NEURAL NETWORKS神经网络
NEURAL NETWORKS(英文缩写 NEURAL NETWORKS),ISSN 0893-6080,eISSN 1879-2782,中文译名:神经网络 是一本学术期刊。本页汇总该期刊的最新影响因子、分区信息以及最新收录于 PubMed 的文献,帮助您快速了解期刊全貌。
发文量统计区间：2025-11-01 至 2026-10-31，按本站收录文献的发表日期统计。
期刊介绍
历年影响因子趋势
| JCR 数据年份 | 影响因子 | JCR 分区 |
|---|---|---|
| 2021 | 9.657 | Q1 |
| 2022 | 7.800 | Q1 |
| 2023 | 6.000 | Q1 |
| 2024 | 6.300 | Q1 |
| 2025 | 7.200 | Q1 |
NEURAL NETWORKS 最新收录文献
-
1. Reinforcement learning-based predefined-time secure control for fractional-order nonlinear systems with sensor and actuator attacks.
1. 基于强化学习的分数阶非线性系统预设时间安全控制：应对传感器和执行器攻击PMID:日期:2026-09-29A secure control problem is investigated for constrained fractional-order nonlinear systems subject to both sensor and actuator attacks. To handle the effects induced by the dual attacks, a nonlinear mapping-based coordinate transformation is elaborately introduced to convert the original state-constrained system into an unconstrained form. Furthermore, by integrating reinforcement learning with predefined-time stability analysis, an adaptive control strategy is developed to achieve predefined-time convergence. The proposed method enhances the transient response of the system while accelerating the convergence of both the tracking error and the policy learning process. A novel reset-based self-triggered mechanism is proposed. On the one hand, the developed mechanism reduces communication load and eliminates the need for continuous state monitoring. On the other hand, the sampling intervals can be adaptively adjusted according to the system states, thereby further alleviating the computational burden. Under the developed control strategy, the closed-loop system is guaranteed to achieve semi-global practical predefined-time stability, and all signals converge to a bounded neighborhood of the origin within the predefined time.
-
2. MODdapter: Spatially-aware text embeddings for zero-shot semantic segmentation.
2. MOD可适应：面向零样本语义分割的空间感知文本嵌入PMID:日期:2026-09-28Vision-language models show promise in zero-shot semantic segmentation, but a key challenge is the disconnect between text and visual features. While text embeddings can roughly localize unseen objects, they often lack the fine-grained detail necessary for accurate segmentation, leading to oversegmentation or undersegmentation. To address this, we propose MODdapter, a Multimodal Adapter with Proportional-Integral-Derivative (PID) control for precise dual-space alignment of vision and language features. Our core idea is to leverage visual cues obtained during testing to complement textual information, improving the alignment between text and visual features. MODdapter extracts visual semantic cues from coarse segmentation results and integrates them as supplementary textual data, allowing for a more accurate projection of text features into the feature space and enhancing fine-grained recognition. To maintain stability and counteract potential oscillations caused by deviations in the extracted instance cues, we incorporate a PID control module that regulates the alignment process. PID control module leverages the combined joint constraints of short-term (derivative) and long-term (integral) memory, achieving fine-grained recognition of unseen target. This strategy significantly enhances the recognition of unseen objects, achieving state-of-the-art performance across multiple datasets. On the challenging COCO-20 dataset, our method outperforms the current state-of-the-art by a significant improvements of 6.9%.
-
3. BGNAS: Backdoored graph neural architecture search.
3. BGNAS：后门图神经架构搜索PMID:日期:2026-09-26Graph Neural Networks (GNNs) have been shown to be vulnerable to backdoor attacks, where attackers implant hidden backdoors into model weights of GNNs. However, backdoors hidden in GNN weights are non-persistent and easily erased by retraining models on clean data. To this end, we investigate a new threat of implanting backdoors in GNNs neural architectures that is independent of the model weights of GNNs. A new backdoored graph neural architecture search model (BGNAS for short) is proposed to use graph neural architecture search algorithms to discover new GNN neural architectures with backdoors. Unlike previous works that rely on manually designed backdoored neural architecture and triggers, BGNAS simultaneously designs backdoored GNN neural architectures associated with their trigger generators, where the generated triggers are indistinguishable within the overall graph data to ensure stealthiness. Theoretical analysis demonstrates that backdoors embedded in graph neural architectures remain effective even after weight re-initialization and retraining on clean data. Experimental results further show that BGNAS achieves high attack effectiveness while maintaining competitive prediction performance on benign data.
-
4. Corrigendum to "DAQSyn: A decentralized adaptive quantization-aware synchronous framework in heterogeneous on-device AI networks" [Neural Networks 205 (2027) 109474].
4. 对“DAQSyn：异构端侧AI网络中的去中心化自适应量化感知同步框架”[Neural Networks 205 (2027) 109474]的勘误PMID:日期:2026-09-26该文献暂无摘要。
-
5. Spectral-temporal dual view graph convolutional networks: A deep learning framework for dynamic bipartite graphs.
5. 谱-时双视图图卷积网络：一种用于动态二部图的深度学习框架PMID:日期:2026-09-25Dynamic bipartite graphs (DBGs) are widely used in real-world scenarios, where representation learning is particularly challenging due to the heterogeneity of node types and the temporal evolution of interactions. A key difficulty lies in jointly capturing non-stationary micro-level preference dynamics and macro-level stable structural evolution. Existing methods typically focus on either temporal dependency modeling or structural pattern extraction. For instance, recurrent or attention-based temporal models struggle to capture abrupt, short-term preference shifts, while spectral or structure-oriented approaches emphasize long-term regularities but overlook fine-grained dynamic variations. This imbalance becomes especially pronounced in scenarios with sparse interactions or sudden behavioral changes. To address these challenges, this paper proposes a spectral-temporal dual-view framework, ST-DYB, for DBG representation learning. By explicitly decoupling macro-level structural evolution and micro-level preference dynamics, the model captures complementary heterogeneous patterns from two perspectives. A cross-view adaptive fusion mechanism further integrates these representations, enabling effective modeling of both implicit preference dependencies and temporal dynamics. Extensive experiments on multiple real-world datasets demonstrate that ST-DYB consistently outperforms state-of-the-art methods in dynamic link prediction and node classification tasks, validating the effectiveness and robustness of the proposed dual-view modeling paradigm.
-
6. Cross-domain few-shot learning for hyperspectral image classification based on mixup foundation model.
PMID:日期:2026-09-25Although cross-domain few-shot learning (CDFSL) for hyperspectral image (HSI) classification has attracted significant research interest, existing works often rely on an unrealistic data augmentation procedure in the form of external noise to enlarge the sample size, thus greatly simplifying the issue of data scarcity. They involve a large number of parameters for model updates, being prone to the overfitting problem. To the best of our knowledge, only one work has explored the strength of the foundation model, yet it still relies on a conventional foundation model for RGB images. This paper proposes the MIxup FOundation MOdel (MIFOMO) for CDFSL of HSI classifications. MIFOMO is built upon the concept of a remote sensing (RS) foundation model, pre-trained across a large scale of RS problems, thus featuring generalizable features. The notion of coalescent projection (CP) is introduced to quickly adapt the foundation model to downstream tasks while freezing the backbone network. The concept of mixup domain adaptation (MDM) is proposed to address the extreme domain discrepancy problem. Last but not least, the label propagation concept is implemented to cope with noisy pseudo-label problems. Our rigorous experiments demonstrate the advantage of MIFOMO, where it beats prior art by up to a 14% margin. The source code of MIFOMO is open-sourced in https://github.com/Naeem-Paeedeh/MIFOMO for reproducibility and further study.
-
7. Transfer learning-based safe Q-learning for discrete-time nonlinear systems with asymmetric state constraints.
PMID:日期:2026-09-25This article investigates the problem of safe control for discrete-time (DT) nonlinear systems subject to asymmetric state constraints. A novel transfer learning-based safe Q-learning (TL-SQL) algorithm is developed within the adaptive dynamic programming (ADP) framework. To support the proposed TL-SQL algorithm, a novel revived transformation formulation for DT systems is derived, serving as the basis for transferring control policies from unconstrained source systems to constrained target systems. Building upon this foundation, the proposed TL-SQL algorithm improves policy transfer under asymmetric constraints in three aspects. Firstly, unlike traditional state transformation methods, the revived transformation alleviates the adverse impact of state transformation on the optimality of the original system. Secondly, compared with control-barrier-function-based approaches, it avoids complicating the utility function structure. Thirdly, by decoupling policy training from constraint handling, it enables efficient adaptation to changing constraints without retraining. Theoretical analysis ensures stability and safety of the transferred policy. Simulation results validate the effectiveness of TL-SQL under asymmetric constraint scenarios.
-
8. Kaczmarz-inspired acceleration for kernel canonical correlation analysis: Theory, algorithms, and its applications to information retrieval.
8. 受Kaczmarz启发的核典型相关分析加速：理论、算法及其在信息检索中的应用PMID:日期:2026-09-24As a fundamental method in multivariate statistical analysis, Canonical Correlation Analysis (CCA) effectively captures linear relationships between paired variables. However, its linearity constraint limits applicability to complex nonlinear dependencies. Kernel CCA (KCCA) addresses this limitation by enabling nonlinear relationship modeling and high-dimensional feature selection. Existing KCCA methods face significant computational bottlenecks when processing large-scale data, primarily due to their prohibitive memory and time complexity. To overcome these constraints, we develop two accelerated KCCA solvers, termed KCCA-GRK and KCCA-MWRK, by adapting the established Greedy Randomized Kaczmarz (GRK) and Maximal Weighted Residual Kaczmarz (MWRK) row-selection principles to the SVD-reformulated KCCA system. The contribution is KCCA-specific: because the reformulated coefficient matrices have orthonormal rows, norm-proportional RK sampling becomes uniform; KCCA-GRK and KCCA-MWRK restore informative row selection through residual-driven randomized and deterministic rules, respectively. We further establish explicit convergence bounds and sufficient iteration estimates for the proposed algorithms. Experiments on one synthetic dataset and four real-world datasets demonstrate that the proposed methods preserve retrieval performance while substantially reducing iteration counts and runtime.
-
9. The role of grammar in transformer encoders: Does it really matter for language understanding process?
9. 语法在变压器编码器中的作用：它对语言理解过程真的重要吗？PMID:日期:2026-09-24Transformer-based methods achieve strong performance on language understanding tasks, yet the role of grammar in encoder representations remains insufficiently explained. While decoder-only models implicitly learn language structure through next-token prediction, encoder-only models still dominate many classification and matching tasks, where grammatical relations can affect the decision boundary. We therefore ask two research questions: (1) how are grammatical structures represented within transformer-based encoders, and (2) how can grammatical regulation improve language understanding in these encoders? We first design an interpretable analysis showing that grammar is weakly present in standard encoders but is not explicitly organised without regulation. We then propose a Semantics-Grammar Coupled Space (SGCS), a low-dimensional space in which token positions encode semantic information and spatial relations encode grammatical dependencies. SGCS is trained through sentiment-based spatial regulation and attention-grammar distribution alignment, and it is mapped back to the encoder representation through an invertible mapping network. Experiments on 12 language understanding datasets with BERT and RoBERTa backbones show that SGCS raises syntactic tree similarity from 23.84% to 80.10% and improves downstream performance on most evaluated tasks, with especially consistent gains on sentiment-oriented and grammar-sensitive benchmarks.
-
10. Adaptive pinning security synchronization for multi-layer networks under asynchronous link attacks.
10. 异步链路攻击下多层网络的自适应牵制安全同步PMID:日期:2026-09-24In response to the complex security threats faced by multi-layer networks (MLNs), this paper investigates the security synchronization problems under asynchronous link attacks, where each layer consists of identical nonlinear node dynamics and inter-layer couplings are bidirectional. Unlike existing works that assume isolated or synchronous attack patterns, we construct a unified link attack model in which intra-layer links and inter-layer links are compromised independently by switching signals with unknown bounds, asynchronous switching instants and bounded attack durations. To compensate for the resulting coupling degradation, we develop an adaptive pinning control strategy equipped with an online state-driven update law for each pinned node. Using Lyapunov stability theory, we prove that the proposed strategy can achieve intra-layer security synchronization (ALSS) and inter-layer security synchronization (RLSS), respectively. Sufficient synchronization conditions are derived as scalar inequalities that explicitly depend on the attack duration. Finally, two numerical simulations verify the effectiveness of the proposed method.