Pub Date : 2024-03-18DOI: 10.1109/TAI.2024.3377172
Yiming Tang;Jianwei Gao;Witold Pedrycz;Xianghui Hu;Lei Xi;Fuji Ren;Min Hu
At present, there exist some problems in granular clustering methods, such as lack of nonlinear membership description and global optimization of granular data boundaries. To address these issues, in this study, revolving around the parabolic granular data, we propose an overall architecture for parabolic granular modeling and clustering. To begin with, novel coverage and specificity functions are established, and then a parabolic granular data structure is proposed. The fuzzy c-means (FCM) algorithm is used to obtain the numeric prototypes, and then particle swarm optimization (PSO) is introduced to construct the parabolic granular data from the global perspective under the guidance of principle of justifiable granularity (PJG). Combining the advantages of FCM and PSO, we propose the parabolic granular modeling and optimization (PGMO) method. Moreover, we put forward attribute weights and sample weights as well as a distance measure induced by the Gaussian kernel similarity, and then come up with the algorithm of weighted kernel fuzzy clustering for parabolic granularity (WKFC-PG). In addition, the assessment mechanism of parabolic granular clustering is discussed. In summary, we set up an overall architecture including parabolic granular modeling, clustering, and assessment. Finally, comparative experiments on artificial, UCI, and high-dimensional datasets validate that our overall architecture delivers a good improvement over previous strategies. The parameter analysis and time complexity are also given for WKFC-PG. In contrast with related granular clustering algorithms, it is observed that WKFC-PG performs better than other granular clustering algorithms and has superior stability in handling outliers, especially on high-dimensional datasets.
{"title":"Modeling and Clustering of Parabolic Granular Data","authors":"Yiming Tang;Jianwei Gao;Witold Pedrycz;Xianghui Hu;Lei Xi;Fuji Ren;Min Hu","doi":"10.1109/TAI.2024.3377172","DOIUrl":"https://doi.org/10.1109/TAI.2024.3377172","url":null,"abstract":"At present, there exist some problems in granular clustering methods, such as lack of nonlinear membership description and global optimization of granular data boundaries. To address these issues, in this study, revolving around the parabolic granular data, we propose an overall architecture for parabolic granular modeling and clustering. To begin with, novel coverage and specificity functions are established, and then a parabolic granular data structure is proposed. The fuzzy c-means (FCM) algorithm is used to obtain the numeric prototypes, and then particle swarm optimization (PSO) is introduced to construct the parabolic granular data from the global perspective under the guidance of principle of justifiable granularity (PJG). Combining the advantages of FCM and PSO, we propose the parabolic granular modeling and optimization (PGMO) method. Moreover, we put forward attribute weights and sample weights as well as a distance measure induced by the Gaussian kernel similarity, and then come up with the algorithm of weighted kernel fuzzy clustering for parabolic granularity (WKFC-PG). In addition, the assessment mechanism of parabolic granular clustering is discussed. In summary, we set up an overall architecture including parabolic granular modeling, clustering, and assessment. Finally, comparative experiments on artificial, UCI, and high-dimensional datasets validate that our overall architecture delivers a good improvement over previous strategies. The parameter analysis and time complexity are also given for WKFC-PG. In contrast with related granular clustering algorithms, it is observed that WKFC-PG performs better than other granular clustering algorithms and has superior stability in handling outliers, especially on high-dimensional datasets.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141630970","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 : 2024-03-18DOI: 10.1109/TAI.2024.3375258
Ayush Aniket;Arpan Chattopadhyay
We study learning in periodic Markov decision process (MDP), a special type of nonstationary MDP where both the state transition probabilities and reward functions vary periodically, under the average reward maximization setting. We formulate the problem as a stationary MDP by augmenting the state space with the period index and propose a periodic upper confidence bound reinforcement learning-2 (PUCRL2) algorithm. We show that the regret of PUCRL2 varies linearly with the period $N$ and as $mathcal{O}(sqrt{T text{log} T})$ with the horizon length $T$ . Utilizing the information about the sparsity of transition matrix of augmented MDP, we propose another algorithm [periodic upper confidence reinforcement learning with Bernstein bounds (PUCRLB) which enhances upon PUCRL2, both in terms of regret ( $O(sqrt{N})$ dependency on period] and empirical performance. Finally, we propose two other algorithms U-PUCRL2 and U-PUCRLB for extended uncertainty in the environment in which the period is unknown but a set of candidate periods are known. Numerical results demonstrate the efficacy of all the algorithms.
{"title":"Online Reinforcement Learning in Periodic MDP","authors":"Ayush Aniket;Arpan Chattopadhyay","doi":"10.1109/TAI.2024.3375258","DOIUrl":"https://doi.org/10.1109/TAI.2024.3375258","url":null,"abstract":"We study learning in periodic Markov decision process (MDP), a special type of nonstationary MDP where both the state transition probabilities and reward functions vary periodically, under the average reward maximization setting. We formulate the problem as a stationary MDP by augmenting the state space with the period index and propose a periodic upper confidence bound reinforcement learning-2 (PUCRL2) algorithm. We show that the regret of PUCRL2 varies linearly with the period \u0000<inline-formula><tex-math>$N$</tex-math></inline-formula>\u0000 and as \u0000<inline-formula><tex-math>$mathcal{O}(sqrt{T text{log} T})$</tex-math></inline-formula>\u0000 with the horizon length \u0000<inline-formula><tex-math>$T$</tex-math></inline-formula>\u0000. Utilizing the information about the sparsity of transition matrix of augmented MDP, we propose another algorithm [periodic upper confidence reinforcement learning with Bernstein bounds (PUCRLB) which enhances upon PUCRL2, both in terms of regret (\u0000<inline-formula><tex-math>$O(sqrt{N})$</tex-math></inline-formula>\u0000 dependency on period] and empirical performance. Finally, we propose two other algorithms U-PUCRL2 and U-PUCRLB for extended uncertainty in the environment in which the period is unknown but a set of candidate periods are known. Numerical results demonstrate the efficacy of all the algorithms.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141630960","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 : 2024-03-18DOI: 10.1109/TAI.2024.3376319
Kanisius Karyono;Badr M. Abdullah;Alison J. Cotgrave;Ana Bras;Jeff Cullen
The artificial intelligence (AI) system faces the challenge of insufficient training datasets and the risk of an uncomfortable user experience during the data gathering and learning process. The unreliable training data leads to overfitting and poor system performance which will result in wasting operational energy. This work introduces a reliable data set for training the AI subsystem for thermal comfort. The most reliable current training data sets for thermal comfort are ASHRAE RP-884 and ASHRAE Global Thermal Comfort Database II, but the direct use of these data for learning will give a poor learning result of less than 60% accuracy. This article presents the algorithm for data filtering and semantic data augmentation for the multiple ASHRAE databases for the supervised learning process. The result was verified with the visual psychrometric chart method that can check for overfitting and verified by developing the Internet of Things (IoT) control system for residential usage based on shallow supervised learning. The AI system was a wide artificial neural network (ANN) which is simple enough to be implemented in a local node. The filtering and semantic augmentation method can increase the accuracy to 96.1%. The control algorithm that was developed based on the comfort zone identification can increase the comfort acknowledgement by 6.06% leading to energy saving for comfort. This work can contribute to 717.2 thousand tonnes of CO 2 equivalent per year which is beneficial for a more sustainable thermal comfort system and the development of a reinforced learning system for thermal comfort.
{"title":"Developing a Reliable Shallow Supervised Learning for Thermal Comfort Using Multiple ASHRAE Databases","authors":"Kanisius Karyono;Badr M. Abdullah;Alison J. Cotgrave;Ana Bras;Jeff Cullen","doi":"10.1109/TAI.2024.3376319","DOIUrl":"https://doi.org/10.1109/TAI.2024.3376319","url":null,"abstract":"The artificial intelligence (AI) system faces the challenge of insufficient training datasets and the risk of an uncomfortable user experience during the data gathering and learning process. The unreliable training data leads to overfitting and poor system performance which will result in wasting operational energy. This work introduces a reliable data set for training the AI subsystem for thermal comfort. The most reliable current training data sets for thermal comfort are ASHRAE RP-884 and ASHRAE Global Thermal Comfort Database II, but the direct use of these data for learning will give a poor learning result of less than 60% accuracy. This article presents the algorithm for data filtering and semantic data augmentation for the multiple ASHRAE databases for the supervised learning process. The result was verified with the visual psychrometric chart method that can check for overfitting and verified by developing the Internet of Things (IoT) control system for residential usage based on shallow supervised learning. The AI system was a wide artificial neural network (ANN) which is simple enough to be implemented in a local node. The filtering and semantic augmentation method can increase the accuracy to 96.1%. The control algorithm that was developed based on the comfort zone identification can increase the comfort acknowledgement by 6.06% leading to energy saving for comfort. This work can contribute to 717.2 thousand tonnes of CO\u0000<sub>2</sub>\u0000 equivalent per year which is beneficial for a more sustainable thermal comfort system and the development of a reinforced learning system for thermal comfort.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141630992","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}
In the era of deep learning, federated learning (FL) presents a promising approach that allows multiinstitutional data owners, or clients, to collaboratively train machine learning models without compromising data privacy. However, most existing FL approaches rely on a centralized server for global model aggregation, leading to a single point of failure. This makes the system vulnerable to malicious attacks when dealing with dishonest clients. In this work, we address this problem by proposing a secure and reliable FL system based on blockchain and distributed ledger technology. Our system incorporates a peer-to-peer voting mechanism and a reward-and-slash mechanism, which are powered by on-chain smart contracts, to detect and deter malicious behaviors. Both theoretical and empirical analyses are presented to demonstrate the effectiveness of the proposed approach, showing that our framework is robust against malicious client-side behaviors.
{"title":"Defending Against Poisoning Attacks in Federated Learning With Blockchain","authors":"Nanqing Dong;Zhipeng Wang;Jiahao Sun;Michael Kampffmeyer;William Knottenbelt;Eric Xing","doi":"10.1109/TAI.2024.3376651","DOIUrl":"https://doi.org/10.1109/TAI.2024.3376651","url":null,"abstract":"In the era of deep learning, federated learning (FL) presents a promising approach that allows multiinstitutional data owners, or clients, to collaboratively train machine learning models without compromising data privacy. However, most existing FL approaches rely on a centralized server for global model aggregation, leading to a single point of failure. This makes the system vulnerable to malicious attacks when dealing with dishonest clients. In this work, we address this problem by proposing a secure and reliable FL system based on blockchain and distributed ledger technology. Our system incorporates a peer-to-peer voting mechanism and a reward-and-slash mechanism, which are powered by on-chain smart contracts, to detect and deter malicious behaviors. Both theoretical and empirical analyses are presented to demonstrate the effectiveness of the proposed approach, showing that our framework is robust against malicious client-side behaviors.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10471193","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141631034","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 : 2024-03-17DOI: 10.1109/TAI.2024.3400751
Zijuan Zhao;Zequn Zhu;Yuan Liu;Jinli Guo;Kai Yang
Heterogeneous graph neural networks (HGNNs) have demonstrated promising capabilities in addressing various problems defined on heterogeneous graphs containing multiple types of nodes or edges. However, traditional HGNN models depend on label information and capture the local structural information of the original graph. In this article, we propose a novel heterogeneous graph contrastive learning method with augmentation graph (AHGCL). Specifically, we construct an augmentation graph by calculating the feature similarity of nodes to capture latent structural information. For the original graph and the augmentation graph, we employ a shared graph neural network (GNN) encoder to extract the semantic features of nodes with different meta-paths. The feature information is aggregated through a semantic-level attention mechanism to generate final node embeddings, which capture latent high-order semantic structural information. Considering the problems of label information for the real-world datasets, we adopt contrastive learning to train the GNN encoder for maximizing the common information between similar nodes from the original graph and the augmentation graph views. We conduct node classification experiments on four real-world datasets, AMiner, Freebase, digital bibliography & library project (DBLP), and association for computing machinery (ACM), to evaluate the performance of AHGCL. The results show that the proposed AHGCL demonstrates excellent stability and capability compared to existing graph representation learning methods.
{"title":"Heterogeneous Graph Contrastive Learning With Augmentation Graph","authors":"Zijuan Zhao;Zequn Zhu;Yuan Liu;Jinli Guo;Kai Yang","doi":"10.1109/TAI.2024.3400751","DOIUrl":"https://doi.org/10.1109/TAI.2024.3400751","url":null,"abstract":"Heterogeneous graph neural networks (HGNNs) have demonstrated promising capabilities in addressing various problems defined on heterogeneous graphs containing multiple types of nodes or edges. However, traditional HGNN models depend on label information and capture the local structural information of the original graph. In this article, we propose a novel heterogeneous graph contrastive learning method with augmentation graph (AHGCL). Specifically, we construct an augmentation graph by calculating the feature similarity of nodes to capture latent structural information. For the original graph and the augmentation graph, we employ a shared graph neural network (GNN) encoder to extract the semantic features of nodes with different meta-paths. The feature information is aggregated through a semantic-level attention mechanism to generate final node embeddings, which capture latent high-order semantic structural information. Considering the problems of label information for the real-world datasets, we adopt contrastive learning to train the GNN encoder for maximizing the common information between similar nodes from the original graph and the augmentation graph views. We conduct node classification experiments on four real-world datasets, AMiner, Freebase, digital bibliography & library project (DBLP), and association for computing machinery (ACM), to evaluate the performance of AHGCL. The results show that the proposed AHGCL demonstrates excellent stability and capability compared to existing graph representation learning methods.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142442995","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 : 2024-03-16DOI: 10.1109/TAI.2024.3402193
Site Qu;Guoqiang Hu
When utilizing end-to-end learn-to-construct methods to solve routing problems for multiagent systems, the model is usually trained individually for different problem scales (i.e., the number of customers to be concurrently served within a map) to make the model adaptive to the corresponding scale, ensuring good solution quality. Otherwise, the model trained for one specific scale can lead to poor performance when applied to another different scale, and this situation can get worse when the scale discrepancy increases. Such a separate training strategy is inefficient and time-intensive. In this article, we propose a mix-scale learning framework that requires only a single training session, enabling the model to effectively plan high-quality routes for various problem scales. Based on the capacitated vehicle routing problem (CVRP), the test results reveal that: for problem scales which are no matter seen or unseen during training, our once-trained model can produce solution routes with performance comparable or even superior to those of individually trained models, and offer the highest average solution quality with improvement ratio ranging from 2.28% to 8.07%, which effectively spares the separate training session for each specific scale. Additionally, the extended comparison analysis with individually trained models on real-world benchmark dataset from CVRPLib further highlights our once-trained model's generalization performance across various problem scales and diverse node distributions.
{"title":"Scalable Learning for Multiagent Route Planning: Adapting to Diverse Task Scales","authors":"Site Qu;Guoqiang Hu","doi":"10.1109/TAI.2024.3402193","DOIUrl":"https://doi.org/10.1109/TAI.2024.3402193","url":null,"abstract":"When utilizing end-to-end learn-to-construct methods to solve routing problems for multiagent systems, the model is usually trained individually for different problem scales (i.e., the number of customers to be concurrently served within a map) to make the model adaptive to the corresponding scale, ensuring good solution quality. Otherwise, the model trained for one specific scale can lead to poor performance when applied to another different scale, and this situation can get worse when the scale discrepancy increases. Such a separate training strategy is inefficient and time-intensive. In this article, we propose a mix-scale learning framework that requires only a single training session, enabling the model to effectively plan high-quality routes for various problem scales. Based on the capacitated vehicle routing problem (CVRP), the test results reveal that: for problem scales which are no matter seen or unseen during training, our once-trained model can produce solution routes with performance comparable or even superior to those of individually trained models, and offer the highest average solution quality with improvement ratio ranging from 2.28% to 8.07%, which effectively spares the separate training session for each specific scale. Additionally, the extended comparison analysis with individually trained models on real-world benchmark dataset from CVRPLib further highlights our once-trained model's generalization performance across various problem scales and diverse node distributions.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142443092","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}
While deep neural networks (DNNs) have been widely used in various X-ray image analytics tasks such as classification, segmentation, detection, etc., there frequently needs to collect and annotate a huge amount of training data to train a model for every single task. In this work, we proposed a multitask self-supervised pretraining strategy MTPret to improve the performance of DNNs in various X-ray analytics tasks. MTPret first trains the backbone to learn visual representations from multiple datasets of different tasks through contrastive learning, then MTPret leverages a multitask continual learning to learn discriminative features from various downstream tasks. To evaluate the performance of MTPret , we collected eleven X-ray image datasets from different body parts, such as heads, chest, lungs, bones, and etc., for various tasks to pretrain backbones, and fine-tuned the networks on seven of the tasks. The evaluation results on top of the seven tasks showed MTPret outperformed a large number of baseline methods, including other initialization strategies, pretrained models, and task-specific algorithms in recent studies. In addition, we also performed experiments based on two external tasks, where the datasets of external tasks have not been used in pretraining. The excellent performance of MTPret further confirmed the generalizability and superiority of the proposed multitask self-supervised pretraining.
虽然深度神经网络（DNN）已被广泛应用于分类、分割、检测等各种 X 射线图像分析任务中，但要为每个任务训练一个模型，往往需要收集和注释大量训练数据。在这项工作中，我们提出了一种多任务自监督预训练策略 MTPret，以提高 DNNs 在各种 X 射线分析任务中的性能。MTPret 首先通过对比学习训练骨干网络从不同任务的多个数据集中学习视觉表征，然后利用多任务持续学习从各种下游任务中学习判别特征。为了评估 MTPret 的性能，我们收集了 11 个不同身体部位（如头部、胸部、肺部、骨骼等）的 X 射线图像数据集，针对不同任务对骨干进行预训练，并在其中 7 个任务上对网络进行了微调。对七项任务的评估结果显示，MTPret优于大量基线方法，包括其他初始化策略、预训练模型和近期研究中的特定任务算法。此外，我们还基于两个外部任务进行了实验，其中外部任务的数据集未用于预训练。MTPret 的出色表现进一步证实了所提出的多任务自监督预训练的普适性和优越性。
{"title":"MTPret: Improving X-Ray Image Analytics With Multitask Pretraining","authors":"Weibin Liao;Qingzhong Wang;Xuhong Li;Yi Liu;Zeyu Chen;Siyu Huang;Dejing Dou;Yanwu Xu;Haoyi Xiong","doi":"10.1109/TAI.2024.3400750","DOIUrl":"https://doi.org/10.1109/TAI.2024.3400750","url":null,"abstract":"While deep neural networks (DNNs) have been widely used in various X-ray image analytics tasks such as classification, segmentation, detection, etc., there frequently needs to collect and annotate a huge amount of training data to train a model for every single task. In this work, we proposed a multitask self-supervised pretraining strategy \u0000<italic>MTPret</i>\u0000 to improve the performance of DNNs in various X-ray analytics tasks. \u0000<italic>MTPret</i>\u0000 first trains the backbone to learn visual representations from multiple datasets of different tasks through contrastive learning, then \u0000<italic>MTPret</i>\u0000 leverages a multitask continual learning to learn discriminative features from various downstream tasks. To evaluate the performance of \u0000<italic>MTPret</i>\u0000, we collected eleven X-ray image datasets from different body parts, such as heads, chest, lungs, bones, and etc., for various tasks to pretrain backbones, and fine-tuned the networks on seven of the tasks. The evaluation results on top of the seven tasks showed \u0000<italic>MTPret</i>\u0000 outperformed a large number of baseline methods, including other initialization strategies, pretrained models, and task-specific algorithms in recent studies. In addition, we also performed experiments based on two external tasks, where the datasets of external tasks have not been used in pretraining. The excellent performance of \u0000<italic>MTPret</i>\u0000 further confirmed the generalizability and superiority of the proposed multitask self-supervised pretraining.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142169724","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 : 2024-03-15DOI: 10.1109/TAI.2024.3400749
Yuenan Li;Xin Tian;Qiang Zhu;Min Wu
This article presents a computational solution that enables continuous cardiac monitoring through cross-modality inference of electrocardiogram (ECG). While some smartwatches now allow users to obtain a 30-s ECG test by tapping a built-in bio-sensor, these short-term ECG tests often miss intermittent and asymptomatic abnormalities of cardiac functions. It is also infeasible to expect persistently active user participation for long-term continuous cardiac monitoring in order to capture these and other types of cardiac abnormalities. To alleviate the need for continuous user attention and active participation, we design a lightweight neural network that infers ECG from the photoplethysmogram (PPG) signal sensed at the skin surface by a wearable optical sensor. We also develop a diagnosis-oriented training strategy to enable the neural network to capture the pathological features of ECG, aiming to increase the utility of reconstructed ECG signals for screening cardiovascular diseases (CVDs). We also leverage model interpretation to obtain insights from data-driven models, for example, to reveal some associations between CVDs and ECG/PPG and to demonstrate how the neural network copes with motion artifacts in the ambulatory application. The experimental results on three datasets demonstrate the feasibility of inferring ECG from PPG, achieving a high fidelity of ECG reconstruction with only about 40 000 parameters.
{"title":"Inferring Electrocardiography From Optical Sensing Using Lightweight Neural Network","authors":"Yuenan Li;Xin Tian;Qiang Zhu;Min Wu","doi":"10.1109/TAI.2024.3400749","DOIUrl":"https://doi.org/10.1109/TAI.2024.3400749","url":null,"abstract":"This article presents a computational solution that enables continuous cardiac monitoring through cross-modality inference of electrocardiogram (ECG). While some smartwatches now allow users to obtain a 30-s ECG test by tapping a built-in bio-sensor, these short-term ECG tests often miss intermittent and asymptomatic abnormalities of cardiac functions. It is also infeasible to expect persistently active user participation for long-term continuous cardiac monitoring in order to capture these and other types of cardiac abnormalities. To alleviate the need for continuous user attention and active participation, we design a lightweight neural network that infers ECG from the photoplethysmogram (PPG) signal sensed at the skin surface by a wearable optical sensor. We also develop a diagnosis-oriented training strategy to enable the neural network to capture the pathological features of ECG, aiming to increase the utility of reconstructed ECG signals for screening cardiovascular diseases (CVDs). We also leverage model interpretation to obtain insights from data-driven models, for example, to reveal some associations between CVDs and ECG/PPG and to demonstrate how the neural network copes with motion artifacts in the ambulatory application. The experimental results on three datasets demonstrate the feasibility of inferring ECG from PPG, achieving a high fidelity of ECG reconstruction with only about 40 000 parameters.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141725567","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 : 2024-03-15DOI: 10.1109/TAI.2024.3400929
Xuezhen Liu;Yongyi Chen;Dan Zhang;Ruqiang Yan;Hongjie Ni
Accurately estimating the remaining useful life (RUL) of aircraft engines can effectively prevent aircraft crashes and human casualties. In some RUL prediction methods, particularly for aircraft engines running under complex conditions, they are difficult to comprehensively characterize the engine degradation process, resulting in poor predicted RUL. To address the above challenge, a multichannel long-term external attention network (MLEAN) is proposed for the RUL prediction of turbofan engines. First, the preprocessed samples are transformed to enable MLEAN to focus on learning inter-sensor correlations within the same degradation stage. To improve the feature representation capability of the network, multichannel time attention network (MTANet) is then designed to realize multiscale and multifrequency feature learning, which effectively achieves multiperspective analysis of long-term dependencies in different channels. Then, external attention block (EAB) is introduced to memorize important degraded features from different samples, which can improve the ability of global feature extraction and generalization ability of the network. The performance of MLEAN is examined on the C-MAPSS public dataset. The evaluation metrics RMSE and score values are 13.71 and 680, respectively. In comparison experiments, the proposed MLEAN performs better than the listed state-of-the-art RUL prediction methods.
{"title":"A Multichannel Long-Term External Attention Network for Aeroengine Remaining Useful Life Prediction","authors":"Xuezhen Liu;Yongyi Chen;Dan Zhang;Ruqiang Yan;Hongjie Ni","doi":"10.1109/TAI.2024.3400929","DOIUrl":"https://doi.org/10.1109/TAI.2024.3400929","url":null,"abstract":"Accurately estimating the remaining useful life (RUL) of aircraft engines can effectively prevent aircraft crashes and human casualties. In some RUL prediction methods, particularly for aircraft engines running under complex conditions, they are difficult to comprehensively characterize the engine degradation process, resulting in poor predicted RUL. To address the above challenge, a multichannel long-term external attention network (MLEAN) is proposed for the RUL prediction of turbofan engines. First, the preprocessed samples are transformed to enable MLEAN to focus on learning inter-sensor correlations within the same degradation stage. To improve the feature representation capability of the network, multichannel time attention network (MTANet) is then designed to realize multiscale and multifrequency feature learning, which effectively achieves multiperspective analysis of long-term dependencies in different channels. Then, external attention block (EAB) is introduced to memorize important degraded features from different samples, which can improve the ability of global feature extraction and generalization ability of the network. The performance of MLEAN is examined on the C-MAPSS public dataset. The evaluation metrics RMSE and score values are 13.71 and 680, respectively. In comparison experiments, the proposed MLEAN performs better than the listed state-of-the-art RUL prediction methods.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142443083","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：