship-Go, an instance-to-image diffusion model, to increase the scale and diversity of the SAR detection datasets. Ship-Go is developed as a multi-conditions denoising diffusion probabilistic model (DDPM), i.e., it takes the proposed visual instances and environment prompt condition variables as constraints to generate backscatter intensity information for each resolution cell, resulting in inpainting a SAR image and generating the corresponding instance-level annotations, which can be directly employed to train detection models. We demonstrate, for the first time, the ability to place the ship objects at any angle, size, and arrangement in the generated background of multiple specified environment types. Importantly, two image generation scenarios are designed to increase the diversity of objects and backgrounds for the original dataset (i.e., in-domain augmentation and out-of-domain augmentation). The experiment verifies that the generated detection datasets boost the performance of multiple classical deep detectors in the different cases of insufficient samples. Qualitatively, we find that Ship-Go improves the diversity of the existing dataset, and has sample distribution transferability among multiple datasets. Code and models are available at: https://github.com/XinZhangRadar/Ship-Go.

Hypothermia vs Normothermia in Patients With Cardiac Arrest and Nonshockable Rhythm

International guidelines recommend body temperature control below 37.8 °C in unconscious patients with out-of-hospital cardiac arrest (OHCA); however, a target temperature of 33 °C might lead to better outcomes when the initial rhythm is nonshockable. To assess whether hypothermia at 33 °C increases survival and improves function when compared with controlled normothermia in unconscious adults resuscitated from OHCA with initial nonshockable rhythm. Individual patient data meta-analysis of 2 multicenter, randomized clinical trials (Targeted Normothermia after Out-of-Hospital Cardiac Arrest [TTM2; NCT02908308] and HYPERION [NCT01994772]) with blinded outcome assessors. Unconscious patients with OHCA and an initial nonshockable rhythm were eligible for the final analysis. The study cohorts had similar inclusion and exclusion criteria. Patients were randomized to hypothermia (target temperature 33 °C) or normothermia (target temperature 36.5 to 37.7 °C), according to different study protocols, for at least 24 hours. Additional analyses of mortality and unfavorable functional outcome were performed according to age, sex, initial rhythm, presence or absence of shock on admission, time to return of spontaneous circulation, lactate levels on admission, and the cardiac arrest hospital prognosis score. Only patients who experienced OHCA and had a nonshockable rhythm with all causes of cardiac arrest were included. Variables from the 2 studies were available from the original data sets and pooled into a unique database and analyzed. Clinical outcomes were harmonized into a single file, which was checked for accuracy of numbers, distributions, and categories. The last day of follow-up from arrest was recorded for each patient. Adjustment for primary outcome and functional outcome was performed using age, gender, time to return of spontaneous circulation, and bystander cardiopulmonary resuscitation. The primary outcome was mortality at 3 months; secondary outcomes included unfavorable functional outcome at 3 to 6 months, defined as a Cerebral Performance Category score of 3 to 5. A total of 912 patients were included, 490 from the TTM2 trial and 422 from the HYPERION trial. Of those, 442 had been assigned to hypothermia (48.4%; mean age, 65.5 years; 287 males [64.9%]) and 470 to normothermia (51.6%; mean age, 65.6 years; 327 males [69.6%]); 571 patients had a first monitored rhythm of asystole (62.6%) and 503 a presumed noncardiac cause of arrest (55.2%). At 3 months, 354 of 442 patients in the hypothermia group (80.1%) and 386 of 470 patients in the normothermia group (82.1%) had died (relative risk [RR] with hypothermia, 1.04; 95% CI, 0.89-1.20; P = .63). On the last day of follow-up, 386 of 429 in the hypothermia group (90.0%) and 413 of 463 in the normothermia group (89.2%) had an unfavorable functional outcome (RR with hypothermia, 0.99; 95% CI, 0.87-1.15; P = .97). The association of hypothermia with death and functional outcome was consistent across the prespecified subgroups. In this individual patient data meta-analysis, including unconscious survivors from OHCA with an initial nonshockable rhythm, hypothermia at 33 °C did not significantly improve survival or functional outcome.

Drivers of the artiodactyl turnover in insular western Europe at the Eocene-Oligocene Transition.

Simultaneously investigating the effects of abiotic and biotic factors on diversity dynamics is essential to understand the evolutionary history of clades. The Grande Coupure corresponds to a major faunal turnover at the Eocene-Oligocene transition (EOT) (~34.1 to 33.55 Mya) and is defined in western Europe as an extinction of insular European mammals coupled with the arrival of crown clades from Asia. Here, we focused on the species-rich group of endemic European artiodactyls to determine the drivers of the Grande Coupure during the major environmental disruptions at the EOT. Using Bayesian birth-death models, we analyzed an original high-resolution fossil dataset (90 species, >2,100 occurrences) from southwestern France (Quercy area) and estimated the regional diversification and diversity dynamics of endemic and immigrant artiodactyls. We show that the endemic artiodactyl radiation was mainly related to the Eocene tropical conditions, combined with biotic controls on speciation and clade-related diversity dependence. We further highlight that the major environmental changes at the transition (77% of species became extinct) and the concurrent increase in seasonality in Europe during the Oligocene were likely the main drivers of their decline. Surprisingly, our results do not support the widely-held hypothesis of active competition between endemic and immigrant artiodactyls but rather suggest a passive or opportunistic replacement by immigrants, which is further supported by morphological clustering of specific ecological traits across the Eocene-Oligocene transition. Our analyses provide insights into the evolutionary and ecological processes driving the diversification and decline of mammalian clades during a major biological and climatic crisis.

ECerto—versatile software for interlaboratory data evaluation and documentation during reference material production

The statistical tool eCerto was developed for the evaluation of measurement data to assign property values and associated uncertainties of reference materials. The analysis is based on collaborative studies of expert laboratories and was implemented using the R software environment. Emphasis was put on comparability of eCerto with SoftCRM, a statistical tool based on the certification strategy of the former Community Bureau of Reference. Additionally, special attention was directed towards easy usability from data collection through processing, archiving, and reporting. While the effects of outlier removal can be flexibly explored, eCerto always retains the original data set and any manipulation such as outlier removal is (graphically and tabularly) documented adequately in the report. As a major reference materials producer, the Bundesanstalt für Materialforschung und -prüfung (BAM) developed and will maintain a tool to meet the needs of modern data processing, documentation requirements, and emerging fields of RM activity. The main features of eCerto are discussed using previously certified reference materials.Graphical

Facial attribute editing method combined with parallel GAN for attribute separation

Facial attribute editing encounters problems of incorrect changes to face regions and artifacts in generated images. We propose a facial attribute editing method combined with parallel GAN for attribute separation. First, the method integrates the U2-net encoder and Trans-GAN decoder as a model encoder to extract and generate facial spatial information effectively. Second, RGB images and semantic mask images are used to train a parallel generator and discriminator respectively. Semantic consistency loss is introduced to ensure that the two branches have consistent semantic output and achieve the effect of convergence in the same direction parallel generator and discriminator. The proposed model, trained on the CelebAMask-HQ original dataset and verified by the CelebA dataset, adopts the separation of the face mask image and the background mask image, to improve the correct rate of face attribute editing. Compared with existing facial attribute editing methods, the proposed method is capable of balancing attribute editing ability and details preservation ability. It can accurately edit the target attribute area and greatly improve the quality of facial images.

Enhancer Recognition: A Transformer Encoder-Based Method with WGAN-GP for Data Augmentation.

Enhancers are located upstream or downstream of key deoxyribonucleic acid (DNA) sequences in genes and can adjust the transcription activity of neighboring genes. Identifying enhancers and determining their functions are important for understanding gene regulatory networks and expression regulatory mechanisms. However, traditional enhancer recognition relies on manual feature engineering, which is time-consuming and labor-intensive, making it difficult to perform large-scale recognition analysis. In addition, if the original dataset is too small, there is a risk of overfitting. In recent years, emerging methods, such as deep learning, have provided new insights for enhancing identification. However, these methods also present certain challenges. Deep learning models typically require a large amount of high-quality data, and data acquisition demands considerable time and resources. To address these challenges, in this paper, we propose a data-augmentation method based on generative adversarial networks to solve the problem of small datasets. Moreover, we used regularization methods such as weight decay to improve the generalizability of the model and alleviate overfitting. The Transformer encoder was used as the main component to capture the complex relationships and dependencies in enhancer sequences. The encoding layer was designed based on the principle of k-mers to preserve more information from the original DNA sequence. Compared with existing methods, the proposed approach made significant progress in enhancing the accuracy and strength of enhancer identification and prediction, demonstrating the effectiveness of the proposed method. This paper provides valuable insights for enhancer analysis and is of great significance for understanding gene regulatory mechanisms and studying disease correlations.

Digestibility of dinosaur food plants revisited and expanded: Previous data, new taxa, microbe donors, foliage maturity, and seasonality

Although the living relatives of the Mesozoic flora were once assumed to constitute a nutritionally poor diet for dinosaur herbivores, in vitro fermentation of their foliage has shown that gymnosperms, ferns, and fern relatives can be as highly digestible as angiosperm grasses and dicots. Because nutritional information cannot be preserved in the fossil record, this laboratory approach, first published in 2008, provides a novel alternative to evaluate the digestive quality of the plants that were available to dinosaur megaherbivores such as sauropods. However, very few further studies have since been conducted to supplement and confirm the high fermentative capacity of nonangiospermous taxa. Here we show that the living relatives of the Araucariaceae and Equistaceae are consistently highly digestible, even between taxa and when influenced by environmental and biological factors, while fern taxa are inconsistent on the family level. These results reinforce previous findings about the high energetic potential of Jurassic-age plant families. Fourteen species of fern and gymnosperm foliage from five Jurassic families were collected in the spring and fall, then analyzed for their digestibility using the in vitro Hohenheim gas test. Equisetum, Araucaria, and Angiopteris were the most digestible genera in both seasons, while Agathis, Wollemia, and Marattia were the least digestible. The season in which specimens were collected was found to have to a significant effect on gas production in four out of 16 samples (P < 0.05). Furthermore, leaf maturity influences digestibility in Marattia attenuata (P < 0.05), yet not in Cyathea cooperi (P = 0.24). Finally, the species of the rumen fluid donor did not influence digestibility (P = 0.74). With the original data set supplemented by one new genus and four species, this study confirms and expands previous results about the nutritional capacity of the living relatives of the Jurassic flora.

EPAG: A novel enhanced move recognition algorithm based on continuous learning mechanism with positional embedding

The identification of abstracts plays a vital role in efficiently locating the content and providing clarity to the article. Existing algorithms for move recognition exhibit a deficiency in their capacity to acquire word adjacent position information when word changes in Chinese expressions to obtain contextual semantics changes. This paper introduces EPAG: a novel enhanced move recognition algorithm with the improved pre-trained framework and downstream model for unstructured abstracts of Chinese scientific and technological papers. The proposed algorithm first performs data segmentation and vocabulary training. The EPAG framework is leveraged to incorporate word positional information, facilitating deep semantic learning and targeted feature extraction. Experimental results demonstrate that the proposed algorithm achieves 13.37% higher accuracy on the split dataset than on the original dataset and a 7.55% improvement in accuracy over the basic comparison model.

Developing a model to predict neonatal respiratory distress syndrome and affecting factors using data mining: A cross-sectional study.

Background: One of the major challenges that hospitals and clinicians face is the early identification of newborns at risk for adverse events. One of them is neonatal respiratory distress syndrome (RDS). RDS is the widest spared respiratory disorder in immature newborns and the main source of death among them. Machine learning has been broadly accepted and used in various scopes to analyze medical information and is very useful in the early detection of RDS. Objective: This study aimed to develop a model to predict neonatal RDS and affecting factors using data mining. Materials and Methods: The original dataset in this cross-sectional study was extracted from the medical records of newborns diagnosed with RDS from July 2017-July 2018 in Alzahra hospital, Tabriz, Iran. This data includes information about 1469 neonates, and their mothers information. The data were preprocessed and applied to expand the classification model using machine learning techniques such as support vector machine, Naïve Bayes, classification tree, random forest, CN2 rule induction, and neural network, for prediction of RDS episodes. The study compares models according to their accuracy. Results: Among the obtained results, an accuracy of 0.815, sensitivity of 0.802, specificity of 0.812, and area under the curve of 0.843 was the best output using random forest. Conclusion: The findings of our study proved that new approaches, such as data mining, may support medical decisions, improving diagnosis in neonatal RDS. The feasibility of using a random forest in neonatal RDS prediction would offer the possibility to decrease postpartum complications of neonatal care.

Identification of spikes in continuous ground-based in situ time series of CO2, CH4 and CO: an extended experiment within the European ICOS Atmosphere network

Abstract. The identification of spikes (i.e., short and high variability in the measured signals due to very local emissions occurring in the proximity of a measurement site) is of interest when using continuous measurements of atmospheric greenhouse gases (GHGs) in different applications like the determination of long-term trends and/or spatial gradients, inversion experiments devoted to the top-down quantification of GHG surface–atmosphere fluxes, the characterization of local emissions, or the quality control of GHG measurements. In this work, we analyzed the results provided by two automatic spike identification methods (i.e., the standard deviation of the background (SD) and the robust extraction of baseline signal (REBS)) for a 2-year dataset of 1 min in situ observations of CO2, CH4 and CO at 10 different atmospheric sites spanning different environmental conditions (remote, continental, urban). The sensitivity of the spike detection frequency and its impact on the averaged mole fractions on method parameters was investigated. Results for both methods were compared and evaluated against manual identification by the site principal investigators (PIs). The study showed that, for CO2 and CH4, REBS identified a larger number of spikes than SD and it was less “site-sensitive” than SD. This led to a larger impact of REBS on the time-averaged values of the observed mole fractions for CO2 and CH4. Further, it could be shown that it is challenging to identify one common algorithm/configuration for all the considered sites: method-dependent and setting-dependent differences in the spike detection were observed as a function of the sites, case studies and considered atmospheric species. Neither SD nor REBS appeared to provide a perfect identification of the spike events. The REBS tendency to over-detect the spike occurrence shows limitations when adopting REBS as an operational method to perform automatic spike detection. REBS should be used only for specific sites, mostly affected by frequent very nearby local emissions. SD appeared to be more selective in identifying spike events, and the temporal variabilities in CO2, CH4 and CO were more consistent with those of the original datasets. Further activities are needed for better consolidating the fitness for purpose of the two proposed methods and to compare them with other spike detection techniques.

PRINCIPAL COMPONENTS VERSUS AUTOENCODERS FOR DIMENSIONALITY REDUCTION: A CASE OF SUPER-RESOLVED OUTPUTS FROM PRISMA HYPERSPECTRAL MISSION DATA

Abstract. This study attempts to solve these issues associated with hyperspectral (HS) data, i.e., coarse spatial resolution and high volume, by understanding the effect of deep learning and traditional dimensionality reduction on super-resolved products generated from the recently launched PRecursore IperSpettrale della Missione Applicativa (PRISMA) HS mission. Four single-frame super-resolution (SR) algorithms have been used to super-resolve a 30 m PRISMA scene of Ahmedabad, India and generate 15 m spatial resolution images with both spatial and spectral fidelity. Iterative back projection (IBP) and sparse representation (SIS) are the best and worst-performing SR algorithms following a comparative assessment and validation protocol. Next, denoising autoencoders and PCT computed using singular and eigenvalue decompositions have been executed on the original PRISMA, IBP and SIS-based super-resolved datasets. The resulting low-dimensional representations have been assessed to preserve the original dataset's topology using label-independent Lee and Verleysen's co-ranking matrix and loss of quality measure. Findings suggest that autoencoders are computationally expensive and require a higher neighbourhood size than PCT and its variants to produce a high-quality encoding. These insights remain significant for urban information extraction as there are few direct comparative assessments between machine learning-based linear and non-linear data compression methods in earlier studies.

Fuzzy self-consistent clustering ensemble

The subject of clustering ensemble (CE) has emerged as a pivotal area of exploration within unsupervised learning, primarily due to its remarkable capacity to integrate multiple base clusterings. However, the conventional CE framework exhibits limitations in effectively handling fuzzy relations and ensuring robustness. This study introduces an innovative and robust fuzzy self-consistent clustering ensemble (FSCE) model. Departing from customary approaches to processing base clustering outcomes, the FSCE model considers the scalable dummy variable representation of base clustering results as a novel feature attributes intrinsic to the original dataset. Subsequently, a γ-fuzzy operator ε+ is formulated, enabling the adjustment of coupling strength contingent upon the uncertainties inherent in practical problems. Leveraging this operator, the model reappraises the fuzzy relationships among objects, thereby engendering a corresponding relation matrix. This matrix profoundly enhances the model's capacity to contend with uncertain relationships. The resultant relation matrix assumes a pivotal role as the primary input in the consensus function, culminating in the derivation of the ensemble outcome. Moreover, the FSCE model deviates from the conventional CE paradigm by introducing a reallocation strategy for fuzzy objects within ensemble outcome. Through comprehensive experimentation, we substantiate that the introduction of γ-fuzzy operator ε+ offers a viable novel approach to enhancing both the overall performance and the fuzzy interpretability of the CE model. The FSCE model distinctly excels in consensus formation and robustness when juxtaposed against eight archetypal and state-of-the-art clustering models.