Pub Date : 2024-04-16DOI: 10.1109/TAI.2024.3356669
K. P. Suba Subbalakshmi;Wojciech Samek;Xia Ben Hu
This special issue brings together seven articles that address different aspects of explainable and interpretable artificial intelligence (AI). Over the years, machine learning (ML) and AI models have posted strong performance across several tasks. This has sparked interest in deploying these methods in critical applications like health and finance. However, to be deployable in the field, ML and AI models must be trustworthy. Explainable and interpretable AI are two areas of research that have become increasingly important to ensure trustworthiness and hence deployability of advanced AI and ML methods. Interpretable AI are models that obey some domain-specific constraints so that they are better understandable by humans. In essence, they are not black-box models. On the other hand, explainable AI refers to models and methods that are typically used to explain another black-box model.
本特刊汇集了七篇文章，探讨了可解释和可解释人工智能（AI）的不同方面。多年来，机器学习（ML）和人工智能模型在多项任务中表现出色。这激发了人们将这些方法部署到健康和金融等关键应用领域的兴趣。然而，要在该领域部署，ML 和 AI 模型必须值得信赖。可解释人工智能和可解释人工智能是两个日益重要的研究领域，可确保先进人工智能和 ML 方法的可信度和可部署性。可解释的人工智能模型遵从某些特定领域的约束条件，因此更容易被人类理解。从本质上讲，它们不是黑盒模型。另一方面，可解释人工智能指的是通常用于解释另一个黑盒模型的模型和方法。
{"title":"Guest Editorial: New Developments in Explainable and Interpretable Artificial Intelligence","authors":"K. P. Suba Subbalakshmi;Wojciech Samek;Xia Ben Hu","doi":"10.1109/TAI.2024.3356669","DOIUrl":"https://doi.org/10.1109/TAI.2024.3356669","url":null,"abstract":"This special issue brings together seven articles that address different aspects of explainable and interpretable artificial intelligence (AI). Over the years, machine learning (ML) and AI models have posted strong performance across several tasks. This has sparked interest in deploying these methods in critical applications like health and finance. However, to be deployable in the field, ML and AI models must be trustworthy. Explainable and interpretable AI are two areas of research that have become increasingly important to ensure trustworthiness and hence deployability of advanced AI and ML methods. Interpretable AI are models that obey some domain-specific constraints so that they are better understandable by humans. In essence, they are not black-box models. On the other hand, explainable AI refers to models and methods that are typically used to explain another black-box model.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"5 4","pages":"1427-1428"},"PeriodicalIF":0.0,"publicationDate":"2024-04-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10500898","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140559370","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 : 2024-04-16DOI: 10.1109/TAI.2024.3389611
Hezhe Sun;Yufei Wang;Huiwen Yang;Kaixuan Huo;Yuzhe Li
Privacy-aware machine learning paradigms have sparked widespread concern due to their ability to safeguard the local privacy of data owners, preventing the leakage of private information to untrustworthy platforms or malicious third parties. This article focuses on characterizing the interactions between the learner and the data owner within this privacy-aware training process. Here, the data owner hesitates to transmit the original gradient to the learner due to potential cybersecurity issues, such as gradient leakage and membership inference. To address this concern, we propose a Stackelberg game framework that models the training process. In this framework, the data owner's objective is not to maximize the discrepancy between the learner's obtained gradient and the true gradient but rather to ensure that the learner obtains a gradient closely resembling one deliberately designed by the data owner, while the learner's objective is to recover the true gradient as accurately as possible. We derive the optimal encoder and decoder using mismatched cost functions and characterize the equilibrium for specific cases, balancing model accuracy and local privacy. Numerical examples illustrate the main results, and we conclude with expanding discussions to suggest future investigations into reliable countermeasure designs.
{"title":"Strategic Gradient Transmission With Targeted Privacy-Awareness in Model Training: A Stackelberg Game Analysis","authors":"Hezhe Sun;Yufei Wang;Huiwen Yang;Kaixuan Huo;Yuzhe Li","doi":"10.1109/TAI.2024.3389611","DOIUrl":"https://doi.org/10.1109/TAI.2024.3389611","url":null,"abstract":"Privacy-aware machine learning paradigms have sparked widespread concern due to their ability to safeguard the local privacy of data owners, preventing the leakage of private information to untrustworthy platforms or malicious third parties. This article focuses on characterizing the interactions between the learner and the data owner within this privacy-aware training process. Here, the data owner hesitates to transmit the original gradient to the learner due to potential cybersecurity issues, such as gradient leakage and membership inference. To address this concern, we propose a Stackelberg game framework that models the training process. In this framework, the data owner's objective is not to maximize the discrepancy between the learner's obtained gradient and the true gradient but rather to ensure that the learner obtains a gradient closely resembling one deliberately designed by the data owner, while the learner's objective is to recover the true gradient as accurately as possible. We derive the optimal encoder and decoder using mismatched cost functions and characterize the equilibrium for specific cases, balancing model accuracy and local privacy. Numerical examples illustrate the main results, and we conclude with expanding discussions to suggest future investigations into reliable countermeasure designs.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"5 9","pages":"4635-4648"},"PeriodicalIF":0.0,"publicationDate":"2024-04-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142169723","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 : 2024-04-16DOI: 10.1109/TAI.2024.3388389
Mingfu Xue;Xin Wang;Yinghao Wu;Shifeng Ni;Leo Yu Zhang;Yushu Zhang;Weiqiang Liu
Intellectual property (IP) protection for deep neural networks (DNNs) has raised serious concerns in recent years. Most existing works embed watermarks in the DNN model for IP protection, which need to modify the model and do not consider/mention interpretability. In this article, for the first time, we propose an interpretable IP protection method for DNN based on explainable artificial intelligence. Compared with existing works, the proposed method does not modify the DNN model, and the decision of the ownership verification is interpretable. We extract the intrinsic features of the DNN model by using deep Taylor decomposition. Since the intrinsic feature is composed of unique interpretation of the model's decision, the intrinsic feature can be regarded as fingerprint of the model. If the fingerprint of a suspected model is the same as the original model, the suspected model is considered as a pirated model. Experimental results demonstrate that the fingerprints can be successfully used to verify the ownership of the model and the test accuracy of the model is not affected. Furthermore, the proposed method is robust to fine-tuning attack, pruning attack, watermark overwriting attack, and adaptive attack.
{"title":"An Explainable Intellectual Property Protection Method for Deep Neural Networks Based on Intrinsic Features","authors":"Mingfu Xue;Xin Wang;Yinghao Wu;Shifeng Ni;Leo Yu Zhang;Yushu Zhang;Weiqiang Liu","doi":"10.1109/TAI.2024.3388389","DOIUrl":"https://doi.org/10.1109/TAI.2024.3388389","url":null,"abstract":"Intellectual property (IP) protection for deep neural networks (DNNs) has raised serious concerns in recent years. Most existing works embed watermarks in the DNN model for IP protection, which need to modify the model and do not consider/mention interpretability. In this article, for the first time, we propose an interpretable IP protection method for DNN based on explainable artificial intelligence. Compared with existing works, the proposed method does not modify the DNN model, and the decision of the ownership verification is interpretable. We extract the intrinsic features of the DNN model by using deep Taylor decomposition. Since the intrinsic feature is composed of unique interpretation of the model's decision, the intrinsic feature can be regarded as fingerprint of the model. If the fingerprint of a suspected model is the same as the original model, the suspected model is considered as a pirated model. Experimental results demonstrate that the fingerprints can be successfully used to verify the ownership of the model and the test accuracy of the model is not affected. Furthermore, the proposed method is robust to fine-tuning attack, pruning attack, watermark overwriting attack, and adaptive attack.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"5 9","pages":"4649-4659"},"PeriodicalIF":0.0,"publicationDate":"2024-04-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142169670","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}
Advances in deep generative models shed light on de novo molecule generation with desired properties. However, molecule generation targeted for dual protein targets still faces formidable challenges including insufficient protein 3-D structure data requisition for conditioned model training, inflexibility of auto-regressive sampling, and model generalization to unseen targets. Here, this study proposed diffusion model for dual targets-based molecule generation (DiffDTM), a novel unified structure-free deep generative framework based on a diffusion model for dual-target based molecule generation to address the above issues. Specifically, DiffDTM receives representations of protein sequences and molecular graphs pretrained on large-scale datasets as inputs instead of protein and molecular conformations and incorporates an information fusion module to achieve conditional generation in a one-shot manner. We perform comprehensive multiview experiments to demonstrate that DiffDTM can generate druglike, synthesis-accessible, novel, and high-binding affinity molecules targeting specific dual proteins, outperforming the state-of-the-art (SOTA) models in terms of multiple evaluation metrics. Furthermore, DiffDTM could directly generate molecules toward dopamine receptor D2 (DRD2) and 5-hydroxytryptamine receptor 1A (HTR1A) as new antipsychotics. Experimental comparisons highlight the generalizability of DiffDTM to easily adapt to unseen dual targets and generate bioactive molecules, addressing the issues of insufficient active molecule data for model training when new targets are encountered.
{"title":"A Unified Conditional Diffusion Framework for Dual Protein Targets-Based Bioactive Molecule Generation","authors":"Lei Huang;Zheng Yuan;Huihui Yan;Rong Sheng;Linjing Liu;Fuzhou Wang;Weidun Xie;Nanjun Chen;Fei Huang;Songfang Huang;Ka-Chun Wong;Yaoyun Zhang","doi":"10.1109/TAI.2024.3387402","DOIUrl":"https://doi.org/10.1109/TAI.2024.3387402","url":null,"abstract":"Advances in deep generative models shed light on \u0000<italic>de novo</i>\u0000 molecule generation with desired properties. However, molecule generation targeted for dual protein targets still faces formidable challenges including insufficient protein 3-D structure data requisition for conditioned model training, inflexibility of auto-regressive sampling, and model generalization to unseen targets. Here, this study proposed diffusion model for dual targets-based molecule generation (DiffDTM), a novel unified structure-free deep generative framework based on a diffusion model for dual-target based molecule generation to address the above issues. Specifically, DiffDTM receives representations of protein sequences and molecular graphs pretrained on large-scale datasets as inputs instead of protein and molecular conformations and incorporates an information fusion module to achieve conditional generation in a one-shot manner. We perform comprehensive multiview experiments to demonstrate that DiffDTM can generate druglike, synthesis-accessible, novel, and high-binding affinity molecules targeting specific dual proteins, outperforming the state-of-the-art (SOTA) models in terms of multiple evaluation metrics. Furthermore, DiffDTM could directly generate molecules toward dopamine receptor D2 (DRD2) and 5-hydroxytryptamine receptor 1A (HTR1A) as new antipsychotics. Experimental comparisons highlight the generalizability of DiffDTM to easily adapt to unseen dual targets and generate bioactive molecules, addressing the issues of insufficient active molecule data for model training when new targets are encountered.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"5 9","pages":"4595-4606"},"PeriodicalIF":0.0,"publicationDate":"2024-04-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142169693","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 : 2024-04-10DOI: 10.1109/TAI.2024.3386498
Varun Kohli;Muhammad Naveed Aman;Biplab Sikdar
The Internet of Things (IoT) has been a popular topic for research and development in the past decade. The resource-constrained and wireless nature of IoT devices presents a large surface of vulnerabilities, and traditional network security methods involving complex cryptography are not feasible. Studies show that Denial of Service (DoS), physical intrusion, spoofing, and node forgery are prevalent threats in the IoT, and there is a need for robust, lightweight device fingerprinting schemes. We identify eight criteria of effective fingerprinting methods for resource-constrained IoT devices and propose an intelligent, lightweight, whitelist-based fingerprinting method that satisfies these properties. The proposed method uses the power-up Static Random Access Memory (SRAM) stack as fingerprint features and autoencoder networks (AEN) for fingerprint registration and verification. We also present a threat mitigation framework based on network isolation levels to handle potential and identified threats. Experiments are conducted with a heterogeneous pool of 10 advanced virtual reduced instruction set computer (AVR) Harvard architecture prover devices from different vendors, and Dell Latitude and Dell XPS 13 laptops are used as verifier testbeds. The proposed method has a 99.9% accuracy, 100% precision, and 99.6% recall on known and unknown heterogeneous devices, which is an improvement over several past works. The independence of fingerprints stored in the AENs enables easy distribution and update, and the observed evaluation latency ( $sim$$10^{-4}$ s) and data collection latency ( $sim$$1$ s) make our method practical for real-world scenarios. Lastly, we analyze the proposed method with regard to the eight criteria and highlight its limitations for future improvement.
{"title":"An Intelligent Fingerprinting Technique for Low-Power Embedded IoT Devices","authors":"Varun Kohli;Muhammad Naveed Aman;Biplab Sikdar","doi":"10.1109/TAI.2024.3386498","DOIUrl":"https://doi.org/10.1109/TAI.2024.3386498","url":null,"abstract":"The Internet of Things (IoT) has been a popular topic for research and development in the past decade. The resource-constrained and wireless nature of IoT devices presents a large surface of vulnerabilities, and traditional network security methods involving complex cryptography are not feasible. Studies show that Denial of Service (DoS), physical intrusion, spoofing, and node forgery are prevalent threats in the IoT, and there is a need for robust, lightweight device fingerprinting schemes. We identify eight criteria of effective fingerprinting methods for resource-constrained IoT devices and propose an intelligent, lightweight, whitelist-based fingerprinting method that satisfies these properties. The proposed method uses the power-up Static Random Access Memory (SRAM) stack as fingerprint features and autoencoder networks (AEN) for fingerprint registration and verification. We also present a threat mitigation framework based on network isolation levels to handle potential and identified threats. Experiments are conducted with a heterogeneous pool of 10 advanced virtual reduced instruction set computer (AVR) Harvard architecture prover devices from different vendors, and Dell Latitude and Dell XPS 13 laptops are used as verifier testbeds. The proposed method has a 99.9% accuracy, 100% precision, and 99.6% recall on known and unknown heterogeneous devices, which is an improvement over several past works. The independence of fingerprints stored in the AENs enables easy distribution and update, and the observed evaluation latency (\u0000<inline-formula><tex-math>$sim$</tex-math></inline-formula>\u0000 \u0000<inline-formula><tex-math>$10^{-4}$</tex-math></inline-formula>\u0000 s) and data collection latency (\u0000<inline-formula><tex-math>$sim$</tex-math></inline-formula>\u0000 \u0000<inline-formula><tex-math>$1$</tex-math></inline-formula>\u0000 s) make our method practical for real-world scenarios. Lastly, we analyze the proposed method with regard to the eight criteria and highlight its limitations for future improvement.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"5 9","pages":"4519-4534"},"PeriodicalIF":0.0,"publicationDate":"2024-04-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142164998","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}
Due to its strong ability in distribution representation, the diffusion model has been incorporated into offline reinforcement learning (RL) to cover diverse trajectories of the complex behavior policy. However, this also causes several challenges. Training the diffusion model to imitate behavior from the collected trajectories suffers from limited stitching capability which derives better policies from suboptimal trajectories. Furthermore, the inherent randomness of the diffusion model can lead to unpredictable control and dangerous behavior for the robot. To address these concerns, we propose the value-learning-based decision diffuser (V-DD), which consists of the trajectory diffusion module (TDM) and the trajectory evaluation module (TEM). During the training process, the TDM combines the state-value and classifier-free guidance to bolster the ability to stitch suboptimal trajectories. During the inference process, we design the TEM to select a feasible trajectory generated by the diffusion model. Empirical results demonstrate that our method delivers competitive results on the D4RL benchmark and substantially outperforms current diffusion model-based methods on the real-world robot task.
由于扩散模型在分布表示方面的强大能力，它已被纳入离线强化学习（RL），以覆盖复杂行为政策的各种轨迹。然而，这也带来了一些挑战。从收集到的轨迹中训练扩散模型来模仿行为，会受到拼接能力的限制，从而从次优轨迹中得出更好的策略。此外，扩散模型固有的随机性可能会导致机器人无法预测的控制和危险行为。为了解决这些问题，我们提出了基于价值学习的决策扩散器（V-DD），它由轨迹扩散模块（TDM）和轨迹评估模块（TEM）组成。在训练过程中，TDM 结合了状态值和无分类器指导，以提高缝合次优轨迹的能力。在推理过程中，我们设计 TEM 来选择由扩散模型生成的可行轨迹。实证结果表明，我们的方法在 D4RL 基准测试中取得了具有竞争力的结果，并且在实际机器人任务中大大优于当前基于扩散模型的方法。
{"title":"Stabilizing Diffusion Model for Robotic Control With Dynamic Programming and Transition Feasibility","authors":"Haoran Li;Yaocheng Zhang;Haowei Wen;Yuanheng Zhu;Dongbin Zhao","doi":"10.1109/TAI.2024.3387401","DOIUrl":"https://doi.org/10.1109/TAI.2024.3387401","url":null,"abstract":"Due to its strong ability in distribution representation, the diffusion model has been incorporated into offline reinforcement learning (RL) to cover diverse trajectories of the complex behavior policy. However, this also causes several challenges. Training the diffusion model to imitate behavior from the collected trajectories suffers from limited stitching capability which derives better policies from suboptimal trajectories. Furthermore, the inherent randomness of the diffusion model can lead to unpredictable control and dangerous behavior for the robot. To address these concerns, we propose the value-learning-based decision diffuser (V-DD), which consists of the trajectory diffusion module (TDM) and the trajectory evaluation module (TEM). During the training process, the TDM combines the state-value and classifier-free guidance to bolster the ability to stitch suboptimal trajectories. During the inference process, we design the TEM to select a feasible trajectory generated by the diffusion model. Empirical results demonstrate that our method delivers competitive results on the D4RL benchmark and substantially outperforms current diffusion model-based methods on the real-world robot task.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"5 9","pages":"4585-4594"},"PeriodicalIF":0.0,"publicationDate":"2024-04-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142165059","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 : 2024-04-10DOI: 10.1109/TAI.2024.3387406
Kangjia He;Li Liu;Youmin Zhang;Ye Wang;Qun Liu;Guoyin Wang
Counterfactual subgraphs explain graph neural networks (GNNs) by answering the question: “How would the prediction change if a certain subgraph were absent in the input instance?” The differentiable proxy adjacency matrix is prevalent in current counterfactual subgraph discovery studies due to its ability to avoid exhaustive edge searching. However, a prediction gap exists when feeding the proxy matrix with continuous values and the thresholded discrete adjacency matrix to GNNs, compromising the optimization of the subgraph generator. Furthermore, the end-to-end learning schema adopted in the subgraph generator limits the diversity of counterfactual subgraphs. To this end, we propose CF-GFNExplainer, a flow-based approach for learning counterfactual subgraphs. CF-GFNExplainer employs a policy network with a discrete edge removal schema to construct counterfactual subgraph generation trajectories. Additionally, we introduce a loss function designed to guide CF-GFNExplainer's optimization. The discrete adjacency matrix generated in each trajectory eliminates the prediction gap, enhancing the validity of the learned subgraphs. Furthermore, the multitrajectories sampling strategy adopted in CF-GFNExplainer results in diverse counterfactual subgraphs. Extensive experiments conducted on synthetic and real-world datasets demonstrate the effectiveness of the proposed method in terms of validity and diversity.
{"title":"Learning Counterfactual Explanation of Graph Neural Networks via Generative Flow Network","authors":"Kangjia He;Li Liu;Youmin Zhang;Ye Wang;Qun Liu;Guoyin Wang","doi":"10.1109/TAI.2024.3387406","DOIUrl":"https://doi.org/10.1109/TAI.2024.3387406","url":null,"abstract":"Counterfactual subgraphs explain graph neural networks (GNNs) by answering the question: “How would the prediction change if a certain subgraph were absent in the input instance?” The differentiable proxy adjacency matrix is prevalent in current counterfactual subgraph discovery studies due to its ability to avoid exhaustive edge searching. However, a prediction gap exists when feeding the proxy matrix with continuous values and the thresholded discrete adjacency matrix to GNNs, compromising the optimization of the subgraph generator. Furthermore, the end-to-end learning schema adopted in the subgraph generator limits the diversity of counterfactual subgraphs. To this end, we propose CF-GFNExplainer, a flow-based approach for learning counterfactual subgraphs. CF-GFNExplainer employs a policy network with a discrete edge removal schema to construct counterfactual subgraph generation trajectories. Additionally, we introduce a loss function designed to guide CF-GFNExplainer's optimization. The discrete adjacency matrix generated in each trajectory eliminates the prediction gap, enhancing the validity of the learned subgraphs. Furthermore, the multitrajectories sampling strategy adopted in CF-GFNExplainer results in diverse counterfactual subgraphs. Extensive experiments conducted on synthetic and real-world datasets demonstrate the effectiveness of the proposed method in terms of validity and diversity.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"5 9","pages":"4607-4619"},"PeriodicalIF":0.0,"publicationDate":"2024-04-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142169597","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 : 2024-04-10DOI: 10.1109/TAI.2024.3386499
Ziyan Zhang;Bo Jiang;Jin Tang;Jinhui Tang;Bin Luo
Graph convolutional networks (GCNs) gain increasing attention on graph data learning tasks in recent years. However, in many applications, graph may come with an incomplete form where attributes of graph nodes are partially unknown/missing. Existing graph convolutions (GCs) are generally designed on complete graphs which cannot deal with attribute-incomplete graph data directly. To address this problem, in this article, we extend standard GC and develop an explicit Partial Graph Convolution (PaGC) for attribute-incomplete graph data. Our PaGC is derived based on the observation that the core neighborhood aggregator in GC operation can be equivalently viewed as an energy minimization model. Based on it, we can define a novel partial aggregation function and derive PaGC for incomplete graph data. Experiments demonstrate the effectiveness and efficiency of the proposed PaGCN.
{"title":"Incomplete Graph Learning via Partial Graph Convolutional Network","authors":"Ziyan Zhang;Bo Jiang;Jin Tang;Jinhui Tang;Bin Luo","doi":"10.1109/TAI.2024.3386499","DOIUrl":"https://doi.org/10.1109/TAI.2024.3386499","url":null,"abstract":"Graph convolutional networks (GCNs) gain increasing attention on graph data learning tasks in recent years. However, in many applications, graph may come with an incomplete form where attributes of graph nodes are partially unknown/missing. Existing graph convolutions (GCs) are generally designed on complete graphs which cannot deal with attribute-incomplete graph data directly. To address this problem, in this article, we extend standard GC and develop an explicit Partial Graph Convolution (PaGC) for attribute-incomplete graph data. Our PaGC is derived based on the observation that the core neighborhood aggregator in GC operation can be equivalently viewed as an energy minimization model. Based on it, we can define a novel \u0000<italic>partial aggregation function</i>\u0000 and derive PaGC for incomplete graph data. Experiments demonstrate the effectiveness and efficiency of the proposed PaGCN.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"5 9","pages":"4315-4321"},"PeriodicalIF":0.0,"publicationDate":"2024-04-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142165011","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 : 2024-04-09DOI: 10.1109/TAI.2024.3386500
Arunava Roy;Dipankar Dasgupta
Generative adversarial networks (GANs) excel in diverse applications such as image enhancement, manipulation, and generating images and videos from text. Yet, training GANs with large datasets remains computationally intensive for standalone systems. Synchronization issues between the generator and discriminator lead to unstable training, poor convergence, vanishing, and exploding gradient challenges. In decentralized environments, standalone GANs struggle with distributed data on client machines. Researchers have turned to federated learning (FL) for distributed-GAN (D-GAN) implementations, but efforts often fall short due to training instability and poor synchronization within GAN components. In this study, we present DRL-GAN, a lightweight Wasserstein conditional distributed relativistic loss-GAN designed to overcome existing limitations. DRL-GAN ensures training stability in the face of nonconvex losses by employing a single global generator on the central server and a discriminator per client. Utilizing Wasserstein-1 for relativistic loss computation between real and fake samples, DRL-GAN effectively addresses issues, such as mode collapses, vanishing, and exploding gradients, accommodating both iid and non-iid private data in clients and fostering strong convergence. The absence of a robust conditional distributed-GAN model serves as another motivation for this work. We provide a comprehensive mathematical formulation of DRL-GAN and validate our claims empirically on CIFAR-10, MNIST, EuroSAT, and LSUN-Bedroom datasets.
{"title":"A Distributed Conditional Wasserstein Deep Convolutional Relativistic Loss Generative Adversarial Network With Improved Convergence","authors":"Arunava Roy;Dipankar Dasgupta","doi":"10.1109/TAI.2024.3386500","DOIUrl":"https://doi.org/10.1109/TAI.2024.3386500","url":null,"abstract":"Generative adversarial networks (GANs) excel in diverse applications such as image enhancement, manipulation, and generating images and videos from text. Yet, training GANs with large datasets remains computationally intensive for standalone systems. Synchronization issues between the generator and discriminator lead to unstable training, poor convergence, vanishing, and exploding gradient challenges. In decentralized environments, standalone GANs struggle with distributed data on client machines. Researchers have turned to federated learning (FL) for distributed-GAN (D-GAN) implementations, but efforts often fall short due to training instability and poor synchronization within GAN components. In this study, we present DRL-GAN, a lightweight Wasserstein conditional distributed relativistic loss-GAN designed to overcome existing limitations. DRL-GAN ensures training stability in the face of nonconvex losses by employing a single global generator on the central server and a discriminator per client. Utilizing Wasserstein-1 for relativistic loss computation between real and fake samples, DRL-GAN effectively addresses issues, such as mode collapses, vanishing, and exploding gradients, accommodating both iid and non-iid private data in clients and fostering strong convergence. The absence of a robust conditional distributed-GAN model serves as another motivation for this work. We provide a comprehensive mathematical formulation of DRL-GAN and validate our claims empirically on CIFAR-10, MNIST, EuroSAT, and LSUN-Bedroom datasets.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"5 9","pages":"4344-4353"},"PeriodicalIF":0.0,"publicationDate":"2024-04-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142165057","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：