Pub Date : 2024-03-22DOI: 10.1109/TAI.2024.3404414
S. Onur Sahin;Suleyman S. Kozat
We study regression (or prediction) of sequential data, which may have missing entries and/or different lengths. This problem is heavily investigated in the machine learning literature since such missingness is a common occurrence in most real-life applications due to data corruption, measurement errors, and similar. To this end, we introduce a novel hierarchical architecture involving a set of long short-term memory (LSTM) networks, which use only the existing inputs in the sequence without any imputations or statistical assumptions on the missing data. To incorporate the missingness information, we partition the input space into different regions in a hierarchical manner based on the “presence-pattern” of the previous inputs and then assign different LSTM networks to these regions. In this sense, we use the LSTM networks as our experts for these regions and adaptively combine their outputs to generate our final output. Our method is generic so that the set of partitioned regions (presence-patterns) that are modeled by the LSTM networks can be customized, and one can readily use other sequential architectures such as gated recurrent unit (GRU) networks and recurrent neural networks (RNNs) as shown in the article. We also provide the computational complexity analysis of the proposed architecture, which is in the same order as a conventional LSTM architecture. In our experiments, our algorithm achieves significant performance improvements on the well-known financial and real-life datasets with respect to the state-of-the-art methods. We also share the source code of our algorithm to facilitate other research and the replicability of our results.
{"title":"Nonlinear Regression With Hierarchical Recurrent Neural Networks Under Missing Data","authors":"S. Onur Sahin;Suleyman S. Kozat","doi":"10.1109/TAI.2024.3404414","DOIUrl":"https://doi.org/10.1109/TAI.2024.3404414","url":null,"abstract":"We study regression (or prediction) of sequential data, which may have missing entries and/or different lengths. This problem is heavily investigated in the machine learning literature since such missingness is a common occurrence in most real-life applications due to data corruption, measurement errors, and similar. To this end, we introduce a novel hierarchical architecture involving a set of long short-term memory (LSTM) networks, which use only the existing inputs in the sequence without any imputations or statistical assumptions on the missing data. To incorporate the missingness information, we partition the input space into different regions in a hierarchical manner based on the “presence-pattern” of the previous inputs and then assign different LSTM networks to these regions. In this sense, we use the LSTM networks as our experts for these regions and adaptively combine their outputs to generate our final output. Our method is generic so that the set of partitioned regions (presence-patterns) that are modeled by the LSTM networks can be customized, and one can readily use other sequential architectures such as gated recurrent unit (GRU) networks and recurrent neural networks (RNNs) as shown in the article. We also provide the computational complexity analysis of the proposed architecture, which is in the same order as a conventional LSTM architecture. In our experiments, our algorithm achieves significant performance improvements on the well-known financial and real-life datasets with respect to the state-of-the-art methods. We also share the source code of our algorithm to facilitate other research and the replicability of our results.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142443130","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}
Stable learning aims to learn a model that generalizes well to arbitrary unseen target domain by leveraging a single source domain. Recent advances in stable learning have focused on balancing the distribution of confounders for each feature to eliminate spurious correlations. However, previous studies treat all features equally without considering the difficulties of confounder balancing associated with different features, and regard irrelevant features as confounders, deteriorating generalization performance. To tackle these issues, this article proposes a novel triplex learning (TriL) based stable learning algorithm, which performs sample reweighting, causal feature selection, and representation learning to remove spurious correlations. Specifically, first, TriL adaptively assigns weights to the confounder balancing term of each feature in accordance with the difficulties of confounder balancing, and aligns the confounder distribution of each feature by learning a group of sample weights. Second, TriL integrates the sample weights into a weighted cross-entropy model to compute causal effects of features for excluding irrelevant features from the confounder set. Finally, TriL relearns a set of sample weights and uses them to guide a new supervised dual-autoencoder containing two classifiers to learn feature representations. TriL forces the results of two classifiers to remain consistent for removing spurious correlations by using a cross-classifier consistency regularization. Extensive experiments on synthetic and two real-world datasets show the superiority of TriL compared with seven methods.
{"title":"Stable Learning via Triplex Learning","authors":"Shuai Yang;Tingting Jiang;Qianlong Dang;Lichuan Gu;Xindong Wu","doi":"10.1109/TAI.2024.3404411","DOIUrl":"https://doi.org/10.1109/TAI.2024.3404411","url":null,"abstract":"Stable learning aims to learn a model that generalizes well to arbitrary unseen target domain by leveraging a single source domain. Recent advances in stable learning have focused on balancing the distribution of confounders for each feature to eliminate spurious correlations. However, previous studies treat all features equally without considering the difficulties of confounder balancing associated with different features, and regard irrelevant features as confounders, deteriorating generalization performance. To tackle these issues, this article proposes a novel triplex learning (TriL) based stable learning algorithm, which performs sample reweighting, causal feature selection, and representation learning to remove spurious correlations. Specifically, first, TriL adaptively assigns weights to the confounder balancing term of each feature in accordance with the difficulties of confounder balancing, and aligns the confounder distribution of each feature by learning a group of sample weights. Second, TriL integrates the sample weights into a weighted cross-entropy model to compute causal effects of features for excluding irrelevant features from the confounder set. Finally, TriL relearns a set of sample weights and uses them to guide a new supervised dual-autoencoder containing two classifiers to learn feature representations. TriL forces the results of two classifiers to remain consistent for removing spurious correlations by using a cross-classifier consistency regularization. Extensive experiments on synthetic and two real-world datasets show the superiority of TriL compared with seven methods.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142442979","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 : 2024-03-22DOI: 10.1109/TAI.2024.3404408
Yu Huang;Jiaxing Liu;Zongshi Zhang;Dui Li;Xuxin Li;Guang Wang
Accurate short-term photovoltaic (PV) power prediction can be crucial for fault detection of the control system and reducing the fault of the PV output control system. However, PV power is highly volatile, and significant power fluctuations cannot be adapted to by the combined model when predicting, thus affecting the stable operation of the PV output control system. In response to this issue, a dynamic combination short-term PV power prediction model of temporal convolutional network (TCN)-bidirectional gated recurrent unit network (BiGRU) and TCN-bidirectional long-short term memory network (BiLSTM) based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is proposed. CEEMDAN is employed to decompose the original PV power data to reduce the volatility of the original data. Constructing two combined models, TCN-BiGRU and TCN-BiLSTM, and training them separately. Introducing ElasticNet, which utilizes both L1 and L2 regularization terms. This approach preserves the sparsity from least absolute shrinkage and selection operator (LASSO) regression regularization while incorporating the smoothness from Ridge regression regularization, effectively avoiding the issue of the combined model getting trapped in a local optimum. In the end, experimental verification is conducted using actual measurement data from a solar power facility in Gansu, China, and another in Xinjiang, China. The simulation results illustrate that the accuracy of PV power prediction can be significantly improved by the proposed forecasting approach. In comparison with the control experiment, the R2 of the Gansu dataset increased by 0.32% at least, and the R2 of the Xinjiang dataset increased by 0.66% at least.
{"title":"Dynamic Combination Forecasting for Short-Term Photovoltaic Power","authors":"Yu Huang;Jiaxing Liu;Zongshi Zhang;Dui Li;Xuxin Li;Guang Wang","doi":"10.1109/TAI.2024.3404408","DOIUrl":"https://doi.org/10.1109/TAI.2024.3404408","url":null,"abstract":"Accurate short-term photovoltaic (PV) power prediction can be crucial for fault detection of the control system and reducing the fault of the PV output control system. However, PV power is highly volatile, and significant power fluctuations cannot be adapted to by the combined model when predicting, thus affecting the stable operation of the PV output control system. In response to this issue, a dynamic combination short-term PV power prediction model of temporal convolutional network (TCN)-bidirectional gated recurrent unit network (BiGRU) and TCN-bidirectional long-short term memory network (BiLSTM) based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is proposed. CEEMDAN is employed to decompose the original PV power data to reduce the volatility of the original data. Constructing two combined models, TCN-BiGRU and TCN-BiLSTM, and training them separately. Introducing ElasticNet, which utilizes both \u0000<italic>L<sub>1</sub></i>\u0000 and \u0000<italic>L<sub>2</sub></i>\u0000 regularization terms. This approach preserves the sparsity from least absolute shrinkage and selection operator (LASSO) regression regularization while incorporating the smoothness from Ridge regression regularization, effectively avoiding the issue of the combined model getting trapped in a local optimum. In the end, experimental verification is conducted using actual measurement data from a solar power facility in Gansu, China, and another in Xinjiang, China. The simulation results illustrate that the accuracy of PV power prediction can be significantly improved by the proposed forecasting approach. In comparison with the control experiment, the \u0000<italic>R<sup>2</sup></i>\u0000 of the Gansu dataset increased by 0.32% at least, and the \u0000<italic>R<sup>2</sup></i>\u0000 of the Xinjiang dataset increased by 0.66% at least.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142442997","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 : 2024-03-21DOI: 10.1109/TAI.2024.3403953
Chunyu Lei;Jifeng Guo;C. L. Philip Chen
Broad learning system (BLS) has to undergo a vectorization operation before modeling image data, which makes it challenging for BLS to learn local semantic features. Thus, various convolutional-based broad learning systems (C-BLSs) have been introduced to address these challenges. Regrettably, the existing C-BLS variants either lack an efficient training algorithm and incremental learning capability or suffer from poor performance. To this end, we propose a novel convolutional broad learning system (ConvBLS) based on the spherical K-means (SKM) algorithm and two-stage multiscale (TSMS) feature fusion, which consists of the convolutional feature layer (CFL), convolutional enhancement layer (CEL), TSMS feature fusion layer, and output layer. First, unlike the current C-BLS, the simple yet efficient SKM algorithm is utilized to learn the weights of CFLs. Compared with random filters, the SKM algorithm enables the CFL to learn more comprehensive spatial features. Second, to further mine the local semantic features, CELs are established to expand the feature space. Third, the TSMS feature fusion layer is proposed to extract more effective multiscale features by integrating deep and broad representations. Thanks to the above elaborate design and the pseudoinverse calculation of the output layer weights, our proposed ConvBLS method is unprecedentedly efficient and effective. Finally, the corresponding incremental learning algorithms are presented for rapid remodeling if the model deems to expand. Experiments and comparisons demonstrate the superiority of our method.
{"title":"ConvBLS: An Effective and Efficient Incremental Convolutional Broad Learning System Combining Deep and Broad Representations","authors":"Chunyu Lei;Jifeng Guo;C. L. Philip Chen","doi":"10.1109/TAI.2024.3403953","DOIUrl":"https://doi.org/10.1109/TAI.2024.3403953","url":null,"abstract":"Broad learning system (BLS) has to undergo a vectorization operation before modeling image data, which makes it challenging for BLS to learn local semantic features. Thus, various convolutional-based broad learning systems (C-BLSs) have been introduced to address these challenges. Regrettably, the existing C-BLS variants either lack an efficient training algorithm and incremental learning capability or suffer from poor performance. To this end, we propose a novel convolutional broad learning system (ConvBLS) based on the spherical K-means (SKM) algorithm and two-stage multiscale (TSMS) feature fusion, which consists of the convolutional feature layer (CFL), convolutional enhancement layer (CEL), TSMS feature fusion layer, and output layer. First, unlike the current C-BLS, the simple yet efficient SKM algorithm is utilized to learn the weights of CFLs. Compared with random filters, the SKM algorithm enables the CFL to learn more comprehensive spatial features. Second, to further mine the local semantic features, CELs are established to expand the feature space. Third, the TSMS feature fusion layer is proposed to extract more effective multiscale features by integrating deep and broad representations. Thanks to the above elaborate design and the pseudoinverse calculation of the output layer weights, our proposed ConvBLS method is unprecedentedly efficient and effective. Finally, the corresponding incremental learning algorithms are presented for rapid remodeling if the model deems to expand. Experiments and comparisons demonstrate the superiority of our method.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142442961","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 : 2024-03-21DOI: 10.1109/TAI.2024.3379969
Minsong Liu;Yuanheng Zhu;Yaran Chen;Dongbin Zhao
Enhancing state representations can effectively mitigate the issue of low sample efficiency in reinforcement learning (RL) within high-dimensional input environments. Existing methods attempt to improve sample efficiency by learning predictive state representations from sequence data. However, there still remain significant challenges in achieving a comprehensive understanding and learning of information within long sequences. Motivated by this, we introduce a transformer-based state predictive representations (TSPR) 1
Our code will be released at https://github.com/gourmet-liu/TSPR
auxiliary task that promotes better representation learning through self-supervised goals. Specifically, we design a transformer-based predictive model to establish unidirectional and bidirectional prediction tasks for predicting state representations within the latent space. TSPR effectively exploits contextual information within sequences to learn more informative state representations, thereby contributing to the enhancement of policy training in RL. Extensive experiments demonstrate that the combination of TSPR with off-policy RL algorithms leads to a substantial improvement in the sample efficiency of RL. Furthermore, TSPR outperforms state-of-the-art sample-efficient RL methods on both the multiple continuous control (DMControl) and discrete control(Atari) tasks.
{"title":"Enhancing Reinforcement Learning via Transformer-Based State Predictive Representations","authors":"Minsong Liu;Yuanheng Zhu;Yaran Chen;Dongbin Zhao","doi":"10.1109/TAI.2024.3379969","DOIUrl":"https://doi.org/10.1109/TAI.2024.3379969","url":null,"abstract":"Enhancing state representations can effectively mitigate the issue of low sample efficiency in reinforcement learning (RL) within high-dimensional input environments. Existing methods attempt to improve sample efficiency by learning predictive state representations from sequence data. However, there still remain significant challenges in achieving a comprehensive understanding and learning of information within long sequences. Motivated by this, we introduce a transformer-based state predictive representations (TSPR)\u0000<xref><sup>1</sup></xref>\u0000<fn><label><sup>1</sup></label><p>Our code will be released at <uri>https://github.com/gourmet-liu/TSPR</uri></p></fn>\u0000 auxiliary task that promotes better representation learning through self-supervised goals. Specifically, we design a transformer-based predictive model to establish unidirectional and bidirectional prediction tasks for predicting state representations within the latent space. TSPR effectively exploits contextual information within sequences to learn more informative state representations, thereby contributing to the enhancement of policy training in RL. Extensive experiments demonstrate that the combination of TSPR with off-policy RL algorithms leads to a substantial improvement in the sample efficiency of RL. Furthermore, TSPR outperforms state-of-the-art sample-efficient RL methods on both the multiple continuous control (DMControl) and discrete control(Atari) tasks.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142165024","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}
The use of deep neural networks (DNNs) in medical images has enabled the development of solutions characterized by the need of leveraging information coming from multiple sources, raising the multimodal deep learning. DNNs are known for their ability to provide hierarchical and high-level representations of input data. This capability has led to the introduction of methods performing data fusion at an intermediate level, preserving the distinctiveness of the heterogeneous sources in modality-specific paths, while learning the way to define an effective combination in a shared representation. However, modeling the intricate relationships between different data remains an open issue. In this article, we aim to improve the integration of data coming from multiple sources. We introduce between layers belonging to different modality-specific paths a transfer module (TM) able to perform the cross-modality calibration of the extracted features, reducing the effects of the less discriminative ones. As case of study, we focus on the axillary lymph nodes (ALNs) metastasis evaluation in malignant breast cancer (BC), a crucial prognostic factor, affecting patient's survival. We propose a multi-input single-output 3-D convolutional neural network (CNN) that considers both images acquired with multiparametric magnetic resonance and clinical information. In particular, we assess the proposed methodology using four architectures, namely BasicNet and three ResNet variants, showing the improvement of the performance obtained by including the TM in the network configuration. Our results achieve up to 90% and 87% of accuracy and area under ROC curve, respectively when the ResNet10 is considered, surpassing various fusion strategies proposed in the literature.
{"title":"Cross-Modality Calibration in Multi-Input Network for Axillary Lymph Node Metastasis Evaluation","authors":"Michela Gravina;Domiziana Santucci;Ermanno Cordelli;Paolo Soda;Carlo Sansone","doi":"10.1109/TAI.2024.3397246","DOIUrl":"https://doi.org/10.1109/TAI.2024.3397246","url":null,"abstract":"The use of deep neural networks (DNNs) in medical images has enabled the development of solutions characterized by the need of leveraging information coming from multiple sources, raising the multimodal deep learning. DNNs are known for their ability to provide hierarchical and high-level representations of input data. This capability has led to the introduction of methods performing data fusion at an intermediate level, preserving the distinctiveness of the heterogeneous sources in modality-specific paths, while learning the way to define an effective combination in a shared representation. However, modeling the intricate relationships between different data remains an open issue. In this article, we aim to improve the integration of data coming from multiple sources. We introduce between layers belonging to different modality-specific paths a transfer module (TM) able to perform the cross-modality calibration of the extracted features, reducing the effects of the less discriminative ones. As case of study, we focus on the axillary lymph nodes (ALNs) metastasis evaluation in malignant breast cancer (BC), a crucial prognostic factor, affecting patient's survival. We propose a multi-input single-output 3-D convolutional neural network (CNN) that considers both images acquired with multiparametric magnetic resonance and clinical information. In particular, we assess the proposed methodology using four architectures, namely BasicNet and three ResNet variants, showing the improvement of the performance obtained by including the TM in the network configuration. Our results achieve up to 90% and 87% of accuracy and area under ROC curve, respectively when the ResNet10 is considered, surpassing various fusion strategies proposed in the literature.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10535471","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142169734","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}
Convolutional neural network (CNN)-based dehazing methods have achieved great success in single image dehazing. However, the absence of real-world haze image datasets hinders the deep development of single image dehazing. To address this issue, we propose a diverse hazy image synthesis method based on generative adversarial network (GAN) and matting. Specially, we train a GAN-based model that can transform a gray image into a hazy image. To boost the diversity of hazy images, we propose to simulate hazy images via image matting, which can fuse a real haze image with another image containing diverse objects. To evaluate the performance of dehazing methods, we propose two novel metrics: part-based peak signal-to-noise ratio (PSNR) and structure similarity index measure (SSIM). Extensive experiments are conducted to show the effectiveness of the proposed model, dataset, and criteria.
基于卷积神经网络（CNN）的去毛刺方法在单幅图像去毛刺方面取得了巨大成功。然而，现实世界中雾霾图像数据集的缺乏阻碍了单幅图像去噪的深入发展。为了解决这个问题，我们提出了一种基于生成式对抗网络（GAN）和消光的多样化雾霾图像合成方法。特别是，我们训练了一个基于 GAN 的模型，该模型可以将灰度图像转化为朦胧图像。为了提高朦胧图像的多样性，我们建议通过图像消隐来模拟朦胧图像，它可以将真实的朦胧图像与另一幅包含不同物体的图像融合在一起。为了评估去雾化方法的性能，我们提出了两个新的指标：基于部分的峰值信噪比（PSNR）和结构相似性指数（SSIM）。我们进行了广泛的实验，以展示所提模型、数据集和标准的有效性。
{"title":"Diverse Hazy Image Synthesis via Coloring Network","authors":"Shengdong Zhang;Xiaoqin Zhang;Shaohua Wan;Wenqi Ren;Liping Zhao;Li Zhao;Linlin Shen","doi":"10.1109/TAI.2024.3379113","DOIUrl":"https://doi.org/10.1109/TAI.2024.3379113","url":null,"abstract":"Convolutional neural network (CNN)-based dehazing methods have achieved great success in single image dehazing. However, the absence of real-world haze image datasets hinders the deep development of single image dehazing. To address this issue, we propose a diverse hazy image synthesis method based on generative adversarial network (GAN) and matting. Specially, we train a GAN-based model that can transform a gray image into a hazy image. To boost the diversity of hazy images, we propose to simulate hazy images via image matting, which can fuse a real haze image with another image containing diverse objects. To evaluate the performance of dehazing methods, we propose two novel metrics: part-based peak signal-to-noise ratio (PSNR) and structure similarity index measure (SSIM). Extensive experiments are conducted to show the effectiveness of the proposed model, dataset, and criteria.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141630909","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 : 2024-03-20DOI: 10.1109/TAI.2024.3403511
Yu Wen;Jiangshuai Huang;Shaoxin Sun;Xiaojie Su
This article presents a systematic design approach to address the challenge of enclosing and tracking a moving target in multirobot systems while accounting for motion and field of view (FOV) constraints. First, a reference trajectory is designed based on relative position measurement which also conforms to the motion and FOV constraints. Subsequently, considering the uncertainty of mobile robots, and combining prescribed performance bound (PPB) technique, an adaptive tracking solutions are designed to force the fleet of robots track and enclose the moving target. Experimental results demonstrate that the robots can efficiently track the provided reference trajectory while ensuring guaranteed transient performance of position and direction tracking errors, account for the motion and FOV constraints, achieve rapid enclosing and tracking of target objects.
{"title":"Enclose and Track a Target of Mobile Robot With Motion and Field of View Constraints Based on Relative Position Measurement","authors":"Yu Wen;Jiangshuai Huang;Shaoxin Sun;Xiaojie Su","doi":"10.1109/TAI.2024.3403511","DOIUrl":"https://doi.org/10.1109/TAI.2024.3403511","url":null,"abstract":"This article presents a systematic design approach to address the challenge of enclosing and tracking a moving target in multirobot systems while accounting for motion and field of view (FOV) constraints. First, a reference trajectory is designed based on relative position measurement which also conforms to the motion and FOV constraints. Subsequently, considering the uncertainty of mobile robots, and combining prescribed performance bound (PPB) technique, an adaptive tracking solutions are designed to force the fleet of robots track and enclose the moving target. Experimental results demonstrate that the robots can efficiently track the provided reference trajectory while ensuring guaranteed transient performance of position and direction tracking errors, account for the motion and FOV constraints, achieve rapid enclosing and tracking of target objects.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142443028","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}
Reinforcement learning (RL) algorithms combined with deep learning architectures have achieved tremendous success in many practical applications. However, the policies obtained by many deep reinforcement learning (DRL) algorithms are seen to suffer from high variance making them less useful in safety-critical applications. In general, it is desirable to have algorithms that give a low iterate-variance while providing a high long-term reward. In this work, we consider the actor–critic (AC) paradigm, where the critic is responsible for evaluating the policy while the feedback from the critic is used by the actor for updating the policy. The updates of both the critic and the actor in the standard AC procedure are run concurrently until convergence. It has been previously observed that updating the actor once after every $L>1$ steps of the critic reduces the iterate variance. In this article, we address the question of what optimal $L$ -value to use in the recursions and propose a data-driven $L$ -update rule that runs concurrently with the AC algorithm with the objective being to minimize the variance of the infinite horizon discounted reward. This update is based on a random search (discrete) parameter optimization procedure that incorporates smoothed functional (SF) estimates. We prove the convergence of our proposed multitimescale scheme to the optimal $L$ and optimal policy pair. Subsequently, through numerical evaluations on benchmark RL tasks, we demonstrate the advantages of our proposed algorithm over multiple state-of-the-art algorithms in the literature.
强化学习（RL）算法与深度学习架构相结合，在许多实际应用中取得了巨大成功。然而，许多深度强化学习（DRL）算法获得的策略都存在高方差的问题，这使得它们在安全关键型应用中的作用大打折扣。一般来说，我们希望算法在提供高长期回报的同时，还能降低迭代方差。在这项工作中，我们考虑了行动者-批评者（AC）范式，即批评者负责评估策略，而行动者则利用批评者的反馈更新策略。在标准 AC 程序中，批判者和行动者的更新同时进行，直至收敛。以前曾观察到，在批判者每执行 $L>1$ 步后更新一次行动者会降低迭代方差。在本文中，我们将讨论在递归中使用什么最优 $L$ 值的问题，并提出一种数据驱动的 $L$ 更新规则，该规则与 AC 算法同时运行，目标是最小化无限期贴现奖励的方差。这种更新基于随机搜索（离散）参数优化程序，其中包含平滑函数（SF）估计值。我们证明了我们提出的多时间尺度方案对最优 $L$ 和最优政策对的收敛性。随后，通过对基准 RL 任务的数值评估，我们证明了我们提出的算法相对于文献中多种最先进算法的优势。
{"title":"Variance-Reduced Deep Actor–Critic With an Optimally Subsampled Actor Recursion","authors":"Lakshmi Mandal;Raghuram Bharadwaj Diddigi;Shalabh Bhatnagar","doi":"10.1109/TAI.2024.3379109","DOIUrl":"https://doi.org/10.1109/TAI.2024.3379109","url":null,"abstract":"Reinforcement learning (RL) algorithms combined with deep learning architectures have achieved tremendous success in many practical applications. However, the policies obtained by many deep reinforcement learning (DRL) algorithms are seen to suffer from high variance making them less useful in safety-critical applications. In general, it is desirable to have algorithms that give a low iterate-variance while providing a high long-term reward. In this work, we consider the actor–critic (AC) paradigm, where the critic is responsible for evaluating the policy while the feedback from the critic is used by the actor for updating the policy. The updates of both the critic and the actor in the standard AC procedure are run concurrently until convergence. It has been previously observed that updating the actor once after every \u0000<inline-formula><tex-math>$L>1$</tex-math></inline-formula>\u0000 steps of the critic reduces the iterate variance. In this article, we address the question of what optimal \u0000<inline-formula><tex-math>$L$</tex-math></inline-formula>\u0000-value to use in the recursions and propose a data-driven \u0000<inline-formula><tex-math>$L$</tex-math></inline-formula>\u0000-update rule that runs concurrently with the AC algorithm with the objective being to minimize the variance of the infinite horizon discounted reward. This update is based on a random search (discrete) parameter optimization procedure that incorporates smoothed functional (SF) estimates. We prove the convergence of our proposed multitimescale scheme to the optimal \u0000<inline-formula><tex-math>$L$</tex-math></inline-formula>\u0000 and optimal policy pair. Subsequently, through numerical evaluations on benchmark RL tasks, we demonstrate the advantages of our proposed algorithm over multiple state-of-the-art algorithms in the literature.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141630910","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 : 2024-03-19DOI: 10.1109/TAI.2024.3375836
Zhongjie Zhu;Lei Zhang;Yongqiang Bai;Yuer Wang;Pei Li
Scene text image superresolution (STISR) aims to enhance the resolution of images containing text within a scene, making the text more readable and easier to recognize. This technique has broad applications in numerous fields such as autonomous driving, document scanning, image retrieval, and so on. However, most existing STISR methods have not fully exploited the multiscale structural and semantic information within scene text images. As a result, the restored text image quality is not sufficient, significantly impacting subsequent tasks such as text detection and recognition. Hence, this article proposes a novel scheme that leverages multiscale structural and semantic priors to efficiently guide text semantic restoration, ultimately yielding high-quality text images. First, a multiscale interaction attention (MSIA) module is designed to capture location-specific details of various-scale structural features and facilitate the recovery of semantic information. Second, a multiscale prior learning module (MSPLM) is developed. Within this module, skip connections are employed among codecs to strengthen both structural and semantic prior features, thereby enhancing the up-sampling and reconstruction capabilities. Finally, building upon the MSPLM, cascaded encoders are connected through residual connections to further enrich the multiscale features and bolster the representational capacity of the prior. Experiments conducted on the standard TextZoom dataset demonstrate that the average recognition accuracies of three evaluators—attentional scene text recognizer (ASTER), convolutional recurrent neural network (CRNN), and multi-object rectified attention network (MORAN)—are 64.4%, 53.5%, and 60.8%, respectively, surpassing most existing methods, including the state-of-the-art ones.
{"title":"Scene Text Image Superresolution Through Multiscale Interaction of Structural and Semantic Priors","authors":"Zhongjie Zhu;Lei Zhang;Yongqiang Bai;Yuer Wang;Pei Li","doi":"10.1109/TAI.2024.3375836","DOIUrl":"https://doi.org/10.1109/TAI.2024.3375836","url":null,"abstract":"Scene text image superresolution (STISR) aims to enhance the resolution of images containing text within a scene, making the text more readable and easier to recognize. This technique has broad applications in numerous fields such as autonomous driving, document scanning, image retrieval, and so on. However, most existing STISR methods have not fully exploited the multiscale structural and semantic information within scene text images. As a result, the restored text image quality is not sufficient, significantly impacting subsequent tasks such as text detection and recognition. Hence, this article proposes a novel scheme that leverages multiscale structural and semantic priors to efficiently guide text semantic restoration, ultimately yielding high-quality text images. First, a multiscale interaction attention (MSIA) module is designed to capture location-specific details of various-scale structural features and facilitate the recovery of semantic information. Second, a multiscale prior learning module (MSPLM) is developed. Within this module, skip connections are employed among codecs to strengthen both structural and semantic prior features, thereby enhancing the up-sampling and reconstruction capabilities. Finally, building upon the MSPLM, cascaded encoders are connected through residual connections to further enrich the multiscale features and bolster the representational capacity of the prior. Experiments conducted on the standard TextZoom dataset demonstrate that the average recognition accuracies of three evaluators—attentional scene text recognizer (ASTER), convolutional recurrent neural network (CRNN), and multi-object rectified attention network (MORAN)—are 64.4%, 53.5%, and 60.8%, respectively, surpassing most existing methods, including the state-of-the-art ones.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-03-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141630968","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：