Data Structure Research Articles

Consensus local graph for multiple kernel clustering

Due to the ability to represent the relationship among data, graph has received extensive attention in clustering field. As illustrated by existing studies, learning graph in kernel space is an effective way to capture the structure of data. Among kernel-based methods, multiple kernel learning (MKL) is increasingly popular since it can automatically utilize the complementary information contained in base kernels. Although many existing MKL-based graph learning algorithms have promising learning abilities, we observe that they (1) put too much attention on learning the consensus kernel which probably contains lots of redundant information and loses the diversity of base kernels; (2) ignore the local structure of the representations in multiple kernel spaces. To eliminate the bad effect of these issues, we propose a novel graph learning method, termed consensus local graph based on MKL (CLGMKL), for clustering. In CLGMKL, a low-rank kernel matrix is used to extract the discriminative information of each base kernel and a local graph is constructed to capture the local structure contained in each new kernel. Then CLGMKL combines these local graphs in a self-weighed way. Since the above processes are associated with each other, we jointly optimize them to obtain the overall optimality. An effective learning scheme with proved convergence is developed to optimize the combined objective function. Finally, extensive experiments on some popular datasets are conducted to test the effectiveness of the presented method. As illustrated, CLGMKL is more competitive than the state-of-the-art algorithms.

Stability of metal-metal interactions in transmetallation intermediates based on electronics of bridging arene ligands determined through pyridine titrations.

In this contribution, we prepare the dinuclear complex [(CNCF)(PPh3)Pt-Au(PPh3)]+ (2-F) supported by an electron deficient derivative of 2,6-diphenylpyridine (CNC), 2,6-di(4-fluorophenyl)pyridine (CNCF). Solution state spectroscopic data and solid-state structural data reveals formation of the desired dinuclear complex occurs and that it remains intact in solution. The solid state structure of 2-F, compared to [(CNC)(PPh3)Pt-Au(PPh3)]+ (2), reveals a substantial change in the C-Au-P bond angle. We postulated that this change in bond angle arises due to a weaker interaction between [(PPh3)Au]+ and (CNCF)Pt(PPh3) (1-F) vs. (CNC)Pt(PPh3) (1). Through pyridine titration experiments, we demonstrate that the interaction is indeed weaker between [(PPh3)Au]+ and 1-Fvs. 1. Cyclic voltammetry (CV) experiments confirm that 1-F is less electron rich than 1. DFT calculations demonstrate that the HOMO of 1 and 1-F is not dz2, helping explain the differences in electrochemical behavior of 1 and 1-F and bonding between 1 and 1-F with [(PPh3)Au]+.

PhAI: A deep-learning approach to solve the crystallographic phase problem.

X-ray crystallography provides a distinctive view on the three-dimensional structure of crystals. To reconstruct the electron density map, the complex structure factors [Formula: see text] of a sufficiently large number of diffracted reflections must be known. In a conventional experiment, only the amplitudes [Formula: see text] are obtained, and the phases ϕ are lost. This is the crystallographic phase problem. In this work, we show that a neural network, trained on millions of artificial structure data, can solve the phase problem at a resolution of only 2 angstroms, using only 10 to 20% of the data needed for direct methods. The network works in common space groups and for modest unit-cell dimensions and suggests that neural networks could be used to solve the phase problem in the general case for weakly scattering crystals.

Ensemble Neural Models for ICD Code Prediction using Unstructured and Structured Healthcare Data

Disease coding is the process of assigning a standardized diagnostic code called ICD (International Classification of Diseases) to clinical notes that are maintained by health practitioners (e.g. clinicians) to track patient prognosis. Such a coding process is often expensive and error-prone, Because human medical coders primarily perform it. Automating ICD coding using Artificial Intelligence is seen as an essential solution in Hospital Information Management Systems (HIMS) and approaches built on Convolutional Neural Networks are currently state-of-the-art (SOTA). In this work, a neural model built on unstructured clinical text for enabling automatic ICD coding for given patient discharge summaries is proposed. We incorporate a structured self-attention mechanism designed to boost learning of label-specific vectors and the significant clinical text snippets associated with a certain label for this purpose. These vectors are then combined with a novel code description pipeline leveraging the descriptions provided for each standardized ICD code. The proposed model achieved SOTA performance in terms of the AUC, F1 and P@k metrics over the MIMIC-III dataset outperforming models based on Longformers and Knowledge graphs. Furthermore, to leverage structured clinical data to enhance the proposed model, and to enable improved ICD code prediction, model ensembling is considered. A neural model built on structured data by leveraging supervised machine learning algorithms such as random forest and boosting, is designed for multi-class classification of ICD codes. Experimental results revealed that the proposed ensemble models show promising performance compared to traditional models that rely solely on unstructured or structured clinical data, emphasizing their suitability for real-world deployment.

Oblique Electrostatic Solitary and Supersolitary Waves in Earth’s Magnetosheath

Nonlinear electrostatic solitary waves (ESWs) are routinely detected at various regions of Earth’s magnetosphere using the Wideband Data plasma wave receivers mounted on board the Cluster satellite(s). This mission has facilitated the observation and analysis of ESW characteristics, such as amplitude and temporal duration within the magnetosheath, while concurrently determining the density and temperature profiles of energetic electrons. These electron parameters, in conjunction with data from ion experiments, have served as input for the ion-acoustic solitary wave model developed in this article, with the ambition to contribute to the understanding of ESW generation mechanisms. Assuming as the starting point a plasma system comprising inertial ion fluid and kappa-distributed electron populations of different temperatures (i.e., “cold” and “hot” electrons), a nonperturbative approach has been adopted to investigate the existence and properties of solitary waves. A thorough parametric investigation has scrutinized the existence conditions for such localized structures in terms of the plasma configuration parameters. An interesting aspect emerges from the analysis, namely, the possibility for the coexistence of positive and negative polarity structures associated with ion-acoustic modes, in fact, manifested as simultaneously occurring positive polarity supersolitary waves and negative polarity regular solitary waves. Furthermore, our study has investigated the combined effect of the magnetic field strength, electron density, and suprathermal electron statistics on wave dynamics. The outcomes of this research are in agreement with observed electrostatic wave phenomena in the magnetosheath region, thus underscoring the intrinsic relevance of electrostatic supersolitary structures in data obtained by Cluster and other satellite missions.

Recommended Nuclear Structure and Decay Data for A=202 Isobars

Evaluated nuclear structure and decay data for all nuclei with mass number A=202 (202Os, 202Ir, 202Pt, 202Au, 202Hg, 202Tl, 202Pb, 202Bi, 202Po, 202At, 202Rn, 202Fr and 202Ra), are presented. All available experimental data are compiled and evaluated, and best values for level and γ-ray energies, quantum numbers, lifetimes, γ-ray intensities and transition probabilities, as well as other nuclear properties, are recommended. Inconsistencies and discrepancies that exist in the literature are discussed. A number of computer codes (https://www-nds.iaea.org/public/ensdf_pgm/) developed by members of the NSDD network were used during the evaluation process. This work supersedes the earlier evaluation by F.G. Kondev (2007Ko42), published in Nuclear Data Sheets108, 1471 (2007).

Comment on: The La Lucette Sb-Au-(W) vein-deposit (Armorican Massif, France): New time and genetic constraints to decipher the 310–295 Ma Sb-Au metallogenic peak in the Variscan Belt by

Recently, Cheval-Garabédian et al. (2023) published a paper in Ore Geology Reviews entitled “The La Lucette Sb-Au-(W) vein-deposit (Armorican Massif, France): new time and genetic constraints to decipher the 310–295 Ma Sb-Au metallogenic peak in the Variscan Belt.” Studies on Sb mineralisation are highly welcomed, especially in France, where there is a lack of updated data hindering a comprehensive understanding of this mineral system. The paper provides new mineralogical, fluid inclusion and geochronological data from historical collection samples of La Lucette mine, a mining site no more accessible since at least 40 years (Serment, 1978). These data, together with structural data from literature, lead the authors to build a metallogenic model related to dextral shear zone during the Carboniferous-Permian transition. The authors further argue for a single and unique 310–295 Ma Sb-Au metallogenic peak in the European Variscan Belt, a late Variscan hydrothermal event coeval with post-thickening tectonics (Munoz et al., 1992; Bouchot et al., 1997, 2005). This assertion is not correct. In this comment, we highlight that (i) several errors and approximations question the validity of their model, and (ii) the authors omit certain information that contradicts their metallogenic model. In our view, their data do not fully support the existence of a generalised and unique Late Carboniferous Sb-Au metallogenic peak in the French Variscan Belt. Furthermore, the idea defended by the authors that ore deposits occurred during dextral strike-slip tectonics that affected the area may better support a possible long-lived history during Palaeozoic times

A new possibilistic-based clustering method for probability density functions and its application to detecting abnormal elements

Cluster analysis can also detect abnormalities besides building a basis for identifying elements into clusters. Detecting abnormalities is a highly developed feature in the field of unsupervised learning. However, existing studies have mainly focused on discrete data, not probability density functions. This paper enables a possibilistic approach to solving the clustering for probability density functions dealing with abnormal elements. First, the data are extracted using the density function. Then, they are passed through the proposed algorithm to produce a possibilistic partition. Finally, a decision rule is established to recognize which function is abnormal. We compare the proposed algorithm with baseline algorithms in clustering PDFs, such as k-means, FCF, and Self-Updated Clustering. The results of three numerical examples applied to the image are typical for this new method. Furthermore, The proposed algorithm reaches accuracy at 100% over simulated benchmark data and outperforms baseline methods. Additionally, two last examples apply to image data reaching G-mean up from 96 to 100% (Sensitivity: 92–100% and Specificity: 100%). The proposed method can be researched and used to understand the internal structures of big data in the digital age through the probability density functions.

Covariate-dependent spatio-temporal covariance models

Geostatistical regression models are widely used in environmental and geophysical sciences to characterize the mean and dependence structures for spatio-temporal data. Traditionally, these models account for covariates solely in the mean structure, neglecting their potential impact on the spatio-temporal covariance structure. This paper addresses a significant gap in the literature by proposing a novel covariate-dependent covariance model within the spatio-temporal random-effects model framework. Our approach integrates covariates into the covariance function through a Cholesky-type decomposition, ensuring compliance with the positive-definite condition. We employ maximum likelihood for parameter estimation, complemented by an efficient expectation conditional maximization algorithm. Simulation studies demonstrate the superior performance of our method compared to conventional techniques that ignore covariates in spatial covariances. We further apply our model to a PM2.5 dataset from Taiwan, highlighting wind speed’s pivotal role in influencing the spatio-temporal covariance structure. Additionally, we incorporate wind speed and sunshine duration into the covariance function for analyzing Taiwan ozone data, revealing a more intricate relationship between covariance and these meteorological variables.

Exploring heterophily in calibration of graph neural networks

With growing applications in safety-critical domains such as autonomous vehicles and healthcare systems, graph neural networks (GNNs) are expected to provide predictions that are not only accurate but also reliable. Confidence calibration is a crucial task closely related to improve the reliability of machine learning models. Current GNN calibration methods primarily address challenges arising from the non-Euclidean topology structure of graph data and a general trend of under-confidence. However, our empirical studies have shown that these methods often fail to deal with local miscalibrations where community homophily diverges from global ones, such as high-homophily nodes in heterophilous graphs. To tackle this issue, we introduce Adaptive Spectral Temperature Scaling (ASTS), a novel spectral-based calibration method tailored for graphs with varying levels of homophily. On top of the node-wise temperature scaling framework, ASTS incorporates two components: (1) dual-channel graph filters that adaptively integrate similar and dissimilar information from neighborhood, (2) an innovative adaptive edge dropout module that alleviates issues induced by low-degree nodes. These designs ensure that ASTS is structure-aware, global trend-aware and well-adapted to different graph structures. Our extensive empirical studies confirm ASTS’s superior calibration performance over existing methods across graphs of diverse homophily ratios.

Piezoresistive nanocomposite sensing using electrical impedance tomography and machine learning

Strain measurements have traditionally been accomplished by connecting many gauges to crucial spots on structures. Installing strain gauges can be costly and limited in providing ‎comprehensive data for large surface areas or complex structures, posing significant challenges. This work explores the use of piezoresistive nanocomposite sensors as a cost-effective and informative solution for wide surface area strain detection. Piezoresistive sensors are created using two methodologies: film-coated and pellet-fed 3D printing technique (PF3DP), with variable gauge factors achieved by using 6 and 10 wt% multi-wall carbon nanotubes (MWCNTs), respectively. An electrical impedance tomography (EIT) setup is utilized to analyze electrical changes with respect to strain and predict the conductivity distribution across the sensor’s surface. This study employs two efficient machine learning (ML) techniques, artificial neural networks (ANN) and convolutional neural networks (CNN), to analyze and compare image reconstruction resolution at different loads. The machine learning models are designed to reconstruct images quickly and reliably. The study demonstrates that the hybrid-ML model, which combines ANN and CNN models, achieves high-resolution image construction with minimal errors compared to outcomes obtained through EIT for the experimental measurements. Sensors created using 3D printing provide higher image resolution and data stability than film-coated sensors. This technology has the potential for accurate strain sensing and measuring the structural health of complex surfaces.

Selective conversion of spent cracking catalyst to faujasite-structured zeolites

Current recycling strategies for spent catalyst in Fluid Catalytic Cracking (FCC) focus on valuable rare-earth elements (REE) and sometimes include a reuse of spent materials by down-cycling, which hinders a circular economy. Herein, a new procedure for the selective conversion of spent FCC catalysts to FAU-type zeolite without secondary crystalline phases such as sodalite, zeolite P or zeolite A is reported. The steps include a leaching process of REE, followed by thermal activation with sodium hydroxide and acid treatment to get an aluminum-rich feed. The synthesis of FAU-type zeolites with varying Si/Al ratios uses the aluminum-rich catalyst feed and commercial water glass. The experiments show that incorporation of silicate from water glass into the zeolite structure is limited to zeolite products with Si/Al ratio ≤ 2.0. It indicates a flocculation of silicates in higher concentration in competition to the crystallization process of FAU-type zeolites. This is reinforced by the resulting crystals, covered and surrounded with small particles with morphology of silica. Moreover, a clear differentiation between zeolite Y and zeolite X is possible from materials properties with the definite evidence of a mixed phase. The kinetics of feed dissolution and incorporation of impurities, alumina and silica are controlled by the pre-treatment temperature, duration and silicon source. This enables flexible adjustment of the aluminum re-use from spent FCC catalyst, the amount of foreign elements (Fe, Ni, V), phase purity, Si/Al ratio, specific surface area and thermal stability of the final zeolite product. Structural and textural data of the resulting zeolites reveals high crystallinity > 80%, Si/Al ratios of 1.6 - 2.0, and specific surface areas up to 780 m2/g.

Universality in the Structure and Dynamics of Water under Lipidic Mesophase Soft Nanoconfinement.

Water under soft nanoconfinement features physical and chemical properties fundamentally different from bulk water; yet, the multitude and specificity of confining systems and geometries mask any of its potentially universal traits. Here, we advance in this quest by resorting to lipidic mesophases as an ideal nanoconfinement system, allowing inspecting the behavior of water under systematic changes in the topological and geometrical properties of the confining medium, without altering the chemical nature of the interfaces. By combining Terahertz absorption spectroscopy experiments and molecular dynamics simulations, we unveil the presence of universal laws governing the physics of nanoconfined water, recapitulating the data collected at varying levels of hydration and nanoconfinement topologies. This geometry-independent universality is evidenced by the existence of master curves characterizing both the structure and dynamics of simulated water as a function of the distance from the lipid-water interface. Based on our theoretical findings, we predict a parameter-free law describing the amount of interfacial water against the structural dimension of the system (i.e., the lattice parameter), which captures both the experimental and numerical results within the same curve, without any fitting. Our results offer insight into the fundamental physics of water under soft nanoconfinement and provide a practical tool for accurately estimating the amount of nonbulk water based on structural experimental data.

A Structure-based Data Set of Protein-peptide Affinities and its Nonredundant Benchmark: Potential Applications in Computational Peptidology.

Peptides play crucial roles in diverse cellular functions and participate in many biological processes by interacting with a variety of proteins, which have also been exploited as a promising class of therapeutic agents to target druggable proteins over the past decades. Understanding the intrinsic association between the structure and affinity of protein-peptide interactions (PpIs) should be considerably valuable for the computational peptidology area, such as guiding protein-peptide docking calculations, developing protein-peptide affinity scoring functions, and designing peptide ligands for specific protein receptors. We attempted to create a data source for relating PpI structure to affinity. By exhaustively surveying the whole protein data bank (PDB) database as well as the ontologically enriched literature information, we manually curated a structure- based data set of protein-peptide affinities, PpI[S/A]DS, which assembled over 350 PpI complex samples with both the experimentally measured structure and affinity data. The data set was further reduced to a nonredundant benchmark consisting of 102 culled samples, PpI[S/A]BM, which only selected those of structurally reliable, functionally diverse and evolutionarily nonhomologous. The collected structures were resolved at a high-resolution level with either Xray crystallography or solution NMR, while the deposited affinities were characterized by dissociation constant, i.e. Kd value, which is a direct biophysical measure of the intermolecular interaction strength between protein and peptide, ranging from subnanomolar to millimolar levels. The PpI samples in the set/benchmark were arbitrarily classified into α-helix, partial α-helix, β-sheet formed through binding, β-strand formed through selffolding, mixed, and other irregular ones, totally resulting in six classes according to the secondary structure of their peptide ligands. In addition, we also categorized these PpIs in terms of their biological function and binding behavior. The PpI[S/A]DS set and PpI[S/A]BM benchmark can be considered a valuable data source in the computational peptidology community, aiming to relate the affinity to structure for PpIs.

Cortical structure and subcortical volumes in conduct disorder: a coordinated analysis of 15 international cohorts from the ENIGMA-Antisocial Behavior Working Group

Conduct disorder is associated with the highest burden of any mental disorder in childhood, yet its neurobiology remains unclear. Inconsistent findings limit our understanding of the role of brain structure alterations in conduct disorder. This study aims to identify the most robust and replicable brain structural correlates of conduct disorder. The ENIGMA-Antisocial Behavior Working Group performed a coordinated analysis of structural MRI data from 15 international cohorts. Eligibility criteria were a mean sample age of 18 years or less, with data available on sex, age, and diagnosis of conduct disorder, and at least ten participants with conduct disorder and ten typically developing participants. 3D T1-weighted MRI brain scans of all participants were pre-processed using ENIGMA-standardised protocols. We assessed group differences in cortical thickness, surface area, and subcortical volumes using general linear models, adjusting for age, sex, and total intracranial volume. Group-by-sex and group-by-age interactions, and DSM-subtype comparisons (childhood-onset vs adolescent-onset, and low vs high levels of callous-unemotional traits) were investigated. People with lived experience of conduct disorder were not involved in this study. We collated individual participant data from 1185 young people with conduct disorder (339 [28·6%] female and 846 [71·4%] male) and 1253 typically developing young people (446 [35·6%] female and 807 [64·4%] male), with a mean age of 13·5 years (SD 3·0; range 7-21). Information on race and ethnicity was not available. Relative to typically developing young people, the conduct disorder group had lower surface area in 26 cortical regions and lower total surface area (Cohen's d 0·09-0·26). Cortical thickness differed in the caudal anterior cingulate cortex (d 0·16) and the banks of the superior temporal sulcus (d -0·13). The conduct disorder group also had smaller amygdala (d 0·13), nucleus accumbens (d 0·11), thalamus (d 0·14), and hippocampus (d 0·12) volumes. Most differences remained significant after adjusting for ADHD comorbidity or intelligence quotient. No group-by-sex or group-by-age interactions were detected. Few differences were found between DSM-defined conduct disorder subtypes. However, individuals with high callous-unemotional traits showed more widespread differences compared with controls than those with low callous-unemotional traits. Our findings provide robust evidence of subtle yet widespread brain structural alterations in conduct disorder across subtypes and sexes, mostly in surface area. These findings provide further evidence that brain alterations might contribute to conduct disorder. Greater consideration of this under-recognised disorder is needed in research and clinical practice. Academy of Medical Sciences and Economic and Social Research Council.

Molecular Docking of Phytomolecules of Grain Amaranth (Amaranthus hypochondriacus) with AKR1C3Protein Involved in Prostate Cancer in Human Beings

Abstract: This study focuses on exploring how the bioactive compounds found in amaranth— phytol, squalene, and α-tocopherol—could potentially offer medicinal benefits in the context of prostate cancer. The investigation involves a docking study with AKRC13, an important target linked to the control of prostate cancer, aiming to uncover their potential effects against this disease. Costeffective and efficient cancer treatment options are crucial because of the high expenses associated with current cancer therapies as well as their side effects. Amaranth (Amaranthus hypochondriacus) is a pseudocereal crop abundant in squalene, α-tocopherol, and phytol, which shows promising foodbased therapy for various diseases, including cancer. Prostate cancer has been a significant contributor to mortality globally, but the introduction of relugolix has emerged as a crucial therapeutic intervention in its treatment. Hence, this study was conducted to investigate the interactions between grain amaranth bioactive compounds squalene, phytol, and α-tocopherol with AKRC13 protein utilizing a molecular docking approach facilitated by Autodock Vina software. Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB-PDB) (http://www.rcsb.org/) was used for retrieving the 3D crystal structure of the target protein, AKRC13 (PDB ID: 7c7f). The 3D structure of bioactive compounds squalene, phytol, and α-tocopherol were retrieved from the PubChem database, following which Open-Babel was used to change the format from .sdf to .pdb. Furthermore, pharmacokinetics characteristics were also considered along with Lipinski’s rule of five using SwissADME (http://www. Swiss adme.ch/index.php) and pkCSM (http://structure.bioc.cam.ac.uk/pkcsm), indicating their potential as a drug candidate in the initial stage. The potential anticancer properties of the ligands were predicted using PASS software. Following the completion of the docking study, it became evident that α-tocopherol demonstrated the most significant binding energy, followed by squalene and phytol, in comparison to the established drug, relugolix. This implies that the chosen bioactive compounds might possess enzyme-inhibiting properties, indicating their potential for further in vivo anticancer screening using model organisms. The findings serve as stepping stones for advancing the potential use of the discussed bioactive compounds as a potential drug candidate for prostate cancer.

Benchmarking Methods for PROTAC Ternary Complex Structure Prediction.

Proteolysis targeting chimeras (PROTACs) are bifunctional compounds that recruit an E3 ligase to a target protein to induce ubiquitination and degradation of the target. Rational optimization of PROTAC requires a structural model of the ternary complex. In the absence of an experimental structure, computational tools have emerged that attempt to predict PROTAC ternary complexes. Here, we systematically benchmark three commonly used tools: PRosettaC, MOE, and ICM. We find that these PROTAC-focused methods produce an array of ternary complex structures, including some that are observed experimentally, but also many that significantly deviate from the crystal structure. Molecular dynamics simulations show that PROTAC complexes may exist in a multiplicity of configurational states and question the use of experimentally observed structures as a reference for accurate predictions. The pioneering computational tools benchmarked here highlight the promises and challenges in the field and may be more valuable when guided by clear structural and biophysical data. The benchmarking data set that we provide may also be valuable for evaluating other and future computational tools for ternary complex modeling.

Multiview Subspace Clustering via Low-Rank Symmetric Affinity Graph.

Multiview subspace clustering (MVSC) has been used to explore the internal structure of multiview datasets by revealing unique information from different views. Most existing methods ignore the consistent information and angular information of different views. In this article, we propose a novel MVSC via low-rank symmetric affinity graph (LSGMC) to tackle these problems. Specifically, considering the consistent information, we pursue a consistent low-rank structure across views by decomposing the coefficient matrix into three factors. Then, the symmetry constraint is utilized to guarantee weight consistency for each pair of data samples. In addition, considering the angular information, we utilize the fusion mechanism to capture the inherent structure of data. Furthermore, to alleviate the effect brought by the noise and the high redundant data, the Schatten p-norm is employed to obtain a low-rank coefficient matrix. Finally, an adaptive information reduction strategy is designed to generate a high-quality similarity matrix for spectral clustering. Experimental results on 11 datasets demonstrate the superiority of LSGMC in clustering performance compared with ten state-of-the-art multiview clustering methods.

Efficient Local Coherent Structure Learning via Self-Evolution Bipartite Graph.

Dimensionality reduction (DR) targets to learn low-dimensional representations for improving discriminability of data, which is essential for many downstream machine learning tasks, such as image classification, information clustering, etc. Non-Gaussian issue as a long-standing challenge brings many obstacles to the applications of DR methods that established on Gaussian assumption. The mainstream way to address above issue is to explore the local structure of data via graph learning technique, the methods based on which however suffer from a common weakness, that is, exploring locality through pairwise points causes the optimal graph and subspace are difficult to be found, degrades the performance of downstream tasks, and also increases the computation complexity. In this article, we first propose a novel self-evolution bipartite graph (SEBG) that uses anchor points as the landmark of subclasses, and learns anchor-based rather than pairwise relationships for improving the efficiency of locality exploration. In addition, we develop an efficient local coherent structure learning (ELCS) algorithm based on SEBG, which possesses the ability of updating the edges of graph in learned subspace automatically. Finally, we also provide a multivariable iterative optimization algorithm to solve proposed problem with strict theoretical proofs. Extensive experiments have verified the superiorities of the proposed method compared to related SOTA methods in terms of performance and efficiency on several real-world benchmarks and large-scale image datasets with deep features.

Interactive Volume Visualization via Multi-Resolution Hash Encoding Based Neural Representation.

Implicit neural networks have demonstrated immense potential in compressing volume data for visualization. However, despite their advantages, the high costs of training and inference have thus far limited their application to offline data processing and non-interactive rendering. In this article, we present a novel solution that leverages modern GPU tensor cores, a well-implemented CUDA machine learning framework, an optimized global-illumination-capable volume rendering algorithm, and a suitable acceleration data structure to enable real-time direct ray tracing of volumetric neural representations. Our approach produces high-fidelity neural representations with a peak signal-to-noise ratio (PSNR) exceeding 30 dB, while reducing their size by up to three orders of magnitude. Remarkably, we show that the entire training step can fit within a rendering loop, bypassing the need for pre-training. Additionally, we introduce an efficient out-of-core training strategy to support extreme-scale volume data, making it possible for our volumetric neural representation training to scale up to terascale on a workstation with an NVIDIA RTX 3090 GPU. Our method significantly outperforms state-of-the-art techniques in terms of training time, reconstruction quality, and rendering performance, making it an ideal choice for applications where fast and accurate visualization of large-scale volume data is paramount.