检索结果-内蒙古大学图书馆

End-to-End Two-Branch Bionic Network for Autonomous Driving

Tsinghua Science and Technology 2025年第5期30卷 2259-2269页

作者： Guoliang Sun Sifa Zheng Xingrui Gong Yijie Pan Rui Yang Yingying Yu Shanshan Pei School of Vehicle and Mobility Tsinghua University Beijing China Faculty of Innovation Engineering Macau University of Science and Technology Macau China Computer Science and Precision Instrument Tsinghua University Beijing China Beijing Information Science & Technology University Beijing China Faculty of Information Technology Macau University of Science and Technology Macau China

Most traffic accidents are caused by improper driver operation, so autonomous driving based on rapidly developing artificial intelligence technology has attracted much attention. Inspired by the biological visual perception and neural decision-making mechanism, this paper constructs a two-branch bionic network for autonomous driving, which learns to map the driver's perspective image directly to the steering commands. On the real-world driving dataset we collected, extensive experiments prove the efficiency, robustness, superior structure and biological interpretability of this end-to-end algorithm. Moreover, the flexible scalability of this network greatly supports real-time inference and deployment.

关键词： Decision making Prediction algorithms Biology inference algorithms Robustness Real-time systems Autonomous vehicles Vehicles Biomedical imaging Visual perception

来源：评论

学校读者我要写书评

暂无评论

Bayesian Variance Change Point Detection With Credible Sets

引用

IEEE Transactions on Pattern Analysis and Machine Intelligence 2025年第6期47卷 4835-4852页

作者： Lorenzo Cappello Oscar Hernan Madrid Padilla Department of Economics and Business Universitat Pompeu Fabra Barcelona Spain Data Science Center Barcelona School of Economics Barcelona Spain Department of Statististics and Data Sciences UCLA Los Angeles CA USA

This paper introduces a novel Bayesian approach to detect changes in the variance of a Gaussian sequence model, focusing on quantifying the uncertainty in the change point locations and providing a scalable algorithm for inference. We do that by framing the problem as a product of multiple single changes in the scale parameter. We fit the model through an iterative procedure similar to what is done for additive models. The novelty is that each iteration returns a probability distribution on time instances, which captures the uncertainty in the change point location. Leveraging a recent result in the literature, we can show that our proposal is a variational approximation of the exact model posterior distribution. We study the convergence of the algorithm and the change point localization rate. Extensive experiments in simulation studies and applications to biological data illustrate the performance of our method.

关键词： Bayes methods Uncertainty Computational modeling Proposals inference algorithms Approximation algorithms Vectors Additives Measurement uncertainty Location awareness

来源：评论

学校读者我要写书评

暂无评论

Joint Beam Search Integrating CTC, Attention, and Transducer Decoders

IEEE Transactions on Audio, Speech and Language Processing

引用

IEEE Transactions on Audio, Speech and Language Processing 2025年 33卷 598-612页

作者： Yui Sudo Muhammad Shakeel Yosuke Fukumoto Brian Yan Jiatong Shi Yifan Peng Shinji Watanabe Research Division Honda Research Institute Japan Company Ltd. Wakō-shi Saitama Japan Language Technologies Institute Carnegie Mellon University Pittsburgh PA USA Department of Electrical and Computer Engineering Carnegie Mellon University Pittsburgh PA USA

End-to-end automatic speech recognition (E2E-ASR) can be classified by its decoder architectures, such as connectionist temporal classification (CTC), recurrent neural network transducer (RNN-T), attention-based encoder-decoder, and Mask-CTC models. Each decoder architecture has advantages and disadvantages, leading practitioners to switch between these different models depending on application requirements. Instead of building separate models, we propose a joint modeling scheme where four decoders (CTC, RNN-T, attention, and Mask-CTC) share the same encoder – we refer to this as 4D modeling. The 4D model is trained jointly, which will bring model regularization and maximize the model robustness thanks to their complementary properties. To efficiently train the 4D model, we introduce a two-stage training strategy that stabilizes the joint training. In addition, we propose three novel joint beam search algorithms by combining three decoders (CTC, RNN-T, and attention) to further improve performance. These three beam search algorithms differ in which decoder is used as the primary decoder. We carefully evaluate the performance and computational tradeoffs associated with each algorithm. Experimental results demonstrate that the jointly trained 4D model outperforms the E2E-ASR models trained with only one individual decoder. Furthermore, we demonstrate that the proposed joint beam search algorithm outperforms the previously proposed CTC/attention decoding.

关键词： Decoding Hidden Markov models Computational modeling Training Computational efficiency Transducers inference algorithms Accuracy Vectors Translation

来源：评论

学校读者我要写书评

暂无评论

Learning From Crowds by Class-Specific Instance Weighting

引用

IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE 2025年

作者： Ji, Qiang Jiang, Liangxiao Zhang, Wenjun Li, Chaoqun China Univ Geosci Sch Comp Sci Wuhan 430074 Peoples R China China Univ Geosci Sch Math & Phys Wuhan 430074 Peoples R China

In crowdsourcing scenarios, we can acquire each instance's multiple noisy label set from crowd workers and then infer its integrated label via label integration. To further enhance the quality of integrated labels, some noise correction algorithms have been proposed in recent years. Most of them aim to partition the dataset into a clean set and a noise set, followed by training one or multiple classifiers on the clean set to rectify the instances in the noise set. However, they overlook the fact that the class distribution of the clean set is often inconsistent with that of the noise set, resulting in the subpar correction performance of trained classifiers. To mitigate this inconsistency, this paper proposes a class-specific instance weighting-based noise correction (CIWNC) algorithm. In CIWNC, each class's weight is computed based on the class distribution of the clean set firstly. Subsequently, a class-specific weight is computed for each instance using the weight of the class that its integrated label belongs to, as well as its multiple noisy label set. Finally, a classifier is trained on the instance weighted clean set to rectify the instances in the noise set. Experimental results on 34 simulated and two real-world datasets demonstrate that CIWNC outperforms existing state-of-the-art noise correction algorithms significantly.

关键词： Noise Classification algorithms Noise measurement Crowdsourcing Prediction algorithms Clustering algorithms Supervised learning Training inference algorithms Partitioning algorithms Crowdsourcing learning noise correction instance weighting class-specific instance weighting

来源：评论

学校读者我要写书评

暂无评论

MBS: A High-Precision Approximation Method for Softmax and Efficient Hardware Implementation

引用

IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS I-REGULAR PAPERS 2025年

作者： Wu, Yuanchen Xie, Zhiheng Pan, Hongbing Wang, Yuxuan Nanjing Univ Sch Elect Sci & Engn Nanjing 210023 Peoples R China

The softmax function needs to be frequently used in the multi-head attention layer of Transformer networks. Compared to DNNs and other networks, Transformers have higher computational complexity, requiring higher accuracy and hardware performance for softmax function calculations. Therefore, we propose mixed-base softmax (MBS) for the first time for the approximation of the softmax function. This method combines exponential functions with bases of 2 and 4, which is advantageous for hardware implementation. MBS has a high similarity to the softmax function and demonstrates advanced performance during inference in Transformer network. Through algorithm transformation and hardware optimization, we have designed a low-complexity and highly parallel hardware architecture, which only occupies few additional hardware resources compared to base-2 softmax but achieves higher accuracy. Experimental results show that, under TSMC 90nm CMOS technology at the frequency of 0.5 GHz, our design can achieve the efficiency of 236.18 Gps/(mm(2)& sdot;mW) with the area of 4234 mu m(2). Furthermore, MBS exhibits higher computational accuracy and inference precision compared with base-2 softmax.

关键词： Hardware Accuracy Transformers Computational modeling Computer architecture Approximation methods Approximation algorithms Linear approximation inference algorithms Frequency modulation Softmax function mixed-base softmax transformer neural network high-efficiency hardware

来源：评论

学校读者我要写书评

暂无评论

A Distributed Topology-Protecting Collaboration Algorithm: Design and Performance Analysis

引用

IEEE TRANSACTIONS ON CYBERNETICS 2025年 PP卷 PP页

作者： Wang, Zitong Li, Yushan Li, Ying Fang, Chongrong He, Jianping Shanghai Jiao Tong Univ Dept Automat Minist Educ China Shanghai 200240 Peoples R China Shanghai Jiao Tong Univ Key Lab Syst Control & Informat Proc Minist Educ China Shanghai 200240 Peoples R China KTH Royal Inst Technol Div Decis & Control Syst S-11428 Stockholm Sweden

The interaction topology of multiagent systems (MASs) is crucial for effective collaboration. Recent advances in topology inference provide a better understanding of the behaviors of the systems. Nevertheless, external attackers can exploit such techniques, posing a severe privacy breach, while the challenges to the topology protection problem remain unresolved. This article proposes a distributed collaboration algorithm for MASs to defend against topology inference attacks. The novelties include: 1) Compared with traditional noise-adding methods that inject decaying random inputs, the proposed algorithm constructs a novel noise term to increase the irregularity of the agents' states in a distributed manner while satisfying the convergence requirements. 2) A weight-selecting strategy is designed to choose the subtopologies to degrade the topology inference accuracy, further improving the topology-protecting performance. Theoretically, we derive the mean-square convergence factor and the nonasymptotic error bounds of our proposed algorithm. Extensive simulations demonstrate the effectiveness of the proposed algorithm in protecting the topology.

关键词： Topology inference algorithms Collaboration Network topology Noise Heuristic algorithms Protection Accuracy Privacy Convergence Multiagent systems (MASs) network privacy and security network topology topology protection

来源：评论

学校读者我要写书评

暂无评论

Accelerating Codec-based Speech Synthesis with Multi-Token Prediction and Speculative Decoding

Accelerating Codec-based Speech Synthesis with Multi-Token P...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Tan Dat Nguyen Ji-Hoon Kim Jeongsoo Choi Shukjae Choi Jinseok Park Younglo Lee Joon Son Chung Korea Advanced Institute of Science and Technology South Korea 42dot Inc South Korea

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

The goal of this paper is to accelerate codec-based speech synthesis systems with minimum sacrifice to speech quality. We propose an enhanced inference method that allows for flexible trade-offs between speed and quality during inference without requiring additional training. Our core idea is to predict multiple tokens per inference step of the AR module using multiple prediction heads, resulting in a linear reduction in synthesis time as the number of heads increases. Furthermore, we introduce a novel speculative decoding technique that utilises a Viterbi-based algorithm to select the optimal sequence of generated tokens at each decoding step. In our experiments, we demonstrate that the time required to predict each token is reduced by a factor of 4 to 5 compared to baseline models, with minimal quality trade-off or even improvement in terms of speech intelligibility. Audio samples are available at: ***/projects/mtp.

关键词： Training Speech coding Signal processing algorithms Signal processing Predictive models Prediction algorithms inference algorithms Acoustics Decoding Speech synthesis

来源：评论

学校读者我要写书评

暂无评论

MCMC-Based Sparse Bayesian Learning for Channel Estimation Using Gaussian Mixture Models

MCMC-Based Sparse Bayesian Learning for Channel Estimation U...

引用

IEEE Conference on Wireless Communications and Networking

作者： Xiaotian Fan Xingyu Zhou Hao Ye Le Liang Shi Jin National Mobile Communications Research Laboratory Southeast University Nanjing China Department of Electrical and Computer Engineering University of California Santa Cruz CA Purple Mountain Laboratories Nanjing China

ISBN: (数字)9798350368369

ISBN: (纸本)9798350368376

This paper investigates the downlink channel estimation problem for frequency division duplex (FDD) multi-user massive multiple-input multiple-output (MIMO) systems. We model this problem within the sparse Bayesian learning (SBL) framework, where all unknowns are treated as random variables. Due to limited scattering at the base station, the channel exhibits sparsity in the angular domain. By introducing Gaussian mixture priors to characterize the user equipment internal sparsity and partially shared sparsity, we develop a Markov chain Monte Carlo (MCMC) method to implement Bayesian inference and accurately estimate all random variables in the model, including the channel matrix. Experimental results demonstrate that the MCMC-based SBL channel estimation algorithm outperforms existing approaches by over 5 dB in multi-user scenarios while reducing pilot overhead.

关键词： Monte Carlo methods Simulation Channel estimation Scattering Massive MIMO Downlink inference algorithms Bayes methods Random variables Sparse matrices

来源：评论

学校读者我要写书评

暂无评论

ResNet-Enhanced TD3-Based Coordinated Control and Transformer-Based Online Fault Detection System

ResNet-Enhanced TD3-Based Coordinated Control and Transforme...

引用

IEEE Information Technology and Mechatronics Engineering Conference (ITOEC)

作者： Jiazheng Zhang Zhuang Fu Zuohao Hua Bang Liu School of Mechanical Engineering Shanghai Jiao Tong University China

ISBN: (数字)9798331529482

ISBN: (纸本)9798331529499

Rocket attitude and trajectory control has long been a key topic in the field of control systems due to the harsh working environments of rockets and the stringent demands for control responsiveness and robustness. These challenges make stabilizing rocket attitude and trajectory particularly difficult. This paper proposes a multi-thruster coordinated control method for rocket attitude and trajectory based on the TD3 architecture, incorporating ResNet into the reinforcement learning framework, achieving superior results compared to traditional models. Furthermore, to enhance system robustness and provide real-time status updates for operators, the paper introduces an online fault detection system based on a transformer network. This system quantitatively assesses the operational status of each thruster in a short time, offering additional insights for system control and enabling operators to monitor the system’s status more accurately and promptly.

关键词： Training Rockets Attitude control Fault detection Reinforcement learning Transformers Control systems Robustness inference algorithms Trajectory

来源：评论

学校读者我要写书评

暂无评论

Problem Solving-Oriented Programming Knowledge Tracing from Behavior to Thought

Problem Solving-Oriented Programming Knowledge Tracing from ...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Tao Huang Linxia Tang Huali Yang Xinjia Ou Shengze Hu Jing Geng Junjie Hu Faculty of Artificial Intelligence in Education Central China Normal University Wuhan China School of Computer Science and Artificial Intelligence Wuhan Textile University Wuhan China School of Education Jianghan University Wuhan China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Programming knowledge tracing (programming KT) aims to analyze the dynamic programming states in solving problems based on historical behaviors and predict future performance. In programming, a student’s thought process can lead to multiple solutions for the same problem. However, current programming KT models attribute learners’ responses only to knowledge mastery and ability, overlooking thought factors, which creates a contradiction between data fitting and the rationalization of the model inference process. To address this, we propose a problem solving-oriented programming KT (SPKT) method that incorporates programming knowledge, computational thinking, and solving ability to improve attribution accuracy. Specifically, we designed a dual-channel attention network based on the principle of initiative gain to retrieve knowledge. Additionally, we utilized the edit tree distance algorithm to capture fine-grained trajectory representations and employed a redress mechanism with gating to update abilities based on code information. Experiments demonstrate SPKT’s superiority.

关键词： Knowledge engineering Accuracy Systematics Computational modeling Fitting Signal processing algorithms Data models inference algorithms Trajectory Programming profession

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：