Pub Date : 2024-07-15DOI: 10.1109/TAI.2024.3428520
Han Wu;Sareh Rowlands;Johan Wahlström
Object detection systems using deep learning models have become increasingly popular in robotics thanks to the rising power of central processing units (CPUs) and graphics processing units (GPUs) in embedded systems. However, these models are susceptible to adversarial attacks. While some attacks are limited by strict assumptions on access to the detection system, we propose a novel hardware attack inspired by Man-in-the-Middle attacks in cryptography. This attack generates a universal adversarial perturbations (UAPs) and injects the perturbation between the universal serial bus (USB) camera and the detection system via a hardware attack. Besides, prior research is misled by an evaluation metric that measures the model accuracy rather than the attack performance. In combination with our proposed evaluation metrics, we significantly increased the strength of adversarial perturbations. These findings raise serious concerns for applications of deep learning models in safety-critical systems, such as autonomous driving.
{"title":"A Human-in-the-Middle Attack Against Object Detection Systems","authors":"Han Wu;Sareh Rowlands;Johan Wahlström","doi":"10.1109/TAI.2024.3428520","DOIUrl":"https://doi.org/10.1109/TAI.2024.3428520","url":null,"abstract":"Object detection systems using deep learning models have become increasingly popular in robotics thanks to the rising power of central processing units (CPUs) and graphics processing units (GPUs) in embedded systems. However, these models are susceptible to adversarial attacks. While some attacks are limited by strict assumptions on access to the detection system, we propose a novel hardware attack inspired by Man-in-the-Middle attacks in cryptography. This attack generates a universal adversarial perturbations (UAPs) and injects the perturbation between the universal serial bus (USB) camera and the detection system via a hardware attack. Besides, prior research is misled by an evaluation metric that measures the model accuracy rather than the attack performance. In combination with our proposed evaluation metrics, we significantly increased the strength of adversarial perturbations. These findings raise serious concerns for applications of deep learning models in safety-critical systems, such as autonomous driving.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142443113","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-28DOI: 10.1109/TAI.2024.3418376
Minjie Liu;Hongjian Wang;Kuk-Jin Yoon;Lin Wang
Event cameras detect the intensity changes and produce asynchronous events with high dynamic range and no motion blur. Recently, several attempts have been made to superresolve the intensity images guided by events. However, these methods directly fuse the event and image features without distinguishing the modality difference and achieve image superresolution (SR) in multiple steps, leading to error-prone image SR results. Also, they lack quantitative evaluation of real-world data. In this article, we present an end-to-end framework, called event-guided image (EGI)-SR to narrow the modality gap and subtly integrate the event and RGB modality features for effective image SR. Specifically, EGI-SR employs three crossmodality encoders (CME) to learn modality-specific and modality-shared features from the stacked events and the intensity image, respectively. As such, EGI-SR can better mitigate the negative impact of modality varieties and reduce the difference in the feature space between the events and the intensity image. Subsequently, a transformer-based decoder is deployed to reconstruct the SR image. Moreover, we collect a real-world dataset, with temporally and spatially aligned events and color image pairs. We conduct extensive experiments on the synthetic and real-world datasets, showing EGI-SR favorably surpassing the existing methods by a large margin.
{"title":"Disentangled Cross-modal Fusion for Event-Guided Image Super-resolution","authors":"Minjie Liu;Hongjian Wang;Kuk-Jin Yoon;Lin Wang","doi":"10.1109/TAI.2024.3418376","DOIUrl":"https://doi.org/10.1109/TAI.2024.3418376","url":null,"abstract":"Event cameras detect the intensity changes and produce asynchronous events with high dynamic range and no motion blur. Recently, several attempts have been made to superresolve the intensity images guided by events. However, these methods directly fuse the event and image features without distinguishing the modality difference and achieve image superresolution (SR) in multiple steps, leading to error-prone image SR results. Also, they lack quantitative evaluation of real-world data. In this article, we present an \u0000<italic>end-to-end</i>\u0000 framework, called \u0000<italic>event-guided image (EGI)-SR</i>\u0000 to narrow the modality gap and subtly integrate the event and RGB modality features for effective image SR. Specifically, EGI-SR employs three crossmodality encoders (CME) to learn modality-specific and modality-shared features from the stacked events and the intensity image, respectively. As such, EGI-SR can better mitigate the negative impact of modality varieties and reduce the difference in the feature space between the events and the intensity image. Subsequently, a transformer-based decoder is deployed to reconstruct the SR image. Moreover, we collect a real-world dataset, with temporally and spatially aligned events and color image pairs. We conduct extensive experiments on the synthetic and real-world datasets, showing EGI-SR favorably surpassing the existing methods by a large margin.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142443093","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-27DOI: 10.1109/TAI.2024.3420261
Yuxiang Zhang;Xiaoling Liang;Dongyu Li;Shuzhi Sam Ge;Bingzhao Gao;Hong Chen;Tong Heng Lee
The approach of (fixed-) time-synchronized control (FTSC) aims at attaining the outcome where all the system state-variables converge to the origin simultaneously/synchronously. This type of outcome can be the highly essential performance desired in various real-world high-precision control applications. Toward this objective, this article proposes and investigates the development of a time-synchronized reinforcement learning algorithm (TSRL) applicable to a particular class of first- and second-order affine nonlinear systems. The approach developed here appropriately incorporates the norm-normalized sign function into the optimal system control design, leveraging on the special properties of this norm-normalized sign function in attaining time-synchronized stability and control. Concurrently, the actor–critic framework in reinforcement learning (RL) is invoked, and the dual quantities of system control and gradient term of the cost function are decomposed with appropriate time-synchronized control items and unknown actor/critic part and to be learned independently. By additionally employing the adaptive dynamic programming technique, the solution of the Hamilton–Jacobi–Bellman equation is iteratively approximated under this actor–critic framework. As an outcome, the proposed TSRL method optimizes the system control while attaining the notable time-synchronized convergence property. The performance and effectiveness of the proposed method are demonstrated to be effectively applicable via detailed numerical studies and on an autonomous vehicle nonlinear system motion control problem.
{"title":"Reinforcement Learning-Based Time-Synchronized Optimized Control for Affine Systems","authors":"Yuxiang Zhang;Xiaoling Liang;Dongyu Li;Shuzhi Sam Ge;Bingzhao Gao;Hong Chen;Tong Heng Lee","doi":"10.1109/TAI.2024.3420261","DOIUrl":"https://doi.org/10.1109/TAI.2024.3420261","url":null,"abstract":"The approach of (fixed-) time-synchronized control (FTSC) aims at attaining the outcome where all the system state-variables converge to the origin simultaneously/synchronously. This type of outcome can be the highly essential performance desired in various real-world high-precision control applications. Toward this objective, this article proposes and investigates the development of a time-synchronized reinforcement learning algorithm (TSRL) applicable to a particular class of first- and second-order affine nonlinear systems. The approach developed here appropriately incorporates the norm-normalized sign function into the optimal system control design, leveraging on the special properties of this norm-normalized sign function in attaining time-synchronized stability and control. Concurrently, the actor–critic framework in reinforcement learning (RL) is invoked, and the dual quantities of system control and gradient term of the cost function are decomposed with appropriate time-synchronized control items and unknown actor/critic part and to be learned independently. By additionally employing the adaptive dynamic programming technique, the solution of the Hamilton–Jacobi–Bellman equation is iteratively approximated under this actor–critic framework. As an outcome, the proposed TSRL method optimizes the system control while attaining the notable time-synchronized convergence property. The performance and effectiveness of the proposed method are demonstrated to be effectively applicable via detailed numerical studies and on an autonomous vehicle nonlinear system motion control problem.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142443107","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-27DOI: 10.1109/TAI.2024.3420262
Hang Yu;Weixu Liu;Nengjun Zhu;Pengbo Li;Xiangfeng Luo
With the development of the e-commerce platform, more and more reviews of its various formats continue to appear. Reviews help people buy the right item faster, and instead, spam reviews reduce the user experience. To be able to detect spam reviews, statistical machine learning-based methods were commonly used in the past, but these approaches ignored the correlation between reviews. With the development of the graph fraud detection model, people have started to graph model the review data. However, typical graph fraud detection models still have problems with interpretability. Therefore, we propose here an interpretable graph fraud detection model for spam reviews, which is also named IN-GFD. As for the interpretability issue, we leveraged the relationship against the predicted score and whether a review is spam or not to build a loss function on top of the feature-embedding matrix, and introduced a scoring difference threshold mechanism, which can allow our model to have antehoc interpretability. In addition, to address class imbalance issues, IN-GFD utilizes the oversampling of the spam nodes to balance them with normal nodes and introduces an edge-loss function to learn new edge relationships. After extensive experiments, our method proves to be better than other state-of-the-arts (SOTA) models in terms of fraud detection and offers the benefit of interpretability. Finally, our study combines detection models with antehoc interpretability, offering a promising direction in review detection. Our approach has wide applicability, detecting spam reviews in datasets with user reviews and providing reasonable interpretations.
{"title":"IN-GFD: An Interpretable Graph Fraud Detection Model for Spam Reviews","authors":"Hang Yu;Weixu Liu;Nengjun Zhu;Pengbo Li;Xiangfeng Luo","doi":"10.1109/TAI.2024.3420262","DOIUrl":"https://doi.org/10.1109/TAI.2024.3420262","url":null,"abstract":"With the development of the e-commerce platform, more and more reviews of its various formats continue to appear. Reviews help people buy the right item faster, and instead, spam reviews reduce the user experience. To be able to detect spam reviews, statistical machine learning-based methods were commonly used in the past, but these approaches ignored the correlation between reviews. With the development of the graph fraud detection model, people have started to graph model the review data. However, typical graph fraud detection models still have problems with interpretability. Therefore, we propose here an interpretable graph fraud detection model for spam reviews, which is also named IN-GFD. As for the interpretability issue, we leveraged the relationship against the predicted score and whether a review is spam or not to build a loss function on top of the feature-embedding matrix, and introduced a scoring difference threshold mechanism, which can allow our model to have antehoc interpretability. In addition, to address class imbalance issues, IN-GFD utilizes the oversampling of the spam nodes to balance them with normal nodes and introduces an edge-loss function to learn new edge relationships. After extensive experiments, our method proves to be better than other state-of-the-arts (SOTA) models in terms of fraud detection and offers the benefit of interpretability. Finally, our study combines detection models with antehoc interpretability, offering a promising direction in review detection. Our approach has wide applicability, detecting spam reviews in datasets with user reviews and providing reasonable interpretations.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142443127","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-26DOI: 10.1109/TAI.2024.3418938
Jie Lan;Hao Wang;Yan-Jun Liu;Shaocheng Tong
An adaptive intelligent resilient distributed output bipartite time-varying formation protocol is proposed for a class of second-order uncertain nonlinear multiagent systems (MASs) with unknown attacks. Actuators and sensors are both vulnerable to unknown false data injection (FDI) attacks, and the proposed protocol does not require the removal of misbehaving agents or strong network connectivity restrictions. However, existing research methods are mainly limited to studying the complete cooperative relationship and attacks only on actuators or sensors. Network interactions are based on directed signed topologies, reflecting cooperation and competition between agents, and the corresponding adjacency matrix is no longer nonnegative, making traditional consensus controls strategy inapplicable and analyzed by gauge transformation matrix. Due to the uncertain nonlinear dynamics with unmeasurable states, unknown attacks would jeopardize the synchronization of bipartite formation control and even deteriorate entire systems. To address this issue, a security state estimator and adaptive intelligent state reconstruction technique are adopted. It not only can estimate and mitigate malicious unknown FDI attacks on both actuators and sensors simultaneously but also achieve uniform ultimate boundedness (UUB) for observer errors and prescribed time-varying bipartite group consistency formation performance. In particular, the proposed method overcomes the restriction that the dynamics must be linear or general Lipschitz-type nonlinear conditions. Finally, employing Riccati equation and linear matrix inequality, the theoretical method is validly proved by constructing proper Lyapunov through transformation matrix. The results of digital simulation can be effectively demonstrated.
{"title":"Adaptive Intelligent Resilient Bipartite Formation Control for Nonlinear Multiagent Systems With False Data Injection Attacks on Actuators and Sensors","authors":"Jie Lan;Hao Wang;Yan-Jun Liu;Shaocheng Tong","doi":"10.1109/TAI.2024.3418938","DOIUrl":"https://doi.org/10.1109/TAI.2024.3418938","url":null,"abstract":"An adaptive intelligent resilient distributed output bipartite time-varying formation protocol is proposed for a class of second-order uncertain nonlinear multiagent systems (MASs) with unknown attacks. Actuators and sensors are both vulnerable to unknown false data injection (FDI) attacks, and the proposed protocol does not require the removal of misbehaving agents or strong network connectivity restrictions. However, existing research methods are mainly limited to studying the complete cooperative relationship and attacks only on actuators or sensors. Network interactions are based on directed signed topologies, reflecting cooperation and competition between agents, and the corresponding adjacency matrix is no longer nonnegative, making traditional consensus controls strategy inapplicable and analyzed by gauge transformation matrix. Due to the uncertain nonlinear dynamics with unmeasurable states, unknown attacks would jeopardize the synchronization of bipartite formation control and even deteriorate entire systems. To address this issue, a security state estimator and adaptive intelligent state reconstruction technique are adopted. It not only can estimate and mitigate malicious unknown FDI attacks on both actuators and sensors simultaneously but also achieve uniform ultimate boundedness (UUB) for observer errors and prescribed time-varying bipartite group consistency formation performance. In particular, the proposed method overcomes the restriction that the dynamics must be linear or general Lipschitz-type nonlinear conditions. Finally, employing Riccati equation and linear matrix inequality, the theoretical method is validly proved by constructing proper Lyapunov through transformation matrix. The results of digital simulation can be effectively demonstrated.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142443132","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-25DOI: 10.1109/TAI.2024.3387583
Ran Cheng;Hugo Jair Escalante;Wei-Wei Tu;Jan N. Van Rijn;Shuo Wang;Yun Yang
The five papers in this special section address different aspects of automated machine learning (AutoML) from fundamental algorithms to real-world applications. Developing high-performance machine learning models is a difficult task that usually requires expertise from data scientists and knowledge from domain experts. To make machine learning more accessible and ease the labor-intensive trial-and-error process of searching for the most appropriate machine learning algorithm and the optimal hyperparameter setting, AutoML was developed and has become a rapidly growing area in recent years. AutoML aims at automation and efficiency of the machine learning process across domains and applications. Nowadays, data is commonly collected over time and susceptible to changes, such as in Internet-of-Things (IoT) systems, mobile phone applications and healthcare data analysis. It poses new challenges to the traditional AutoML with the assumption of data stationarity. Interesting research questions arise around whether, when and how to effectively and efficiently deal with non-stationary data in AutoML.
{"title":"Guest Editorial: AutoML for Nonstationary Data","authors":"Ran Cheng;Hugo Jair Escalante;Wei-Wei Tu;Jan N. Van Rijn;Shuo Wang;Yun Yang","doi":"10.1109/TAI.2024.3387583","DOIUrl":"https://doi.org/10.1109/TAI.2024.3387583","url":null,"abstract":"The five papers in this special section address different aspects of automated machine learning (AutoML) from fundamental algorithms to real-world applications. Developing high-performance machine learning models is a difficult task that usually requires expertise from data scientists and knowledge from domain experts. To make machine learning more accessible and ease the labor-intensive trial-and-error process of searching for the most appropriate machine learning algorithm and the optimal hyperparameter setting, AutoML was developed and has become a rapidly growing area in recent years. AutoML aims at automation and efficiency of the machine learning process across domains and applications. Nowadays, data is commonly collected over time and susceptible to changes, such as in Internet-of-Things (IoT) systems, mobile phone applications and healthcare data analysis. It poses new challenges to the traditional AutoML with the assumption of data stationarity. Interesting research questions arise around whether, when and how to effectively and efficiently deal with non-stationary data in AutoML.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10571781","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141474817","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-20DOI: 10.1109/TAI.2024.3417389
Dipanshu Naware;Arghya Mitra
The need for data-driven technologies such as artificial intelligence (AI), machine learning (ML), and deep learning (DL) in various sectors has been soaring for over a decade. The amount of data released by the smart grid itself has been enormous, making these cutting-edge technologies highly efficient and reliable. This article proposes an orderly review of data-driven technology applications for smart residential households. It underpins the importance of forecasting studies with demand-side management (DSM)-aided tools such as demand response (DR), over a secure energy transaction platform. For the publications reviewed, the outcomes suggest the urgent need for household-level forecasting as it accounts for only 21% of the publications reviewed while DL dominates the forecasting studies (57%) with scope towards its hybridization with decomposition techniques. Similarly, the DSM/DR domain needs to be actively implemented at the retail level over a secure network. The outcomes suggest that baseline prediction (4.76%) and self-learning DR (19%) are crucial but the least focused issues, hence AI/ML/DL could be the solutions. Likewise, scalability (24.3%) turns out to be the major issue for assessing the security of the utility grid. However, deep reinforcement learning (DRL) could be a suitable tool as it is adaptive, independent of the system dynamics, and works best in a model-free dynamic environment. The overall findings suggest that the smart household community is the least focused entity and needs prompt attention to address the associated challenges. Additionally, several distinct insights such as dataset features, model parameters, performance metrics, customer-centricity, customer diversity, and mitigation are mapped with applications. Besides, this article points out various shortcomings and tries to postulate probable solutions to the best of capacity.
{"title":"Data-Driven Technology Applications in Planning, Demand-Side Management, and Cybersecurity for Smart Household Community","authors":"Dipanshu Naware;Arghya Mitra","doi":"10.1109/TAI.2024.3417389","DOIUrl":"https://doi.org/10.1109/TAI.2024.3417389","url":null,"abstract":"The need for data-driven technologies such as artificial intelligence (AI), machine learning (ML), and deep learning (DL) in various sectors has been soaring for over a decade. The amount of data released by the smart grid itself has been enormous, making these cutting-edge technologies highly efficient and reliable. This article proposes an orderly review of data-driven technology applications for smart residential households. It underpins the importance of forecasting studies with demand-side management (DSM)-aided tools such as demand response (DR), over a secure energy transaction platform. For the publications reviewed, the outcomes suggest the urgent need for household-level forecasting as it accounts for only 21% of the publications reviewed while DL dominates the forecasting studies (57%) with scope towards its hybridization with decomposition techniques. Similarly, the DSM/DR domain needs to be actively implemented at the retail level over a secure network. The outcomes suggest that baseline prediction (4.76%) and self-learning DR (19%) are crucial but the least focused issues, hence AI/ML/DL could be the solutions. Likewise, scalability (24.3%) turns out to be the major issue for assessing the security of the utility grid. However, deep reinforcement learning (DRL) could be a suitable tool as it is adaptive, independent of the system dynamics, and works best in a model-free dynamic environment. The overall findings suggest that the smart household community is the least focused entity and needs prompt attention to address the associated challenges. Additionally, several distinct insights such as dataset features, model parameters, performance metrics, customer-centricity, customer diversity, and mitigation are mapped with applications. Besides, this article points out various shortcomings and tries to postulate probable solutions to the best of capacity.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142442994","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-18DOI: 10.1109/TAI.2024.3416420
Junren Pan;Qiankun Zuo;Bingchuan Wang;C.L. Philip Chen;Baiying Lei;Shuqiang Wang
One of the main reasons for Alzheimer's disease (AD) is the disorder of some neural circuits. Existing methods for AD prediction have achieved great success, however, detecting abnormal neural circuits from the perspective of brain networks is still a big challenge. In this work, a novel decoupling generative adversarial network (DecGAN) is proposed to detect abnormal neural circuits for AD. Concretely, a decoupling module is designed to decompose a brain network into two parts: one part is composed of a few sparse graphs that represent the neural circuits largely determining the development of AD; the other part is a supplement graph, whose influence on AD can be ignored. Furthermore, the adversarial strategy is utilized to guide the decoupling module to extract the feature more related to AD. Meanwhile, by encoding the detected neural circuits to hypergraph data, an analytic module associated with the hyperedge neurons algorithm is designed to identify the neural circuits. More importantly, a novel sparse capacity loss based on the spatial-spectral hypergraph similarity is developed to minimize the intrinsic topological distribution of neural circuits, which can significantly improve the accuracy and robustness of the proposed model. Experimental results demonstrate that the proposed model can effectively detect the abnormal neural circuits at different stages of AD, which is helpful for pathological study and early treatment.
阿尔茨海默病（AD）的主要原因之一是某些神经回路的紊乱。现有的阿尔茨海默病预测方法取得了巨大成功，但从大脑网络的角度检测异常神经回路仍是一大挑战。本研究提出了一种新型解耦生成对抗网络（DecGAN）来检测 AD 的异常神经回路。具体来说，设计了一个解耦模块，将大脑网络分解为两部分：一部分由一些稀疏图组成，这些稀疏图代表了在很大程度上决定注意力缺失症发展的神经回路；另一部分是补充图，这些补充图对注意力缺失症的影响可以忽略不计。此外，还利用对抗策略引导解耦模块提取与注意力缺失症更相关的特征。同时，通过将检测到的神经回路编码为超图数据，设计了一个与超edge 神经元算法相关的分析模块来识别神经回路。更重要的是，基于空间-光谱超图相似性开发了一种新的稀疏容量损失，以最小化神经回路的内在拓扑分布，从而显著提高了所提模型的准确性和鲁棒性。实验结果表明，所提出的模型能有效地检测出AD不同阶段的异常神经回路，有助于病理研究和早期治疗。
{"title":"DecGAN: Decoupling Generative Adversarial Network for Detecting Abnormal Neural Circuits in Alzheimer's Disease","authors":"Junren Pan;Qiankun Zuo;Bingchuan Wang;C.L. Philip Chen;Baiying Lei;Shuqiang Wang","doi":"10.1109/TAI.2024.3416420","DOIUrl":"https://doi.org/10.1109/TAI.2024.3416420","url":null,"abstract":"One of the main reasons for Alzheimer's disease (AD) is the disorder of some neural circuits. Existing methods for AD prediction have achieved great success, however, detecting abnormal neural circuits from the perspective of brain networks is still a big challenge. In this work, a novel decoupling generative adversarial network (DecGAN) is proposed to detect abnormal neural circuits for AD. Concretely, a decoupling module is designed to decompose a brain network into two parts: one part is composed of a few sparse graphs that represent the neural circuits largely determining the development of AD; the other part is a supplement graph, whose influence on AD can be ignored. Furthermore, the adversarial strategy is utilized to guide the decoupling module to extract the feature more related to AD. Meanwhile, by encoding the detected neural circuits to hypergraph data, an analytic module associated with the hyperedge neurons algorithm is designed to identify the neural circuits. More importantly, a novel sparse capacity loss based on the spatial-spectral hypergraph similarity is developed to minimize the intrinsic topological distribution of neural circuits, which can significantly improve the accuracy and robustness of the proposed model. Experimental results demonstrate that the proposed model can effectively detect the abnormal neural circuits at different stages of AD, which is helpful for pathological study and early treatment.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142443087","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-18DOI: 10.1109/TAI.2024.3415550
Youness Boutyour;Abdellah Idrissi
Multiagent reinforcement learning (MARL) poses unique challenges in real-world applications, demanding the adaptation of reinforcement learning principles to scenarios where agents interact in dynamically changing environments. This article presents a novel approach, “decentralized policy with attention” (ADPA), designed to address these challenges in large-scale multiagent environments. ADPA leverages an attention mechanism to dynamically select relevant information for estimating critics while training decentralized policies. This enables effective and scalable learning, supporting both cooperative and competitive settings, and scenarios with nonglobal states. In this work, we conduct a comprehensive evaluation of ADPA across a range of multiagent environments, including cooperative treasure collection and rover-tower communication. We compare ADPA with existing centralized training methods and ablated variants to showcase its advantages in terms of scalability, adaptability to various environments, and robustness. Our results demonstrate that ADPA offers a promising solution for addressing the complexities of large-scale MARL, providing the flexibility to handle diverse multiagent scenarios. By combining decentralized policies with attention mechanisms, we contribute to the advancement of MARL techniques, offering a powerful tool for real-world applications in dynamic and interactive multiagent systems.
{"title":"Adaptive Decentralized Policies With Attention for Large-Scale Multiagent Environments","authors":"Youness Boutyour;Abdellah Idrissi","doi":"10.1109/TAI.2024.3415550","DOIUrl":"https://doi.org/10.1109/TAI.2024.3415550","url":null,"abstract":"Multiagent reinforcement learning (MARL) poses unique challenges in real-world applications, demanding the adaptation of reinforcement learning principles to scenarios where agents interact in dynamically changing environments. This article presents a novel approach, “decentralized policy with attention” (ADPA), designed to address these challenges in large-scale multiagent environments. ADPA leverages an attention mechanism to dynamically select relevant information for estimating critics while training decentralized policies. This enables effective and scalable learning, supporting both cooperative and competitive settings, and scenarios with nonglobal states. In this work, we conduct a comprehensive evaluation of ADPA across a range of multiagent environments, including cooperative treasure collection and rover-tower communication. We compare ADPA with existing centralized training methods and ablated variants to showcase its advantages in terms of scalability, adaptability to various environments, and robustness. Our results demonstrate that ADPA offers a promising solution for addressing the complexities of large-scale MARL, providing the flexibility to handle diverse multiagent scenarios. By combining decentralized policies with attention mechanisms, we contribute to the advancement of MARL techniques, offering a powerful tool for real-world applications in dynamic and interactive multiagent systems.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-06-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142442959","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：