Large Datasets Research Articles

Four-year Outcomes of Class III and IV Anterior Restorations Based on a Subset of German Health Insurance Data.

Numerous studies report on the outcome performance of posterior composite restorations. However, there are fewer studies providing data for anterior restorations. The aim of this study was to evaluate the clinical outcome performance of anterior permanent restorations by analyzing a large dataset from a German national health insurance company. Routine claims data from a major German national health insurance company were assessed. Fee codes were used for tracing restoration careers on a day-count basis. The treatment was defined as a placed restoration (Class III and IV) on a mesial or distal tooth surface, irrespective of the extension. The restorations were placed between January 1, 2010 and December 31, 2013. Statistical analyses were conducted using Kaplan-Meier survival analysis to determine cumulative 4-year survival rates. The primary outcome was re-intervention on the same surface. Secondary outcomes were crowning and extraction which were analyzed separately. A total of 2,417,791 restorations involving mesial surfaces and a number of 2,409,031 restorations involving distal surfaces were observed. At 4 years, the cumulative survival rates concerning the primary outcome 're-intervention' were 79.9% for mesial and 80.9% for distal restorations. The respective annual failure rates (AFR) were 5.5% and 5.2%. Four-year survival rates for the secondary outcome 'crown' were 93.8% for mesial and 94.1% for distal anterior restorations. The respective AFRs were 1.6% and 1.5%. For the secondary outcome 'extraction,' the respective rates were 94.6% for mesial and 93.9% for distal restorations. The respective AFRs were 1.4% and 1.6%. The performance of permanent anterior restorations which were placed in general dental practices in Germany can be rated as acceptable.

Promoting Collaborative Scholarship During the COVID-19 Pandemic Through an Innovative COVID-19 Data Explorer and Repository at Yale School of Medicine: Development and Usability Study.

The COVID-19 pandemic sparked a surge of research publications spanning epidemiology, basic science, and clinical science. Thanks to the digital revolution, large data sets are now accessible, which also enables real-time epidemic tracking. However, despite this, academic faculty and their trainees have been struggling to access comprehensive clinical data. To tackle this issue, we have devised a clinical data repository that streamlines research processes and promotes interdisciplinary collaboration. This study aimed to present an easily accessible up-to-date database that promotes access to local COVID-19 clinical data, thereby increasing efficiency, streamlining, and democratizing the research enterprise. By providing a robust database, a broad range of researchers (faculty and trainees) and clinicians from different areas of medicine are encouraged to explore and collaborate on novel clinically relevant research questions. A research platform, called the Yale Department of Medicine COVID-19 Explorer and Repository (DOM-CovX), was constructed to house cleaned, highly granular, deidentified, and continually updated data from over 18,000 patients hospitalized with COVID-19 from January 2020 to January 2023, across the Yale New Haven Health System. Data across several key domains were extracted including demographics, past medical history, laboratory values during hospitalization, vital signs, medications, imaging, procedures, and outcomes. Given the time-varying nature of several data domains, summary statistics were constructed to limit the computational size of the database and provide a reasonable data file that the broader research community could use for basic statistical analyses. The initiative also included a front-end user interface, the DOM-CovX Explorer, for simple data visualization of aggregate data. The detailed clinical data sets were made available for researchers after a review board process. As of January 2023, the DOM-CovX Explorer has received 38 requests from different groups of scientists at Yale and the repository has expanded research capability to a diverse group of stakeholders including clinical and research-based faculty and trainees within 15 different surgical and nonsurgical specialties. A dedicated DOM-CovX team guides access and use of the database, which has enhanced interdepartmental collaborations, resulting in the publication of 16 peer-reviewed papers, 2 projects available in preprint servers, and 8 presentations in scientific conferences. Currently, the DOM-CovX Explorer continues to expand and improve its interface. The repository includes up to 3997 variables across 7 different clinical domains, with continued growth in response to researchers' requests and data availability. The DOM-CovX Data Explorer and Repository is a user-friendly tool for analyzing data and accessing a consistently updated, standardized, and large-scale database. Its innovative approach fosters collaboration, diversity of scholarly pursuits, and expands medical education. In addition, it can be applied to other diseases beyond COVID-19.

An analysis of Data mining techniques, multidimensional modelling on agriculture

This articles analysis the “An analysis of Data mining techniques, multidimensional modelling on agriculture.” The sole method for using and obtaining knowledge from large data sets is data mining. Samples are taken in increments from various locations within a bag. A consignment can contain up to 30 bags, of which one must be selected at random from each bag. For consignments containing 50 bags, however, samples of the Magadh division blocks in Bihar must be obtained from each bag. additionally to investigate how changing weather affects agriculture and how data mining in government platforms is formed. Whether there is a positive correlation between weather changes and agricultural and big data mining. Results shows that the Big Data mining has positive co-relation (r=.8) on weather changing on agriculture. The results confirms the hypothesis.

Unlocking the potential of AI: Machine learning and deep learning models for predicting carcinogenicity of chemicals.

The escalating apprehension surrounding the carcinogenic potential of chemicals emphasizes the imperative need for efficient methods of assessing carcinogenicity. Conventional experimental approaches such as in vitro and in vivo assays, albeit effective, suffer from being costly and time-consuming. In response to this challenge, new alternative methodologies, notably machine learning and deep learning techniques, have attracted attention for their potential in developing carcinogenicity prediction models. This article reviews the progress in predicting carcinogenicity using various machine learning and deep learning algorithms. A comparative analysis on these developed models reveals that support vector machine, random forest, and ensemble learning are commonly preferred for their robustness and effectiveness in predicting chemical carcinogenicity. Conversely, models based on deep learning algorithms, such as feedforward neural network, convolutional neural network, graph convolutional neural network, capsule neural network, and hybrid neural networks, exhibit promising capabilities but are limited by the size of available carcinogenicity datasets. This review provides a comprehensive analysis of current machine learning and deep learning models for carcinogenicity prediction, underscoring the importance of high-quality and large datasets. These observations are anticipated to catalyze future advancements in developing effective and generalizable machine learning and deep learning models for predicting chemical carcinogenicity.

Unveiling nuances in data analysis to illuminate marine pilot strain

Maritime studies, encompassing a range of disciplines, increasingly rely on advanced data analytics, particularly in the context of navigation. As technology advances, the statistical averaging of large datasets has become a critical component of these analyses. However, recent studies have highlighted discrepancies between statistical predictions and observable realities, especially in high-stress environments like port approach procedures conducted by marine pilots. This study analyzed physiological responses recorded during simulation exercises involving experienced marine pilots. The focus was not on the specific outcomes of the simulations but on the potential faults arising from conventional statistical signal processing, particularly mean-centered approaches. A large dataset of signals was generated, including one signal with a dominant characteristic intentionally designed to introduce imbalance, mimicking the uneven distribution of real-world data. Initial analysis suggested that the average physiological response of the pilots followed an S-shaped curve, indicative of a psycho-physiological reaction to stress. However, further post hoc analysis revealed that this pattern was primarily influenced by a single participant’s data. This finding raises concerns about the generalizability of the S-curve as a typical stress response in maritime pilots. The results underscore the limitations of relying solely on conventional statistical methods, such as mean-centered approaches, in interpreting complex datasets. The study calls into question the validity of standardizing data interpretations based on dominant characteristic curves, particularly in environments as intricate as maritime navigation. The research highlights the need for a re-evaluation of these methods to ensure more reliable and nuanced conclusions in maritime studies. This study contributes to the ongoing discourse on data interpretation in maritime research, emphasizing the critical need to re-assess conventional statistical signal processing techniques. By recognizing the potential pitfalls in data generalization, the study advocates for more robust analytical approaches to better capture the complexities of real-world maritime challenges.

M5C-Seq: Machine learning-enhanced profiling of RNA 5-methylcytosine modifications

Epigenetic modifications, particularly RNA methylation and histone alterations, play a crucial role in heredity, development, and disease. Among these, RNA 5-methylcytosine (m5C) is the most prevalent RNA modification in mammalian cells, essential for processes such as ribosome synthesis, translational fidelity, mRNA nuclear export, turnover, and translation. The increasing volume of nucleotide sequences has led to the development of machine learning-based predictors for m5C site prediction. However, these predictors often face challenges related to training data limitations and overfitting due to insufficient external validation. This study introduces m5C-Seq, an ensemble learning approach for RNA modification profiling, designed to address these issues. m5C-Seq employs a meta-classifier that integrates 15 probabilities generated from a novel, large dataset using systematic encoding methods to make final predictions. Demonstrating superior performance compared to existing predictors, m5C-Seq represents a significant advancement in accurate RNA modification profiling. The code and the newly established datasets are made available through GitHub at https://github.com/Z-Abbas/m5C-Seq.

A New Approach to Detect Driver Distraction to Ensure Traffic Safety and Prevent Traffic Accidents: Image Processing and MCDM

One of the factors that threaten traffic safety and cause various traffic problems is distracted drivers. Various studies have been carried out to ensure traffic safety and, accordingly, to reduce traffic accidents. This study aims to determine driver-distraction classes and detect driver violations with deep learning algorithms and decision-making methods. Different driver characteristics are included in the study by using a dataset created from five different countries. Weight classification in the range of 0–1 is used to determine the most important classes using the AHP method, and the most important 9 out of 23 classes are determined. The YOLOv8 algorithm is used to detect driver behaviors and distraction action classes. The YOLOv8 algorithm is examined according to performance-measurement criteria. According to mAP 0.5:0.95, an accuracy rate of 91.17% is obtained. In large datasets, it is seen that a successful result is obtained by using the AHP method, which is used to reduce transaction complexity, and the YOLOv8 algorithm, which is used to detect driver distraction. By detecting driver distraction, it is possible to partially avoid traffic accidents and the negative situations they create. While detecting and preventing driver distraction makes a significant contribution to traffic safety, it also provides a significant improvement in traffic accidents and traffic congestion, increasing transportation efficiency and the sustainability of cities. It also serves sustainable development goals such as energy efficiency and reducing carbon emissions.

Comparing morphology and cranial osteology in two divergent clades of dice snakes from continental Europe (Squamata: Natricidae: Natrix tessellata)

Abstract The Western Palearctic harbours a diverse snake fauna, including numerous endemic species and yet unnamed clades, identified through molecular analyses. However, morphological characteristics of these clades, even of common species, often remain relatively unexplored. In this study, we provide an examination of the morphology and cranial anatomy of the semi-aquatic snake species Natrix tessellata (Laurenti, 1768), with a focus on populations of the so-called ‘Europe’ and ‘Greece’ clades. Utilising both museum collections and field data, we first morphologically examined 541 individuals of N. tessellata, categorising them according to previously established clades and lineages that resulted in relatively low morphometric and meristic variation across the species’ range. When assessing the 448 specimens from the ‘Europe’ and the ‘Greece’ clades separately, we similarly observed little variation in meristic characteristics. On the other hand, individuals of the ‘Greece’ clade displayed smaller and more slender body and head proportions compared to those of the ‘Europe’ clade and the pigmentation of the labial scales is distinctively paler in the ‘Greece’ clade, whereas the overall body colouration remains largely similar between the two. Our osteological analysis of 47 N. tessellata skulls also indicated slight differences in the frontoparietal portion of the braincase between the ‘Europe’ and the ‘Greece’ clades, warranting further examination with a larger dataset and extending to other skull components. These findings hold significance for ongoing enquiries into the species’ biogeography, morphology and ecological adaptations. In summary, the integration of morphological and osteological data with genetic information offers a promising avenue for potential taxonomic revisions of N. tessellata in the future.

Parallel fractional dominance MOEAs for feature subset selection in big data

In this paper, we solve the feature subset selection (FSS) problem with three objective functions namely, cardinality, area under receiver operating characteristic curve (AUC) and Matthews correlation coefficient (MCC) using novel multi-objective evolutionary algorithms (MOEAs). MOEAs often encounter poor convergence due to the increase in non-dominated solutions and getting entrapped in the local optima. This situation worsens when dealing with large, voluminous big and high-dimensional datasets. To address these challenges, we propose parallel, fractional dominance-based MOEAs for FSS under Spark. Further, to improve the exploitation of MOEAs, we introduce a novel batch opposition-based learning (BOP) along with a cardinality constraint on the opposite solution. Accordingly, we propose two variants, namely, BOP1 and BOP2. In BOP1, a single neighbour is randomly chosen in the opposite solution space, whereas in BOP2, a group of randomly chosen neighbours in the opposite solution space. In either case, the opposite solutions are evaluated to improve the exploitation capability of the underlying MOEAs. We observe that in terms of mean optimal objective function values and across all datasets, the proposed BOP2 variant of parallel fractional dominance-based algorithms emerges as the top performer in obtaining efficient solutions. Further, we introduce a novel metric, namely the ratio of hypervolume (HV) and inverted generated distance (IGD), HV/IGD, that combines both diversity and convergence. With respect to the mean HV/IGD computed over 20 runs and Formula 1 racing, the BOP1 variants of fractional dominance-based MOEAs outperformed other algorithms.

Identify the digitalization technology opportunity of low-carbon energy technologies: Using the patent data and collaborative filtering.

The digitalization of low-carbon energy technologies (LCET) provides important technical support for the transition to a greener energy system. Digitalization addresses the phenomenon of the growing application of information and communications technologies (ICT) across the economy, which is regarded as the technology convergence between ICT and other technologies. Scholars have revealed the signs that LCET and ICT are becoming increasingly interlinked, which raises the challenges for predicting and identifying the technology opportunities for innovations in the converged technology area. To address the challenges, this paper proposes a collaborative filtering approach to identify the digitalization technology opportunity of low-carbon energy technologies using patent classification and patent citation information. We applied the proposed collaborative filtering approach using a large LCET patent dataset derived from the United States Patent and Trademark Office (USPTO). The results indicate that the proposed method can effectively identify digitalization technology opportunities of LCET, and the current LCET digitalization technology opportunities identified based on this approach are mainly concentrated in the Energy storage field. The advantages of the proposed approach are that its underlying data are more readily available and its technical complexity is relatively lower, and thus, more replicable for other technology fields.

Global estimates of 100-year return values of daily precipitation from ensemble weather prediction data

Abstract. High-impact river floods are often caused by very extreme precipitation events with return periods of several decades or centuries, and the design of flood protection measures thus relies on reliable estimates of the corresponding return values. However, calculating such return values from observations is associated with large statistical uncertainties due to the limited length of observational time series, uneven spatial distributions of rain gauges and trends associated with anthropogenic climate change. Here, 100-year return values of daily precipitation are estimated on a global grid based on a large data set of model-generated precipitation events from ensemble weather prediction. In this way, the statistical uncertainties in the return values can be substantially reduced compared to observational estimates due to the substantially longer time series. In spite of a general agreement in spatial patterns, the model-generated data set leads to systematically higher return values than the observations in many regions, with statistically significant differences, for instance, over the Amazon, western Africa, the Arabian Peninsula and India. This might be linked to an overestimation of tropical extreme precipitation in the model or an underestimation of extreme precipitation events in observations, which, if true, would have important consequences for practical water management.

Accelerating joint species distribution modelling with Hmsc-HPC by GPU porting.

Joint species distribution modelling (JSDM) is a widely used statistical method that analyzes combined patterns of all species in a community, linking empirical data to ecological theory and enhancing community-wide prediction tasks. However, fitting JSDMs to large datasets is often computationally demanding and time-consuming. Recent studies have introduced new statistical and machine learning techniques to provide more scalable fitting algorithms, but extending these to complex JSDM structures that account for spatial dependencies or multi-level sampling designs remains challenging. In this study, we aim to enhance JSDM scalability by leveraging high-performance computing (HPC) resources for an existing fitting method. Our work focuses on the Hmsc R-package, a widely used JSDM framework that supports the integration of various dataset types into a single comprehensive model. We developed a GPU-compatible implementation of its model-fitting algorithm using Python and the TensorFlow library. Despite these changes, our enhanced framework retains the original user interface of the Hmsc R-package. We evaluated the performance of the proposed implementation across various model configurations and dataset sizes. Our results show a significant increase in model fitting speed for most models compared to the baseline Hmsc R-package. For the largest datasets, we achieved speed-ups of over 1000 times, demonstrating the substantial potential of GPU porting for previously CPU-bound JSDM software. This advancement opens promising opportunities for better utilizing the rapidly accumulating new biodiversity data resources for inference and prediction.

Event-driven nearshore and shoreline coastline detection on SpiNNaker neuromorphic hardware

Abstract Coastline detection is vital for coastal management, involving frequent observation and assessment to understand coastal dynamics and inform decisions on environmental protection. 
Continuous streaming of high-resolution images demands robust data processing and storage solutions to manage large datasets efficiently, posing challenges that require innovative solutions for real-time analysis and meaningful insights extraction.
This work leverages low-latency event-based vision sensors coupled with neuromorphic hardware in an attempt to decrease a two-fold challenge, reducing the computational burden to ~0.375 mW whilst obtaining a coastline detection map in as little as 20 ms.
The proposed Spiking Neural Network (SNN) runs on the SpiNNaker neuromorphic platform using a total of 18040 neurons reaching 98.33% accuracy.
The model has been characterised and evaluated by computing the accuracy of Intersection over Union (IoU) scores over the ground truth of a real-world coastline dataset across different time windows. The system's robustness was further assessed by evaluating its ability to avoid coastline detection in non-coastline profiles and funny shapes, achieving a success rate of 97.3%.

Dung beetle optimization with deep learning approach for solving inverse problems in predicting financial futures

Leveraging deep learning (DL) to inverse problems has proven transformative in predicting financial futures, mainly in stock price prediction. In terms of financial markets, where predicting stock price is a challenging inverse problem, DL methods like long short-term memory networks (LSTM) and recurrent neural networks (RNN) demonstrate notable prowess. These techniques successfully capture complex patterns and dependencies in historical stock price information, learning to discern subtle correlations and fundamental market trends. DL facilitates the extraction of nuanced features by leveraging large datasets and sophisticated techniques, enabling accurate prediction of future stock prices. This intersection of advanced ML approaches and financial predicting signifies a paradigm shift in leveraging technology to navigate the complexity of the financial market and improves decision-making for traders and investors. This manuscript introduces a Dung Beetle Optimization with a Deep Learning Approach for Solving Inverse Problems in Predicting Financial Futures (DBODL-SIPPFF) technique. The DBODL-SIPPFF technique resolves the inverse problem and predicts the stock prices adequately. In addition, the DBODL-SIPPFF technique exploits data consistency (DC) optimizer to solve inverse problems. Primarily, linear scaling normalization (LSN) is applied for data normalization. The honey badger algorithm (HBA) is utilized for the feature selection. Moreover, convolutional long short-term memory (ConvLSTM) is used for stock price prediction. Lastly, the DBO technique is utilized for the optimum hyperparameter selection of the ConvLSTM model. The empirical assessment of the DBODL-SIPPFF technique takes place under diverse stock datasets. The obtained values stated that the DBODL-SIPPFF technique performs optimally over compared methods.

Prevalence and Risk Factors for Malignant Nodal Involvement in Early esophago-gastric Adenocarcinoma: Results from the Multicenter Retrospective Congress Study (endosCopic resectiON, esophaGectomy or Gastrectomy For Early Esophagogastric Cancers).

The aim of this study was to quantify LNM risk and outcomes following treatment of early esophago-gastric (EG) adenocarcinoma. The standard of care for early T1N0 EG cancer is endoscopic resection (ER). Radical surgical resection is recommended for patients perceived to be at risk of lymph node metastasis (LNM). Current models to select organ-preserving vs. surgical treatment are inconsistent. CONGRESS is a UK-based multicentre retrospective cohort study. Patients diagnosed with clinical or pathological T1N0 EG adenocarcinoma from 2015-2022 were included. Outcomes and rates of LNM were assessed. Cox regression was performed to assess the impact of prognostic and treatment factors on overall survival. 1,601 patients from 26 centres were included, with median follow-up 32 months(IQR 14-53). 1285/1612(80.3%) underwent ER, 497/1601(31.0%) underwent surgery. Overall rate of LNM was 13.5%. On ER staging, tumour depth (T1bsm2-3 17.6% vs. T1a 7.1%), lymphovascular invasion (17.2% vs. 12.6%), or signet cells (28.6% vs. 13.0%) were associated with LNM. In multivariable regression analysis, these were not significantly associated with LNM rates or survival. Adjusting for demographic and tumour variables, surgery after ER was associated with significant survival benefit, HR 0.33(0.15-0.77),P=0.010. This large multicentre dataset suggests that early EG adenocarcinoma is associated with significant risk of LNM. This data is representative of current real clinical practice with ER-based staging, and suggests previously held beliefs regarding reliability of predictive factors for LNM may need to be reconsidered. Further research to identify patients who may benefit from organ-preserving vs. surgical treatment is urgently required.

Statistical Metamodel of Liner Acoustic Impedance Based on Neural Network and Probabilistic Learning for Small Datasets

The main novelty of this paper consists of presenting a statistical artificial neural network (ANN)-based model for a robust prediction of the frequency-dependent aeroacoustic liner impedance using an aeroacoustic computational model (ACM) dataset of small size. The model, focusing on percentage of open area (POA) and sound pressure level (SPL) at a zero Mach number, takes into account uncertainties using a probabilistic formulation. The main difficulty in training an ANN-based model is the small size of the ACM dataset. The probabilistic learning carried out using the probabilistic learning on manifolds (PLoM) algorithm addresses this difficulty as it allows constructing a very large training dataset from learning the probabilistic model from a small dataset. A prior conditional probability model is presented for the PCA-based statistical reduced representation of the frequency-sampled vector of the log-resistance and reactance. It induces some statistical constraints that are not straightforwardly taken into account when training such an ANN-based model by classical optimizations methods under constraints. A second novelty of this paper consists of presenting an alternate solution that involves using conditional statistics estimated with learned realizations from PLoM. A numerical example is presented.

Analisis Sentimen Media Sosial Terhadap Aplikasi Perbankan Untuk Mengetahui Kepuasan Pengguna Aplikasi: Studi Kasus Pada Livin by Mandiri dan BCA Mobile

In this study, the researchers aimed to explore how sentiment expressed on social media platforms towards e-servqual provided by Livin' by Mandiri and BCA Mobile applications. This research uses sentiment analysis by collecting data from Twitter. Naïve Bayes is the method chosen for this research due to its user-friendly nature and ability to handle large datasets. Regarding the results of sentiment analysis on the e-servqual dimensions for the BCA Mobile application, overall opinions tend to be negative, although there are some positive opinions, especially on the variables of user friendliness, personal nedss, and efficiency. However, sentiments towards the dimensions of site organization, responsiveness, and reliability are mostly filled with negative sentiments. In contrast, sentiments towards the Livin by Mandiri application mostly contain positive sentiments. In the e-servqual dimension, Personal Needs, user frendlines and Responsiveness, and Site Organization all show positive sentiments, while, Efficiency, and Reliability show negative sentiments. In addition, based on the results of sentiment analysis, this research also examines topic modeling which is used to see what topics often appear in both applications.

Hydraulically linked reservoirs simultaneously fed the 1975–1984 Krafla Fires eruptions: Insights from petrochemistry

The 1975–1984 Krafla Fires in northeast Iceland was the first plate-boundary rifting episode to be tracked using seismic and geodetic monitoring. Geophysical observations from this episode have inspired conceptual models of magma transport during plate spreading, but a lack of complementary petrologic insights has hindered a holistic understanding of the events. To address this knowledge gap, we studied the petrochemistry of all nine Krafla Fires basaltic eruptions. Our large dataset of new whole-rock, matrix glass and mineral analyses from samples collected during or shortly after each eruption reveal a clear compositional bimodality in the erupted magmas that persisted across the episode, with evolved quartz tholeiite (MgO = 5.7–6.4 wt.%) erupted inside Krafla caldera, and more primitive (usually olivine-normative) tholeiite (MgO = 6.4–8.7 wt%) erupted north of the caldera margin. Barometric calculations indicate tapping of these magmas from distinct reservoirs: a primitive lower-crustal reservoir at a most probable depth of ∼14–19 km, and a more evolved, shallower reservoir at a most probable depth of ∼7–9 km beneath the caldera. These reservoirs were tapped simultaneously in several of the nine eruptions, and in three events the two magma types mixed near the northern caldera margin. Varying levels of trace element depletion in the deep-sourced primitive melts reflect incomplete mixing of diverse mantle-derived melts at depth; the most enriched of these melts could be parental to evolved inside-caldera magma via fractional crystallization. Clinopyroxene rims on gabbroic nodules from primitive September 1984 lavas record lower crustal pressures, while diffusion models suggest that these rims grew up to within a few months before eruption. Ascent of the primitive magma from the lower crust thus occurred over timescales much shorter than eruptive repose periods, without prolonged stalling at shallow depths. These observations are inconsistent with the view that the eruptions were entirely fed by lateral magma outflow from the shallow reservoir. They instead require some decoupling of the flow paths of the two magma types: the primitive magma either bypassed the sub-caldera reservoir laterally or ascended vertically beneath the northern vents. The two reservoirs nonetheless shared a hydraulic connection and jointly responded to rifting. Comparison of the Krafla Fires with other rifting events and eruptions highlights the complexity and diversity of magma transport during plate boundary rifting events, which is not yet captured by a generalizable model. Integration of petrologic, geochemical and geophysical data is essential to provide a holistic view of future rifting events in Iceland and at other spreading centres.

HTR-VT: Handwritten text recognition with vision transformer

We explore the application of Vision Transformer (ViT) for handwritten text recognition. The limited availability of labeled data in this domain poses challenges for achieving high performance solely relying on ViT. Previous transformer-based models required external data or extensive pre-training on large datasets to excel. To address this limitation, we introduce a data-efficient ViT method that uses only the encoder of the standard transformer. We find that incorporate a Convolutional Neural Network (CNN) for feature extraction instead of the original patch embedding and employ Sharpness-Aware Minimization (SAM) optimizer to ensure that the model can converge towards flatter minima yield notable enhancements. Furthermore, our introduction of the span mask technique, which masks interconnected features in the feature map, acts as an effective regularizer. Empirically, our approach competes favorably with traditional CNN-based models on small datasets like IAM and READ2016. Additionally, it establishes a new benchmark on the LAM dataset, currently the largest dataset with 19,830 training text lines. The code will be publicly available at: https://github.com/YutingLi0606/HTR-VT.

Optimizing the thermal performance of phase change materials in building applications using deep reinforcement learning and Bayesian optimization

This research presents a novel methodology for Deep Reinforcement Learning (DRL) and Bayesian Optimisation of the thermal performance of PCMs in building operations. The developed models utilise a unique and large dataset comprising 1500 building thermal profiles, simulated for various climates and building setups obtained from our industry partners and as open data. PCM-based systems are deployed for thermal insulation of building envelopes to regulate indoor temperature conditions and reduce the need for heating and cooling systems, resulting in enhanced energy efficiency. Through the real-time management of the thermal efficacy of PCMs using the DRL method trained on the large dataset and fine-tuning of the underlying model parameters using Bayesian Optimisation, the optimised system achieves energy saving in heating and cooling load of up to 45 %, along with the induced reduction in CO2 emission. At the same time, DRL contributes to decreasing the thermal fluctuation in the indoor temperature and keeps it in the narrow range of 1.2 °C in case of high thermal variability scenarios. Currently, the best performance is reported in the literature. This research exemplifies the potential of DRL and Bayesian optimisation in sustainable building. It depicts the applications of advanced intelligent computing algorithms with big building energy data as a novel, robust and superior approach for optimising real-world building energy management systems. The methodology and the improvements in energy savings in thermal and energy management of buildings highlight the novelty and potential benefit of the implemented research as a new intellectual property towards sustainable building design.