检索结果-内蒙古大学图书馆

Defending Adversarial Examples by Negative Correlation Ensemble

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Luo, Wenjian Zhang, Hongwei Kong, Linghao Chen, Zhijian Tang, Ke The School of Computer Science and Technology Harbin Institute of Technology Guangdong Shenzhen518055 China Peng Cheng Laboratory Shenzhen518055 China Guangdong Provincial Key Laboratory of Brain-Inspired Intelligent Computation The School of Computer Science and Engineering Southern University of Science and Technology Guangdong Shenzhen518055 China

The security issues in DNNs, such as adversarial examples, have attracted much attention. Adversarial examples refer to the examples which are capable to induce the DNNs return completely predictions by introducing carefully designed perturbations. Obviously, adversarial examples bring great security risks to the development of deep learning. Recently, Some defense approaches against adversarial examples have been proposed, however, in our opinion, the performance of these approaches are still limited. In this paper, we propose a new ensemble defense approach named the Negative Correlation Ensemble (NCEn), which achieves compelling results by introducing gradient directions and gradient magnitudes of each member in the ensemble negatively correlated and at the same time, reducing the transferability of adversarial examples among them. Extensive experiments have been conducted, and the results demonstrate that NCEn can improve the adversarial robustness of ensembles effectively. Copyright © 2022, The Authors. All rights reserved.

关键词： Deep learning

ESVAE: An Efficient Spiking Variational Autoencoder with Reparameterizable Poisson Spiking Sampling

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Zhan, Qiugang Tao, Ran Xie, Xiurui Liu, Guisong Zhang, Malu Tang, Huajin Yang, Yang The Complex Laboratory of New Finance and Economics School of Computing and Artificial Intelligence Southwestern University of Finance and Economics Chengdu611130 China The Laboratory of Intelligent Collaborative Computing University of Electronic Science and Technology of China Chengdu611731 China The School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu611731 China The College of Computer Science The State Key Laboratory of Brain-Machine Intelligence Zhejiang University Hangzhou310027 China

In recent years, studies on image generation models of spiking neural networks (SNNs) have gained the attention of many researchers. Variational autoencoders (VAEs), as one of the most popular image generation models, have attracted a lot of work exploring their SNN implementation. Due to the constrained binary representation in SNNs, existing SNN VAE methods implicitly construct the latent space by an elaborated autoregressive network and use the network outputs as the sampling variables. However, this unspecified implicit representation of the latent space will increase the difficulty of generating high-quality images and introduce additional network parameters. In this paper, we propose an efficient spiking variational autoencoder (ESVAE) that constructs an interpretable latent space distribution and designs a reparameterizable spiking sampling method. Specifically, we construct the prior and posterior of the latent space as a Poisson distribution using the firing rate of the spiking neurons. Subsequently, we propose a reparameterizable Poisson spiking sampling method, which is free from the additional network. Comprehensive experiments have been conducted, and the experimental results show that the proposed ESVAE outperforms previous SNN VAE methods in reconstructed & generated image quality. In addition, experiments demonstrate that the encoder of ESVAE can retain the original image information more efficiently and is more robust. The source code is available at https://***/QgZhan/ESVAE. Copyright © 2023, The Authors. All rights reserved.

关键词： Poisson distribution

DANI-Net: Uncalibrated Photometric Stereo by Differentiable Shadow Handling, Anisotropic Reflectance Modeling, and Neural Inverse Rendering

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Li, Zongrui Zheng, Qian Shi, Boxin Pan, Gang Jiang, Xudong School of Electrical and Electronic Engineering Nanyang Technological University Singapore The State Key Lab of Brain-Machine Intelligence Zhejiang University Hangzhou China College of Computer Science and Technology Zhejiang University Hangzhou China National Key Laboratory for Multimedia Information Processing School of Computer Science Peking University Beijing China National Engineering Research Center of Visual Technology School of Computer Science Peking University Beijing China

关键词： Reflection

PolSAM: Polarimetric Scattering Mechanism Informed Segment Anything Model

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Wang, Yuqing Huang, Zhongling Yang, Shuxin Tang, Hao Qiu, Xiaolan Han, Junwei Zhang, Dingwen Brain and Artificial Intelligence Laboratory School of Automation Northwestern Polytechnical University Xi’an710072 China the National Key Laboratory for Multimedia Information Processing School of Computer Science Peking University Beijing100871 China the National Key Laboratory of Microwave Imaging Technology Chinese Academy of Sciences Beijing100090 China the Aerospace Information Research Institute Chinese Academy of Sciences Beijing100094 China

PolSAR data presents unique challenges due to its rich and complex characteristics. Existing data representations, such as complex-valued data, polarimetric features, and amplitude images, are widely used. However, these formats often face issues related to usability, interpretability, and data integrity. While most feature extraction networks for PolSAR attempt to address these issues, they are typically small in size, which limits their ability to effectively extract features. To overcome these limitations, we introduce the large, powerful Segment Anything Model (SAM), which excels in feature extraction and prompt-based segmentation. However, SAM’s application to PolSAR is hindered by modality differences and limited integration of domain-specific knowledge. To address these challenges, we propose the Polarimetric Scattering Mechanism-Informed SAM (PolSAM), which incorporates physical scattering characteristics and a novel prompt generation strategy to enhance segmentation performance with high data efficiency. Our approach includes a new data processing pipeline that utilizes polarimetric decomposition and semantic correlations to generate Microwave Vision Data (MVD) products, which are lightweight, physically interpretable, and information-dense. We extend the basic SAM architecture with two key contributions: The Feature-Level Fusion Prompt (FFP) module merges visual tokens from the pseudo-colored SAR image and its associated MVD, enriching them with supplementary information. When combined with a dedicated adapter, it addresses modality incompatibility in the frozen SAM encoder. Additionally, we propose the Semantic-Level Fusion Prompt (SFP) module, a progressive mechanism that leverages semantic information within MVD to generate sparse and dense prompt embeddings, refining segmentation details. Experimental evaluation on the newly constructed PhySAR-Seg datasets shows that PolSAM outperforms existing SAM-based models (without MVD) and other multimodal fusion

关键词： Polarimeters

Evolutionary Multiparty Distance Minimization

学校读者我要写书评

暂无评论

arXiv 2022年

作者： She, Zeneng Luo, Wenjian Lin, Xin Chang, Yatong Shi, Yuhui School of Computer Science and Technology Harbin Institute of Technology Guangdong Shenzhen518055 China School of Computer Science and Technology The University of Science and Technology of China Anhui Hefei230027 China Guangdong Provincial Key Laboratory of Brain-inspired Intelligent Computation Department of Computer Science and Engineering Southern University of Science and Technology Shenzhen518055 China

In the field of evolutionary multiobjective optimization, the decision maker (DM) concerns conflicting objectives. In the real-world applications, there usually exist more than one DM and each DM concerns parts of these objectives. Multiparty multiobjective optimization problems (MPMOPs) are proposed to depict the MOP with multiple decision makers involved, where each party concerns about certain some objectives of all. However, in the evolutionary computation field, there is not much attention paid on MPMOPs. This paper constructs a series of MPMOPs based on distance minimization problems (DMPs), whose Pareto optimal solutions can be vividly visualized. To address MPMOPs, the new proposed algorithm OptMPNDS3 uses the multiparty initializing method to initialize the population and takes JADE2 operator to generate the offsprings. OptMPNDS3 is compared with OptAll, OptMPNDS and OptMPNDS2 on the problem suite. The result shows that OptMPNDS3 is strongly comparable to other algorithms. Copyright © 2022, The Authors. All rights reserved.

关键词： Multiobjective optimization

DMNER: Biomedical Named Entity Recognition by Detection and Matching

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Bian, Junyi Jiang, Rongze Zhai, Weiqi Huang, Tianyang Zhou, Hong Zhu, Shanfeng School of Computer Science Fudan University Shanghai200433 China Atypon Systems LLC United Kingdom Institute of Science and Technology for Brain-Inspired Intelligence Fudan University China Key Laboratory of Computational Neuroscience and Brain-Inspired Intelligence Fudan University Ministry of Education Shanghai200433 China MOE Frontiers Center for Brain Science Fudan University Shanghai200433 China Zhangjiang Fudan International Innovation Center Shanghai200433 China Shanghai Key Lab of Intelligent Information Processing Fudan University Shanghai200433 China

Biomedical named entity recognition (BNER) serves as the foundation for numerous biomedical text mining tasks. Unlike general NER, BNER require a comprehensive grasp of the domain, and incorporating external knowledge beyond training data poses a significant challenge. In this study, we propose a novel BNER framework called DMNER. By leveraging existing entity representation models SAP-BERT, we tackle BNER as a two-step process: entity boundary detection and biomedical entity matching. DMNER exhibits applicability across multiple NER scenarios: 1) In supervised NER, we observe that DMNER effectively rectifies the output of baseline NER models, thereby further enhancing performance. 2) In distantly supervised NER, combining MRC and AutoNER as span boundary detectors enables DMNER to achieve satisfactory results. 3) For training NER by merging multiple datasets, we adopt a framework similar to DS-NER but additionally leverage ChatGPT to obtain high-quality phrases in the training. Through extensive experiments conducted on 10 benchmark datasets, we demonstrate the versatility and effectiveness of DMNER. Copyright © 2023, The Authors. All rights reserved.

关键词： Character recognition

Extraction and analysis of abnormal EEG features in children with amblyopia

学校读者我要写书评

暂无评论

Clinical Neurophysiology 2025年 175卷 2110765页

作者： Xiaoke Niu Jinxiong Zhang Yanyan Peng Ying Kong Yadong Li Yonghao Han Li Shi Guangying Zheng Henan Key Laboratory of Brain Science and Brain-Computer Interface Technology School of Electrical and Information Engineering Zhengzhou University Zhengzhou 450001 China Department of Ophthalmology the First Affiliated Hospital Zhengzhou University Zhengzhou 450052 China Department of Automation Tsinghua University Beijing 100084 China

Objective Early and accurate diagnosis of amblyopia is crucial for the healthy development of children. Existing clinical diagnostic methods rely on patient cooperation, which can easily lead to misdiagnosis. The commonly used features derived from visual evoked potentials (VEP) only provided limited information for characterizing the whole brain, highlighting the need for integrating additional data sources, such as brain network metrics, to achieve a more comprehensive understanding. Methods We extracted 488 features from 64-channel EEG data recorded from thirty amblyopic children. The features mainly derived from a weighted functional brain network based on coherence across different frequency bands. Feature selection and linear classification techniques were employed to assess their effectiveness in distinguishing amblyopia from normal children. Results Abnormal EEG features were distributed not only in the occipital lobe but also in non-visual regions, with a higher prevalence in the alpha and beta bands. Their decoding performance surpassed traditional VEP features, and their combination achieved the highest accuracy (89.00%). Moreover, features beyond the occipital lobe exhibited limited decoding performance when considered individually, yet they still have an obvious contribution. Conclusions The study identified novel abnormal EEG features associated with amblyopia and demonstrated the potential of multi-channel EEG recordings to assist in the diagnosis of amblyopia. Significance The study suggests amblyopia may impair more abilities beyond visual cognition and further provides objective biomarkers for diagnosing amblyopia, which is essential for effective treatment.

关键词： Amblyopic children Decoding EEG Functional connection Weighted brain networks

DISD-Net: A Dynamic Interactive Network with Self-distillation for Cross-subject Multi-modal Emotion Recognition

学校读者我要写书评

暂无评论

IEEE Transactions on Multimedia 2025年

作者： Cheng, Cheng Liu, Wenzhe Wang, Xinying Feng, Lin Jia, Ziyu Liaoning Normal University Research Center of Brain and Cognitive Neuroscience Dalian116029 China Anhui University Anhui Provincial Key Laboratory of Multimodal Cognitive Computation Hefei230601 China Key Laboratory of Biomedical Engineering of Hainan Province Haikou570228 China Huzhou University School of Information Engineering Huzhou313000 China Dalian University of Technology School of Computer Science and Technology Dalian116081 China Institute of Automation Chinese Academy of Sciences Brainnetome Center Beijing100190 China

Multi-modal Emotion Recognition (MER) has demonstrated competitive performance in affective computing, owing to synthesizing information from diverse modalities. However, many existing approaches still face unresolved challenges, such as: (i) how to learn compact yet representative features from multi-modal data simultaneously, and (ii) how to address differences among subjects and enhance the generalization of the emotion recognition model, given the diverse nature of individual biological signals. To this end, we propose a Dynamic Interactive Network with Self-Distillation (DISD-Net) for cross-subject MER. The DISD-Net incorporates a dynamin interactive module to capture the intra- and inter-modal interactions from multi-modal data. Additionally, to enhance compactness in modal representations, we leverage the soft labels generated by the DISD-Net model as supplemental training guidance. This involves incorporating self-distillation, aiming to transfer the knowledge that the DISD-Net model contains hard and soft labels to each modality. Finally, domain adaptation (DA) is seamlessly integrated into the dynamic interactive and self-distillation components, forming a unified framework to extract subject-invariant multi-modal emotional features. Experimental results indicate that the proposed model achieves a mean accuracy of 75.00% with a standard deviation of 7.68% for the DEAP dataset and a mean accuracy of 65.65% with a standard deviation of 5.08% for the SEED-IV dataset. © 1999-2012 IEEE.

关键词： domain adaptation (DA) dynamic interactive Multi-modal Emotion Recognition (MER) self-distillation

SLC2-SLAM: Semantic-guided Loop Closure using Shared Latent Code for NeRF SLAM

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Ming, Yuhang Ma, Di Dai, Weichen Yang, Han Fan, Rui Zhang, Guofeng Kong, Wanzeng School of Computer Science Key Laboratory of Brain Machine Collaborative Intelligence of Zhejiang Province Hangzhou Dianzi University Hangzhou310018 China College of Electronics & Information Engineering Shanghai Institute of Intelligent Science and Technology Shanghai Research Institute for Intelligent Autonomous Systems The State Key Laboratory of Intelligent Autonomous Systems Frontiers Science Center for Intelligent Autonomous Systems Tongji University Shanghai201804 China State Key Lab of CAD&CG Zhejiang University Hangzhou310058 China

Targeting the notorious cumulative drift errors in NeRF SLAM, we propose a Semantic-guided Loop Closure using Shared Latent Code, dubbed SLC2-SLAM. We argue that latent codes stored in many NeRF SLAM systems are not fully exploited, as they are only used for better reconstruction. In this paper, we propose a simple yet effective way to detect potential loops using the same latent codes as local features. To further improve the loop detection performance, we use the semantic information, which are also decoded from the same latent codes to guide the aggregation of local features. Finally, with the potential loops detected, we close them with a graph optimization followed by bundle adjustment to refine both the estimated poses and the reconstructed scene. To evaluate the performance of our SLC2-SLAM, we conduct extensive experiments on Replica and ScanNet datasets. Our proposed semantic-guided loop closure significantly outperforms the pre-trained NetVLAD and ORB combined with Bag-of-Words, which are used in all the other NeRF SLAM with loop closure. As a result, our SLC2-SLAM also demonstrated better tracking and reconstruction performance, especially in larger scenes with more loops, like ScanNet. Copyright © 2025, The Authors. All rights reserved.

关键词： SLAM robotics