检索结果-内蒙古大学图书馆

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Lei, Shiye Zhang, Sen Tao, Dacheng School of Computer Science The University of Sydney Australia College of Computing & Data Science Nanyang Technological University Singapore

Massive reinforcement learning (RL) data are typically collected to train policies offline without the need for interactions, but the large data volume can cause training inefficiencies. To tackle this issue, we formulate offline behavior distillation (OBD), which synthesizes limited expert behavioral data from sub-optimal RL data, enabling rapid policy learning. We propose two naive OBD objectives, DBC and PBC, which measure distillation performance via the decision difference between policies trained on distilled data and either offline data or a near-expert policy. Due to intractable bi-level optimization, the OBD objective is difficult to minimize to small values, which deteriorates PBC by its distillation performance guarantee with quadratic discount complexity O(1/(1 − γ)2). We theoretically establish the equivalence between the policy performance and action-value weighted decision difference, and introduce action-value weighted PBC (Av-PBC) as a more effective OBD objective. By optimizing the weighted decision difference, Av-PBC achieves a superior distillation guarantee with linear discount complexity O(1/(1 − γ)). Extensive experiments on multiple D4RL datasets reveal that Av-PBC offers significant improvements in OBD performance, fast distillation convergence speed, and robust cross-architecture/optimizer generalization. The code is available at https://***/LeavesLei/OBD. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

End-to-end Compilation is All FPGAs Need: A Unified Overlay-based FPGA Compiler for Deep Learning 25

End-to-end Compilation is All FPGAs Need: A Unified Overlay-...

引用

30th Asia and South Pacific Design Automation Conference, ASP-DAC 2025

作者： Qian, Kai Lu, Haodong Liu, Yinqiu Zhang, Zexu Wang, Kun School of Microelectronics Fudan University China College of Computing and Data Science Nanyang Technological University Singapore

ISBN: (纸本)9798400706356

Field-Programmable Gate Array (FPGA) has shown great application potential in deploying Neural Networks (NNs) due to the characteristics of programmability, low power consumption, etc. However, deploying NNs on FPGA is non-trivial because (1) Mainstream NNs pose significant FPGA architecture design challenges due to their large number of parameters, complex operations, and the need for data optimization, and (2) Supporting the deployment of different machine learning frameworks to FPGA requires significant manual effort, consuming a large amount of time. In this paper, we propose AutoCompiler, a unified compiler for mapping NNs to different FPGAs, along with overlay techniques to enable fast and efficient implementation. To the best of our knowledge, we are the first work to support both Deep Neural Networks (DNNs) and Transformer-based networks for overlay-based FPGA deployment. AutoCompiler comprises three integrated enablers: (1) Model Translator, built on top of a topology-based NNs representation, which can optimize the topology and data representation of the models from an algorithmic level based on different hardware configurations, e.g., DSP utilization, (2) Instruction Generator, which generates pipeline data streams according to various FPGA resource configurations by manipulating the instruction set at the upper level rapidly, and (3) End-to-end optimization, which moves as much of the computational processes as possible onto the FPGA chip and minimizes the interaction between CPU and FPGA. Extensive experiments on various Xilinx FPGAs show that AutoCompiler outperforms state-of-the-art overlay-based compiler by 1.2× - 1.35× and same-level GPUs by 1.15× - 1.59× for classic DNN models, and ViT inference, respectively. © 2025 Copyright is held by the owner/author(s). Publication rights licensed to ACM.

关键词： Program compilers

来源：评论

学校读者我要写书评

暂无评论

Reasoning with Trees: Faithful Question Answering over Knowledge Graph 31

Reasoning with Trees: Faithful Question Answering over Knowl...

引用

31st International Conference on Computational Linguistics, COLING 2025

作者： Shen, Tiesunlong Wang, Jin Zhang, Xuejie Cambria, Erik School of Information Science and Engineering Yunnan University Kunming China College of Computing and Data Science Nanyang Technological University Singapore

ISBN: (纸本)9798891761964

Recent advancements in large language models (LLMs) have shown remarkable progress in reasoning capabilities, yet they still face challenges in complex, multi-step reasoning tasks. This study introduces Reasoning with Trees (RwT), a novel framework that synergistically integrates LLMs with knowledge graphs (KGs) to enhance reasoning performance and interpretability. RwT reformulates knowledge graph question answering (KGQA) as a discrete decision-making problem, leveraging Monte Carlo Tree Search (MCTS) to iteratively refine reasoning paths. This approach mirrors human-like reasoning by dynamically integrating the LLM's internal knowledge with external KG information. We propose a real-data guided iteration technique to train an evaluation model that assesses action values, improving the efficiency of MCTS. Experimental results on two benchmark KGQA datasets demonstrate that RwT significantly outperforms existing state-of-the-art methods, with an average performance improvement of 9.81%. Notably, RwT achieves these improvements without requiring complete LLM retraining, offering a more efficient and adaptable approach to enhancing LLM reasoning capabilities. © 2025 Association for Computational Linguistics.

关键词： Knowledge graph

来源：评论

学校读者我要写书评

暂无评论

FPS Killer - A Moving Target Detection Algorithm in FPS Game Based on Yolov3 Frame and Gradient Descent Algorithm 2

FPS Killer - A Moving Target Detection Algorithm in FPS Game...

引用

2nd International Conference on Advancements in Smart, Secure and Intelligent computing, ASSIC 2024

作者： Huang, Jiayu Liu, Aofan School of Computing Jiangsu Vocational College of Commerce Jiangsu China School of Computing & Data Science Xiamen University Malaysia Sepang Malaysia

ISBN: (纸本)9798350370188

With the continuous development of the economic level of the society, the game industry is getting more and more extensive development among young people. As one of the mainstreams in the game industry, FPS games are warmly welcomed by the majority of players. In order to ensure the fairness of the game, anti-cheat technologies have emerged, such as 'BattlEye' and 'ValveAnti-Cheat'. This article applies the current popular computer vision technology to FPS games and proposes a cheating technology that cannot be detected by traditional anti-cheating methods. However, this project will only be used for academic research and discussion, and it is forbidden to use it for commercial sales and esports. © 2024 IEEE.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

JDC-GCN: joint diversity and consistency graph convolutional network

引用

Neural computing and Applications 2025年 1-17页

作者： Lin, Renjie Yao, Jie Wang, Shiping Guo, Wenzhong College of Computer and Data Science Fuzhou University Fuzhou350116 China Fujian Provincial Key Laboratory of Network Computing and Intelligent Information Processing Fuzhou University Fuzhou350116 China

In real-world scenarios, multi-view data comprises heterogeneous features, with each feature corresponding to a specific view. The objective of multi-view semi-supervised classification is to enhance classification performance by leveraging the inherent complementary and consistent information present within diverse views. Nevertheless, many existing frameworks primarily focus on assigning suitable weights to different views while neglecting the importance of consistent information. In this paper, a multi-view semi-supervised classification framework called joint diversity and consistency graph convolutional network (JDC-GCN) is proposed. Firstly, the structure of graph convolutional network is introduced to the multi-view semi-supervised classification, capable of propagating the label information over the topological structure of multi-view data. Secondly, the proposed JDC-GCN captures the complementary and consistent information from multiple views through two indispensable sub-modules, Diversity-GCN and Consistency-GCN, respectively. Finally, the attention mechanism is leveraged to dynamically adjust the weights of various views, allowing us to measure the significance of heterogeneous features and the consistent graph without introducing additional parameters. Comprehensive experiments on eight multi-view datasets are conducted to validate the effectiveness of the JDC-GCN algorithm. The results show that the proposed method exhibits superior classification performance compared to other state-of-the-art methods. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2025.

关键词： Self-supervised learning

来源：评论

学校读者我要写书评

暂无评论

A New Carry Look-Ahead Adder Architecture Enabling Improved Speed and Energy Efficiency

A New Carry Look-Ahead Adder Architecture Enabling Improved ...

引用

2024 IEEE Pacific Rim Conference on Communications, Computers and Signal Processing, PACRIM 2024

作者： Balasubramanian, Padmanabhan Maskell, Douglas College of Computing and Data Science Nanyang Technological University 50 Nanyang Avenue Singapore639798 Singapore

ISBN: (纸本)9798350362312

We present a new carry look-ahead adder (NCLA) architecture that makes use of non-uniform-size CLA modules in contrast to the existing CLA architecture which typically makes use of uniform-size CLA modules. This helps to achieve improved speed and energy efficiency for the NCLA architecture compared to the existing CLA architecture without incurring notable area and power penalties. To demonstrate the advantage of NCLA, several adders were implemented ranging from the slow ripple carry adder to the generally considered fastest parallel-prefix adder viz. the Kogge-Stone adder (KSA), and their performance metrics were compared. The 32-bit addition was considered an example, and the different adders were implemented in a semi-custom design method using a 28-nm CMOS standard digital cell library. Based on the synthesis results, the NCLA was found to facilitate significant optimizations in design metrics compared to its high-speed counterparts. To give some examples, the NCLA achieved: (i) a 14.7% reduction in delay and a 13.4% reduction in energy compared to a high-speed CLA while occupying slightly more area, (ii) a 42.1% reduction in delay and a 58.3% reduction in energy compared to a conditional sum adder while occupying 8% more area, (iii) a 14.7% reduction in delay and a 37.7% reduction in energy compared to a speed-optimized carry select adder while requiring 37% less area, and (iv) a 20.2% reduction in energy and a 55.4% reduction in area compared to the KSA. © 2024 IEEE.

关键词： CMOS integrated circuits

来源：评论

学校读者我要写书评

暂无评论

You & AI: A Research computing Hackathon: Poster Abstract 36

You & AI: A Research Computing Hackathon: Poster Abstract

引用

36th International Conference on Software Engineering Education and Training, CSEE and T 2024

作者： Deslauriers, James Michalickova, Katerina Pinney, John Gao, Liam Cooling, Chris Research Computing & Data Science Graduate School. Imperial College London London United Kingdom

ISBN: (纸本)9798350378979

In June 2023, the Imperial college London Graduate School's Research computing and data science group invited thirty PhD students for a one-day Hackathon on AI-assisted programming. This poster abstract presents our experiences in planning and running the event, and shares student and organiser reflections on the event and where it might lead. © 2024 IEEE.

关键词： ai-assisted programming generative ai training

来源：评论

学校读者我要写书评

暂无评论

Fast Bipartitioned Hybrid Adder Utilizing Carry Select and Carry Lookahead Logic 9

Fast Bipartitioned Hybrid Adder Utilizing Carry Select and C...

引用

9th International Conference on Mathematics and Computers in sciences and Industry, MCSI 2024

作者： Balasubramanian, Padmanabhan Maskell, Douglas L. Nanyang Technological University College of Computing and Data Science 50 Nanyang Avenue Singapore639798 Singapore

ISBN: (纸本)9798350377224

We present a novel fast bipartitioned hybrid adder (FBHA) that utilizes carry-select and carry-lookahead logic. The proposed FBHA is an accurate adder with a significant part and a less significant part joined together by a carry signal. In an N-bit FBHA, the K-bit less significant part is realized using carry-lookahead adder logic, and the (N - K)-bit significant part is realized using carry-select adder logic. The 32-bit addition was considered as an example operation for this work. Many 32-bit adders ranging from the slow ripple carry adder to the fast parallel-prefix Kogge-Stone adder and the proposed adder were synthesized using a 28-nm CMOS standard cell library and their design metrics were compared. A well-optimized FBHA achieved significant optimizations in design metrics compared to its high-speed adder counterparts and some examples are mentioned as follows: (a) 19.8% reduction in delay compared to a carry-lookahead adder;(b) 19.8% reduction in delay, 24.4% reduction in area, and 19.4% reduction in power compared to a carry-select adder;(c) 45.6% reduction in delay, and 13.5% reduction in power compared to a conditional sum adder;and (d) 46.5% reduction in area, and 29.3% reduction in power compared to the Kogge-Stone adder © 2024 IEEE.

关键词： CMOS integrated circuits

来源：评论

学校读者我要写书评

暂无评论

Multilingual Synopses of Movie Narratives: A dataset for Vision-Language Story Understanding

Multilingual Synopses of Movie Narratives: A Dataset for Vis...

引用

2024 Conference on Empirical Methods in Natural Language Processing, EMNLP 2024

作者： Sun, Yidan Yu, Jianfei Li, Boyang College of Computing and Data Science Nanyang Technological University Singapore School of Computer Science and Engineering Nanjing University of Science and Technology China

ISBN: (纸本)9798891761681

Story video-text alignment, a core task in computational story understanding, aims to align video clips with corresponding sentences in their descriptions. However, progress on the task has been held back by the scarcity of manually annotated video-text correspondence and the heavy concentration on English narrations of Hollywood movies. To address these issues, in this paper, we construct a large-scale multilingual video story dataset named Multilingual Synopses of Movie Narratives (M-SYMON), containing 13,166 movie summary videos from 7 languages, as well as manual annotation of fine-grained video-text correspondences for 101.5 hours of video. Training on the human annotated data from M-SYMON outperforms the SOTA methods by 15.7 and 16.2 percentage points on Clip Accuracy and Sentence IoU scores, respectively, demonstrating the effectiveness of the annotations. As benchmarks for future research, we create 6 baseline approaches with different multilingual training strategies, compare their performance in both intra-lingual and cross-lingual setups, exemplifying the challenges of multilingual video-text alignment. The dataset is released at: https://***/insundaycathy/M-SyMoN. © 2024 Association for Computational Linguistics.

关键词： Visual languages

来源：评论

学校读者我要写书评

暂无评论

Feature Contrast Difference and Enhanced Network for RGB-D Indoor Scene Classification in Internet of Things

引用

IEEE Internet of Things Journal 2025年第11期12卷 17610-17621页

作者： Zhou, Wujie Jian, Bitao Liu, Yuanyuan Zhejiang University of Science and Technology School of Information and Electronic Engineering Hangzhou310023 China China University of Geosciences School of Computer Science and Technology Wuhan430074 China Nanyang Technological University College of Computing and Data Science 308232 Singapore

The era of smart connectivity spawned by the Internet of Things (IoT) has made the need to achieve environmental perception and understanding of different scenarios increasingly urgent. Among the many scenarios, indoor scene classification has attracted much attention because of its its relevance to the daily lives of people, ranging from comfort regulation in living spaces to the optimal allocation of resources in offices, and a variety of approaches for this task have emerged. However, increasing accuracy remains a crucial objective due to the complexity and disorder of indoor scenes. Therefore, we propose a feature contrast difference and enhanced network for RGB-D indoor scene classification, FCDENet. First, the RGB and Depth images express different information. Therefore, we built a feature contrast difference module for the first two low-level features to extend the receptive fields of the different features, utilizing differential contrast to complement each other. Second, the high-level feature semantic information is abstract. Therefore, we introduced information cluster blocks, which are used to aggregate feature points with similar attributes into compact clusters after being parsed by an initial frequency transform, enabling instantiated representations of the semantic information. Finally, to further enhance the integrated features, we introduced a wavelet transform block in the cross-layer decoding process. In contrast to conventional decoding methods, we employed a wavelet transform for initial denoising cross-layer features and used multiple pooling structures to supplement local information, gradually weighting to achieve higher prediction accuracy. Extensive experiments on two typical indoor datasets, NYUDv2 and SUN RGB-D, show that our results exhibit excellent performance. In addition, to better demonstrate the reliability of the method, we conducted generalizability experiments on other datasets, and the proposed method provides a robust soluti

关键词： Wavelet decomposition

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：