67639 research outputs found
Sort by
Improved Random Features for Dot Product Kernels
No full textDot product kernels, such as polynomial and exponential (softmax) kernels, are among the most widely used kernels in machine learning, as they enable modeling the interactions between input features, which is crucial in applications like computer vision, natural language processing, and recommender systems. We make several novel contributions for improving the efficiency of random feature approximations for dot product kernels, to make these kernels more useful in large scale learning. First, we present a generalization of existing random feature approximations for polynomial kernels, such as Rademacher and Gaussian sketches and TensorSRHT, using complex-valued random features. We show empirically that the use of complex features can significantly reduce the variances of these approximations. Second, we provide a theoretical analysis for understanding the factors affecting the efficiency of various random feature approximations, by deriving closed-form expressions for their variances. These variance formulas elucidate conditions under which certain approximations (e.g., TensorSRHT) achieve lower variances than others (e.g., Rademacher sketches), and conditions under which the use of complex features leads to lower variances than real features. Third, by using these variance formulas, which can be evaluated in practice, we develop a data-driven optimization approach to improve random feature approximations for general dot product kernels, which is also applicable to the Gaussian
kernel. We describe the improvements brought by these contributions with extensive experiments on a variety of tasks and datasets
Permeability and Acoustic Velocity Evolution of Ultratight Jurassic Anhydrite Caprock (Saudi Arabia) in Response to CO2 Exposure Under Reservoir Conditions: Implications to CO2 Storage
No full textSaudi Arabia’s Jurassic reservoir-caprock sequences play an important role for CO2 sequestration as they have the potential to store and seal large volumes of CO2 in saline aquifers. Containment longevity relies on caprock integrity, which is composed mainly of anhydrite layers with a thickness of several 10s of meters. To investigate some of the complex chemo-thermohydro- mechanical coupled processes associated with injection and assess risk for caprock failure, we conducted an experimental study that tests several Hith anhydrite core samples retrieved from a shallow, near outcrop well drilled close to the capital Riyadh.
The goal is to observe how permeability evolves under in situ reservoir conditions as well as the presence of supercritical CO2.
Initial gas filled anhydrite porosities are around 2% or less. Five specimens were tested either dry or brined saturated by flowing supercritical CO2 at two constant mean effective stress (sm) values. These varied from ambient to presumed in situ reservoir conditions (i.e., sm from 1,000 psi to 4,000 psi and constant temperature of 50°C). Each test lasted between 3 and 6 days while monitoring upstream/downstream flowrates, fluid pressures, differential pressure (DP), acoustic p-wave velocities and amplitudes, and calculated upstream/downstream permeability to ensure a steady state condition is achieved.
The following summarizes our experimental observations:
• initially permeabilities (in the nD to low mD range) change up to two orders of magnitudes due to CO₂-mineral and CO₂-fluid
interactions;
• rehydration reduces permeability;
• compressional velocities and associated amplitudes provide a real time indication of CO₂ saturation changes and its impact
on rock moduli: velocities vary up to ±50 m/s (or ±1%); these relatively small variations suggest that saturation as well rock
frame (matrix) changes within a week’s testing period due to CO₂ injection do not have a marked impact on the rock’s elastic
constitutive behavior;
• given the small anhydrite porosity saturating its available pore space with brine accelerates acoustic energy transmission
Layered double hydroxide helps LaFeO3 photocatalyst activate peroxymonosulfate to efficiently degrade dyes
No full textPhotocatalysis is recognized as an environmentally benign technology for water treatment, yet its limited efficiency has impeded its wider application. Combining it with persulfate oxidation - a technique known for rapid reaction kinetics and potent contaminant removal capabilities - may enhance the efficacy of photocatalysis. However, the development of highly efficient catalysts for the synergistic effect of photocatalysis and persulfate oxidation remains challenging. To this end, we report a novel catalyst consisting of layered double hydroxides (LDH) and lanthanum ferrite (LFO). This LDH@LFO catalyst not only effectively combines photocatalysis and persulfate oxidation, but also establishes a p-n heterojunction. Optical and photoelectrochemical analysis shows that the catalyst significantly improves the photogenic carrier separation. We performed a thorough study to understand the influence of various factors on the catalysts’ performance, e.g., reaction systems, LFO loading, photocatalyst dosage, peroxymonosulfate (PMS) and dye concentration, pH value, coexisting anions, catalyst reusability and applicability. Under optimal conditions, the catalyst achieved a 93% reduction of the dye AR27 in 20 min, enhancing the degradation rate by at least 1.6-fold compared to conventional photocatalysis or persulfate oxidation alone. Investigations into the mechanism, through batch comparison tests, oxygen species quenching assays and electron paramagnetic resonance, have illuminated the essential roles of 1O2 and h+ etc., with Co and Fe redox cycles on LDH critical for PMS activation. Overall, this research illustrates the design of a heterojunction photocatalyst for efficiently mitigating environmental pollution.The authors sincerely thank the support from the National Natural Science Foundation of China (52100072), the Beijing Natural Science Foundation (2232051), the R&D Program of Beijing Municipal Education Commission (KM202110017008), the Young Elite Scientists Sponsorship Program by BAST (BYESS2023258), theShijiazhuang high-level science and technology innovation and entrepreneurship talent project (No. 08202303). Science and Technology Project of Yingze District in 2023 (Taiyuan City, Shanxi Province), King Abdullah University of Science and Technology (KAUST), and Beijing Institute of Petrochemical Technology
Sediment routing systems of the eastern red sea rifted margin
No full textWe investigate the sediment routing systems of the eastern Red Sea rifted margin by constraining the sediment accumulation history of the offshore depositional domain from regional seismic sections, wells, and outcrops observations and comparing it to the denudation history of the onshore erosional domain (Arabian Shield - Delaunay et al. (2024, submitted in this issue). We show that the rift (28–16 Ma) was segmented into a northern high-relief segment (28°N-21.5°N) under marine depositional environments and a southern low-relief magmatic segment (21.5°N-13°N) under continental depositional environments. The late syn-rift (Transition period - 16-14 Ma) was associated with a decrease in the magmatic and tectonic activity of the rifted margins, marine flooding of the entire basin, and the onset of the evaporitic sequence resulting from partial isolation from the global ocean. During the early post-rift (14–5 Ma), the southern segment underwent an uplift and a significant increase of the siliciclastic accumulation (15 fold), suggesting a rapid retreat of the escarpment initially corresponding to the former rift shoulder. During the late post-rift (5–0 Ma) the siliciclastic accumulation dropped and carbonate sedimentation returned suggesting a shift towards arid climatic conditions and a slowdown in the retreat of the onshore escarpment.This study was completed under KAUST's Competitive Research Grant 4082 on “Geologic Evolution of the Red Sea and Gulf of Aqaba” leaded by Professor Abdulkader M. Afifi. Petroleum companies operating in the Red Sea are acknowledged for the opportunity to interpret their data for scientific purposes. We would also like to thank Julien Bailleul for inviting us to participate in this special issue and two anonymous reviewers for their constructive comments, which helped to improve the quality of our manuscript
Faradaic Rectification in Electrochemical Deionization and Its Influence on Cyclic Stability
No full textCapacitive deionization (CDI) is a typical configuration of electrochemical deionization, which suffers from severe desalination capacity degradation derived from uncontrolled parasitic reactions. In this work, Faradaic rectification, the phenomenon by which electrode potentials and side reactions are dynamically regulated due to the asymmetrical anode/cathode Faradaic reactions, was studied under various CDI operation conditions. It was found that the Faradaic rectification in CDI would lead to capacity degradation indirectly by accelerating carbon anode oxidation and would be influenced by the cell voltage, flow rate, and asymmetric electrode construction. We also found an unconventional degradation mechanism in Faradaic cathode hybrid-CDI (HCDI) caused by the dramatic electrode-potential redistribution, which is derived from Faradaic rectification rather than the electrode structure decay. By adding a cation-exchange membrane to block the dissolved oxygen from cathode, the Faradaic rectification was suppressed successfully, and thus, the cyclic performance of CDI and HCDI was significantly increased by 59 and 46%, respectively (in 100 h cycling). This study provides an insight into understanding the Faradaic rectification in electrochemical deionization and its influence on CDI/HCDI cyclic stability, which should be of value to future explore cost-competitive membrane-less electrochemical deionization construction.This work was financially supported by the National Natural Science Foundation of China (grant number 2021YFD1700800). The authors are also grateful to KAUST for the very generous financial support. The authors wish to acknowledge Li Wang (College of Environmental Science and Engineering, Tongji University), for his insightful comments on this manuscript
Open Access Week Presentation: FAIR Data Management for Computational Materials Science using NOMAD
No full textScientific research is becoming increasingly data centric, which requires more effort to manage, share, and publish data. NOMAD is a web-based platform that provides research data management (RDM) for materials-science data. In addition to core RDM functions like uploading and sharing files, NOMAD automatically extracts structured data from supported file formats, normalizes, and converts data from these formats. NOMAD provides an extendable framework for managing not just files, but structured machine-actionable harmonized and interoperable data. This is the basis for a faceted search with domain-specific filters, a comprehensive API, structured data entry via customizable ELNs, integrated data-analysis and machine-learning tools. NOMAD is run as a free public service and can additionally be operated by research institutes. Connecting NOMAD installations through the public services will allow a federated data infrastructure to share data between research institutes and further harmonize RDM within a large research domain such as materials science.
Speakers
Joseph F. Rudzinski, Coordinator of Computational Data Infrastructure FAIRmat, Humboldt University of Berlin, Germany.
Joseph Rudzinski is the Coordinator of Computation for the FAIRmat consortium. He studied Chemistry and Mathematics at UC Santa Barbara and earned his Ph.D. in theoretical chemistry from Penn State, focusing on software for coarse-grained simulation models. After a Humboldt Postdoctoral Fellowship at the Max Planck Institute for Polymer Research (MPIP), where he studied coarse-grained dynamics, he became a group leader at MPIP, emphasizing data-driven techniques for molecular simulation analysis. In 2022, he joined FAIRmat, leading a team to develop research data infrastructure tools for computational materials science within the NOMAD software. His team has expanded NOMAD to support various advanced methodologies and implemented standardized workflows for efficient data provenance and AI-ready dataset curation.
Alex Fuchs, The Lead Developer of NOMAD CAMELS, FAIRmat, Humboldt University of Berlin, Germany.
Alex Fuchs is the Lead Developer of NOMAD CAMELS for the NFDI Consortium FAIRmat Since 2021. He studied Physics at the University of Stuttgart and earned his Ph.D. in Applied Physics from Friedrich-Alexander University (FAU) Erlangen-Nürnberg
Modeling transient natural convection in heterogeneous porous media with Convolutional Neural Networks
No full textConvolutional Neural Networks (CNNs) are gaining significant attention in applications related to coupled flow and transfer processes in porous media, especially when dealing with image-like data. In this context, the most important applications are related to surrogate modeling, where data obtained from simulators is used to train a CNN model. CNNs are also used as optimizers for inverse modeling or parameter estimation. For natural convection in porous media, applications of CNNs are scarce and limited to steady-state data. The main goal of this paper is to extend the applications of CNNs to transient data, by developing new CNN models that allow for integrating time-variant images. Thus, we suggest using an Encoder-Decoder CNN (ED-CNN) for surrogate modeling and a 3D-CNN for inverse modeling. Besides surrogate and inverse modeling, we suggest using CNN for time prediction by coupling it with long short-term memory (LSTM). The performances of these suggested approaches are investigated by applying them to the benchmark of natural convection in porous cavity with heterogeneous property fields, and by comparing the suggested approaches to other alternatives such as standard deep neural network (DNN) and 2D-CNN trained on steady-state data. The results show that, for surrogate modeling, with the same amount of data and equivalent training times, ED-CNN is more practical than DNN because it provides spatially distributed prediction while DNN is limited to local data. The transient data allows for improving the performance of CNN in inverse modeling because it provides more information about heat transfer across the different material zones and thus heterogeneity. The 3D-CNN approach is more efficient than 2D-CNN as it allows for considering the time progress in the training. For instance, the error with 3D-CNN and transient data is about 11 %, while it is about 18 % with 2D-CNN. Coupling CNN with LSTM allows for improving the performance of CNN in time series prediction.We thank Aqeel Afzal Chaudhry from the Geotechnical Institute at TU Bergakademie Freiberg for help in setting up the OGS simulations. We further acknowledge funding by the German Research Foundation (DFG, project INFRA NA1528/2–1 and MA4450/5–1). This work was motivated by results from a joint MSc thesis of the first author between Technische Universität Bergakademie Freiberg and the University of Strasbourg which was supported by the PROCOPE program
The KAUST Repository: Evolving Beyond Traditional Publications
No full textThe KAUST Repository serves as the foundation for scholarly communication at KAUST, collecting, preserving, and disseminating the research output of faculty, researchers, and students. This paper explores the evolution of the repository beyond traditional publications (conference papers, journal articles, dissertations, and theses) to include a broader range of research-related works, including valuable data collections.
Recognizing the evolving needs of research, a flexible approach to managing scholarly information is essential to support the mission of research institutions. As cross-disciplinary research becomes increasingly data- and software-intensive, traditional library services and infrastructure, primarily designed to handle textual documents, are inadequate.
The KAUST Repository has developed a process to collect and register information about KAUST-affiliated datasets and software, including machine-readable relationships with publications. The workflow tracks both datasets and code related to KAUST publications and standalone materials, updating active publication tracking procedures. It also queries external services such as DataCite, Crossref, GitHub, and GenBank to retrieve KAUST-affiliated work.
This paper will focus on two case studies that exemplify our collaborative efforts in curating and making such valuable data discoverable. These projects were achieved through close collaboration with KAUST researchers, faculty, and community members.
1. Local Birds Observations Collection: This project aimed to convert unstructured bird observation data into a standardized, structured format using the Darwin Core metadata standard. By integrating this data with the eBird global citizen science platform, we facilitated the collection of new observational records. We will discuss the tools employed, including KNIME for data extraction and conversion, and explore how this project serves as a foundation for future biodiversity information curation efforts within our institution.
2. Coral Specimen Collection: This project focuses on archiving information of coral specimens maintained by researchers at the KAUST Red Sea Center. The collection contains a comprehensive range of data, including the taxonomic classification of the coral (species, genera, subspecies, varieties, etc.), geographic location data, and images of both live animals and coral skeletons. Furthermore, we capture the links between specimen records and relevant publications, as well as GenBank accession pages.
This paper will highlight key lessons learned from these projects and outline strategic next steps and priorities to facilitate the collection, curation, and dissemination of a broader range of research materials. By embracing collaboration and utilizing innovative approaches, we aim to strengthen the KAUST Repository's role as an important hub for scholarly communication within our institution
Low diversity and abundance of predatory fishes in a peripheral coral reef ecosystem
No full textSemi-enclosed seas are often associated with elevated local threats and distinct biogeographic patterns among marine fishes, but our understanding of how fish assemblage dynamics vary in relation to relatively small semi-enclosed seas (e.g., the Gulf of Aqaba) remains limited. Baited remote underwater video surveys (n = 111) were conducted across ~300 km of coral reef habitats in the Gulf of Aqaba and the northern Red Sea. A total of 55 predatory fish species were detected, with less than half of all species (n = 23) observed in both basins. Relative abundance patterns between the Gulf of Aqaba and the northern Red Sea were variable among taxa, but nearly twice as many predatory fish were observed per unit of effort in the northern Red Sea. In general, assemblages in both basins were dominated by three taxa (Epinephelinae, Carangidae, and Lethrinidae). Large-bodied and threatened species were recorded at very low abundances. Multivariate analysis revealed distinct assemblage structuring of coral reef predators between the Gulf of Aqaba and the northern Red Sea. Most of the species driving these differences were recorded in both basins, but occurred at varying levels of abundance. Environmental factors were largely unsuccessful in explaining variation in assemblage structuring. These findings indicate that biological assemblages in the Gulf of Aqaba are more distinct than previously reported and that reef fish assemblage structuring can occur even within a relatively small semi-enclosed sea. Despite inter-basin assemblage structuring, the overall low abundance of vulnerable fish species is suggestive of overexploitation in both the Gulf of Aqaba and the northern Red Sea of Saudi Arabia. As the region surveyed is currently undergoing large-scale coastal development, the results presented herein aim to guide spatial management and recovery plans for these coral reef systems in relation to this development.The authors thank OceanX and the crew of the OceanXplorer for their hospitality and logistical assistance during field operations. We are also grateful for field assistance from Prof. Francesca Benzoni, Dr. Tullia Isotta Terraneo, and Silvia Vimercati from the Habitat and Benthic Biodiversity lab (KAUST), as well as Prof. Sam Purkis (University of Miami). Thamer Habis (Saudi Water Sports & Diving Federation) was vital for this project in providing essential logistical support. Dr. Alison Green and Alexander Kattan provided valuable feedback on the initial manuscript. The authors would also like to thank Sean Ruggeri (OceanX Media) for his photographic documentation (see Figure S3)