期刊,Knowledge-based systems 2025年309卷Jan.30期_国家学术搜索

期刊信息/Journal information

Knowledge-based systems

Elsevier Science

主办单位：Elsevier Science

国际刊号：0950-7051

Knowledge-based systems/Journal Knowledge-based systemsSCIAHCIISTPEI

正式出版

收录年代

Classification of cognitive syndromes in a Southeast Asian population: Interpretable graph convolutional neural networks

Ong, Charlene Zhi LinVipin, AshwatiLeow, Yi JinTanoto, Pricilia...

1.1-1.11页

查看更多>>摘要：Dementia is a debilitating disease that afflicts a large population worldwide. Early diagnosis of cognitive impairment can allow for preventative measures to be taken to slow down or prevent the progression to dementia. In this study, we devise an interpretable graph convolutional neural network approach, GCNEnsemble, using both non-clinical variables such as MRI preprocessed features including cortical thickness and gray matter volumes, and clinical features from a community-dwelling Southeast Asian population in Singapore aged between 30 and 95 years from the Biomarker and Cognition study (BIOCIS), to classify participants into cognitively normal, subjective cognitive decline, and mild cognitive impairment. We further conducted ablation studies and varied the quantities of labeled data to understand the contribution of the non-clinical features and the applicability of GCNEnsemble in low to high labeled data availability scenarios. GCNEnsemble was able to attain the highest accuracy and Matthew's correlation coefficient compared to existing state-of-the-art methods. Feature interpretability via Integrated Gradients identified features such as visual cognitive assessment test (VCAT), systolic and diastolic blood pressure, and cerebrospinal fluid volume as key features for the classification, with VCAT having the highest feature importance. There was higher median cerebrospinal fluid volume, right frontal pole thickness, left pallidum volume, and right hippocampal fissure volume but lower VCAT for the mild cognitive impairment group than the two other groups. In conclusion, GCNEnsemble can be used as a semisupervised interpretable classification tool for cognitive syndrome in a Southeast Asian population.

原文链接:

NETL
NSTL
Elsevier

Graph protection under multiple simultaneous attacks: A heuristic approach

Djukanovic M.Matic D.Kapunac S.Kartelj A....

1.1-1.18页

查看更多>>摘要：© 2024 Elsevier B.V.This work focuses on developing a meta-heuristic approach to protect network nodes from simultaneous attacks, specifically addressing the k-strong Roman domination problem. The objective is to assign integer weights to the nodes, representing the number of stationed armies, to meet protection constraints while minimizing the total number of armies. A network is protected if it can repel any simultaneous attack on k nodes. A node is protected if it can defend itself or if a neighboring node provides an army while retaining at least one army for self-defense. This problem formulation can be used in practical scenarios, e.g. developing counter-terrorism strategies or in coping with supply chain disruptions. The problem is difficult as even verifying the feasibility of a single solution generally requires an exponential time. Two exact approaches are proposed in the literature but applicable to small random graphs. For larger graphs, we propose an effective variable neighborhood search, where the feasibility of a solution is verified by introducing the concept of relaxed feasibility. Experiments are conducted with random networks from the literature and two introduced ad-hoc wireless and real-world networks. Extensive experimental evaluations show the robustness of the proposed approach compared to the existing approaches from the literature by significantly outperforming them in all three benchmark sets. Furthermore, we demonstrate the practical application of the proposed variable neighborhood search approach, where its solution is used to position fire stations within the city so that simultaneous fires can be extinguished efficiently while reducing the number of required fire trucks.

原文链接:

NETL
NSTL
Elsevier

DMAdam: Dual averaging enhanced adaptive gradient method for deep neural networks

Jiang W.Xu D.Liu J.Zhang N....

1.1-1.19页

查看更多>>摘要：© 2024 Elsevier B.V.Deep neural networks (DNNs) have achieved remarkable success in a wide range of fields, largely due to their stable and efficient optimizers. We propose a novel optimizer called Dual Momentum Adam (DMAdam), which combines the stability of dual averaging with the efficiency of adaptive gradient techniques. DMAdam adaptively tunes the learning rate and employs dual averaging updates, effectively balancing stability and convergence rate. This strategy enhances the control of DMAdam over gradient updates, resulting in superior performance in a variety of optimization tasks. Theoretically, we investigate the convergence properties of DMAdam for non-convex models and obtain the non-ergodic convergence of its gradient sequence. Numerically, we demonstrate the impressive performance of DMAdam on CIFAR-10 and CIFAR-100 datasets for image classification tasks. Additionally, DMAdam shows robust performance in natural language processing and object detection tasks. The PyTorch code of DMAdam is available at: https://github.com/Wenhan-Jiang/DMAdam.git.

原文链接:

NETL
NSTL
Elsevier

A multivariate time series anomaly detection method with Multi-Grain Dynamic Receptive Field

Chen L.Gao X.Wang T.Lu J....

1.1-1.16页

查看更多>>摘要：© 2024 Elsevier B.V.The multivariate time series (MTS) anomaly detection methods based on masked reconstruction pose challenges in model training by setting unknown areas to the data, compelling the model to explore deeper patterns to enhance its performance. Due to the low information density of MTS, point-masked methods relying on timestamps can only capture a limited amount of data information, while patch-masked methods based on segments can more effectively uncover advanced semantic features of underlying trends in MTS. However, patch-masked methods process MTS with either fixed or random masks, which may not only sacrifice the known information but also impose restrictions on the size of mask blocks during the reconstruction process. In this paper, a multivariate time series anomaly detection method with Multi-Grain Dynamic Receptive Field (MGDRF) is proposed. MGDRF designs multi-grain mask strategies to excavate semantic features of MTS ranging from lower to higher levels. The dynamic receptive fields are specifically crafted to mitigate information loss encountered in existing methods, thereby facilitating learning of temporal and dimensional relationships of the data. Furthermore, MGDRF incorporates the receptive-field-based and model-based layered losses. It establishes primary losses for each single-grain receptive field, enabling the extraction of different semantic features. Based on ensemble learning, MGDRF constructs a model-based loss through the fusion of outcomes from multiple grains of dynamic receptive fields, thereby further learning the interaction characteristics among different grains of MTS. Extensive experiments on five representative public datasets demonstrate that the proposed algorithm exhibits more advanced performance compared to 18 typical MTS anomaly detection methods.

原文链接:

NETL
NSTL
Elsevier

DE-PSA: Learning from unlabeled data by dual-stage label propagation for positive selection algorithm

Chen W.Yang Y.Liu L.

1.1-1.15页

查看更多>>摘要：© 2024Artificial immune detectors are the basic classification units for self/nonself discrimination. Traditional immune detector generation algorithms adopt supervised learning paradigm, relying on a large number of labeled samples to fully train the candidate detectors. However, in practical applications it is often difficult to obtain sufficient labeled training samples, resulting in model's insufficient learning problems. In the paper, we proposed a semi-supervised detector generation algorithm DE-PSA, which generates immune detectors using some initially labeled samples and a large number of unlabeled samples. DE-PSA consists of two main steps: dual-stage label propagation and detector generation. In the first stage of label propagation, pseudo labels are propagated to each unlabeled sample from its k-nearest labeled neighbors based on category influence calculations. According to the calculation results, we can select samples with highly credible pseudo (HCP) labels and partial labeled (PAL) samples which have multiple candidate labels. In the second label propagation stage, following partial label learning theory, category probabilities are iteratively propagated from the initially labeled and HCP labeled samples to the PAL samples to achieve label disambiguation; Subsequently, self (positive) and nonself (negative) samples are selected from the initially labeled samples, HCP labeled samples, and disambiguated PAL samples to constitute training set. Based on the set, DE-PSA generates self-detectors with variable radii using a positive selection process. Comprehensive tests on 10 standard datasets are carried out to test DE-PSA, and the true positive rates of self/nonself samples: TPS and TPN are taken as evaluation metrics. The results show that DE-PSA outperforms traditional algorithms, such that compared with newly proposed DGA-PSO, SA-PSA,HI-Detector, the average true positive rate of DE-PSA is raised by 23%, 19.5%, 16.5% respectively, and when compared with state-of-the-art algorithm co-PSA, only with 0.1‰ initially labeled training samples, DE-PSA and co-PSA has similar TPS, but DE-PSA's TPN is raised by 30%.

原文链接:

NETL
NSTL
Elsevier

Dual alignment feature embedding network for multi-omics data clustering

Xiao Y.Zou X.Tang C.Yang D....

1.1-1.11页

查看更多>>摘要：© 2024 Elsevier B.V.Multi-omics data clustering, with its capability to utilize the biological information of cross-omics to partition cells into their respective clusters, has attracted considerable attention due to its effectiveness for pathological analysis. Aside from cross-omics discrepancy, existing methods suffer from distribution differences, making it difficult to learn high-quality cross-omics consistent information. To tackle this issue, we propose a novel dual alignment feature embedding network for multi-omics data clustering (DAMIC). Specifically, we first utilize an attention-induced feature fusion mechanism to capture intra-omics specific and inter-omics structural information for more discriminative features. Moreover, we maximize the mutual information between the unified target distribution and other omics-specific assignments by simultaneously optimizing contrastive learning loss and Kullback–Leibler (KL) divergence loss. Finally, we can extract omics-invariant features with robust and rich common embeddings for multi-omics clustering. Extensive experimental results on six real-world benchmark datasets demonstrate that our approach surpasses existing state-of-the-art methods in multi-omics data clustering analysis, which provides effective pathologic analysis way for tumors such as Leukemia and Colorectal Neoplasms. The source code is available at https://github.com/YuangXiao/DAMIC.

原文链接:

NETL
NSTL
Elsevier

RVFL-LSTM: A lightweight model with long-short term memory for time series

Liu Q.Wang Q.Wang X.

1.1-1.12页

查看更多>>摘要：© 2024 Elsevier B.V.Neural networks have been widely used for time series prediction due to their excellent ability to capture the sequential relationship between the past and the future in time series data. However, the existing neural networks, such as long-short term memory(LSTM) and recurrent ones, are often criticized for their complex structure and strict training mode, but some lightweight network models often cannot achieve satisfying prediction performance. In order to tackle this challenge, this paper proposes a lightweight model with long-short term memory for time series, named RVFL-LSTM, which is trained for moving auto-regression task. To highlight the long and short-term patterns in time series data, RVFL-LSTM adjusts the input weights to learn the short-term patterns and updates the output weights gradually to capture the long-term patterns. Experimental results show that the proposed method can capture the long-term and short-term patterns efficiently, and it is very competitive in time series prediction with respect to prediction accuracy and computational complexity.

原文链接:

NETL
NSTL
Elsevier

A Multi-scale neighbourhood feature interaction network for photovoltaic cell defect detection

Liu Y.C.Hua Q.Chen L.L.Dong C.R....

1.1-1.13页

查看更多>>摘要：© 2024 Elsevier B.V.Photovoltaic power generation is a critical component in the industrial sector, with the efficiency of energy production being influenced by surface defects in photovoltaic cells. Recent advancements in defect detection have been largely driven by the widespread use of deep learning models. However, detecting defects at multiple scales especially small ones remains challenging due to the varying sizes of defects on photovoltaic cells. Additionally, the presence of significant noise in the images further complicates the extraction of distinguishable features. To address these challenges, this study proposes a novel, one-stage multi-scale neighbourhood feature interaction network (MNFI-Net) designed to detect defects of various sizes against complex backgrounds. The MNFI-Net architecture includes the following components: (1) Ghost cross-stage module, aimed at reducing redundant information; (2) neighbourhood feature interaction module, which enhances the model's ability to detect defects of different sizes; (3) global attention mechanism that focuses on highlighting key features in the fused feature maps. Additionally, for multi-scale defect detection tasks, we introduced a new balanced efficient loss function. Extensive comparison experiments and ablation studies were conducted on the public photovoltaic electroluminescence image dataset. The experimental results demonstrate that MNFI-Net achieves 94.0% precision and 95.5% mean average precision, outperforming existing state-of-the-art methods in defect classification and detection. The code and proposed models in this study can be accessed at https://github.com/lyc686/MNFI-Net.

原文链接:

NETL
NSTL
Elsevier

Enhancing associative classification on imbalanced data through ontology-based feature extraction and resampling

Mba Kouhoue J.Lonlac J.Doniec A.Lesage A....

1.1-1.12页

查看更多>>摘要：© 2024Associative classification models are valuable for discovering relationships within heterogeneous data systems, making them particularly useful for data integration tasks. However, they struggle with imbalanced and sparse data. This paper addresses the problem of imbalanced classification in building maintenance data by providing several updates based both on algorithms and preprocessing. Experiments conducted on real maintenance datasets demonstrate significant improvements in accuracy and precision.

原文链接:

NETL
NSTL
Elsevier

Adaptive random tree ensemble for evolving data stream classification

Paim, Aldo M.Enembreck, Fabricio

1.1-1.16页

查看更多>>摘要：Data stream mining with concept drift is a significant challenge in machine learning because this scenario requires the ability to handle unlimited and ever-changing data and real-time processing. An often employed strategy in data stream mining involves utilizing ensembles due to their capability to tackle concept drift and attain remarkably accurate predictions. However, developing a precise and efficient ensemble for data stream mining poses a significant challenge, as state-of-the-art algorithms are often highly inefficient, consuming excessive memory and processing time. In this study, we propose a novel ensemble-based classification algorithm for data streams named Adaptive Random Tree Ensemble (ARTE). The algorithm explores approaches that promote high prediction accuracy using a random-sized feature subspace for each element of the ensemble, online bagging, random choice of the cut-point for splitting the trees, and a method of classifier selection for final ensemble voting. This study also presents analyses on the contribution of the choice of subspace size and the random cut-point for splitting the tree's nodes to the ensemble's diversity. Following an extensive experimental investigation, ARTE exhibited high predictive performance and outperformed state-of-the-art ensembles on data streams for real and synthetic datasets while requiring fewer computational resources.

原文链接:

NETL
NSTL
Elsevier