检索结果-内蒙古大学图书馆

Optimizing over multiple distributions under generalized quasar-convexity condition 24

学校读者我要写书评

暂无评论

Optimizing over multiple distributions under generalized qua...

Proceedings of the 38th International Conference on Neural Information Processing Systems

作者： Shihong Ding Long Yang Luo Luo Cong Fang State Key Lab of General AI School of Intelligence Science and Technology Peking University School of Data Science Fudan University and Shanghai Key Laboratory for Contemporary Applied Mathematics State Key Lab of General AI School of Intelligence Science and Technology Peking University and Institute for Artificial Intelligence Peking University

ISBN: (纸本)9798331314385

We study a typical optimization model where the optimization variable is composed of multiple probability distributions. Though the model appears frequently in practice, such as for policy problems, it lacks specific analysis in the general setting. For this optimization problem, we propose a new structural condition/landscape description named generalized quasar-convexity (GQC) beyond the realms of convexity. In contrast to original quasar-convexity [24], GQC allows an individual quasar-convex parameter γi for each variable block i and the smaller of γi implies less block-convexity. To minimize the objective function, we consider a generalized oracle termed as the internal function that includes the standard gradient oracle as a special case. We provide optimistic mirror descent (OMD) for multiple distributions and prove that the algorithm can achieve an adaptive Õ(Σdi=1 1/γi)ε-1 iteration complexity to find an ε-suboptimal global solution without pre-known the exact values of γi when the objective admits "polynomial-like" structural. Notably, it achieves iteration complexity that does not explicitly depend on the number of distributions and strictly faster (σdi=1 1/γi v.s. d maxi∈[1:d] 1/γi) than mirror decent methods. We also extend GQC to the minimax optimization problem proposing the generalized quasar-convexity-concavity (GQCC) condition and a decentralized variant of OMD with regularization. Finally, we show the applications of our algorithmic framework on discounted Markov Decision Processes problem and Markov games, which bring new insights on the landscape analysis of reinforcement learning.

关键词：

Xiezhi: An Ever-Updating Benchmark for Holistic Domain Knowledge Evaluation

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Gu, Zhouhong Zhu, Xiaoxuan Ye, Haoning Zhang, Lin Wang, Jianchen Zhu, Yixin Jiang, Sihang Xiong, Zhuozhi Li, Zihan Wu, Weijie He, Qianyu Xu, Rui Huang, Wenhao Liu, Jingping Wang, Zili Wang, Shusen Zheng, Weiguo Feng, Hongwei Xiao, Yanghua Shanghai Key Laboratory of Data Science School of Computer Science Fudan University China School of Information Science and Engineering East China University of Science and Technology China School of Data Science Fudan University China Fudan-Aishu Cognitive Intelligence Joint Research Center China Research Group of Computational and AI Communication Institute for Global Communications and Integrated Media Fudan University China

New Natural Langauge Process (NLP) benchmarks are urgently needed to align with the rapid development of large language models (LLMs). We present Xiezhi, the most comprehensive evaluation suite designed to assess holistic domain knowledge. Xiezhi comprises multiple-choice questions across 516 diverse disciplines ranging from 13 different subjects with 249,587 questions and accompanied by Xiezhi-Specialty with 14,041 questions and Xiezhi-Interdiscipline with 10,746 questions. We conduct evaluation of the 47 cutting-edge LLMs on Xiezhi. Results indicate that LLMs exceed average performance of humans in science, engineering, agronomy, medicine, and art, but fall short in economics, jurisprudence, pedagogy, literature, history, and management. © 2023, CC BY.

关键词： Domain Knowledge

Token2vec: A Joint Self-Supervised Pre-Training Framework Using Unpaired Speech and Text

学校读者我要写书评

暂无评论

Token2vec: A Joint Self-Supervised Pre-Training Framework Us...

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

作者： Xianghu Yue Junyi Ao Xiaoxue Gao Haizhou Li Department of Electrical and Computer Engineering National University of Singapore Singapore Shenzhen Research Institute of Big Data Shenzhen China School of Data Science The Chinese University of Hong Kong Shenzhen China

Self-supervised pre-training has been successful in both text and speech processing. Speech and text offer different but complementary information. The question is whether we are able to perform a speech-text joint pre-training on unpaired speech and text. In this paper, we take the idea of self-supervised pre-training one step further and propose token2vec, a novel joint pre-training framework for unpaired speech and text based on discrete representations of speech. Specifically, we introduce two modality-specific tokenizers for speech and text. Based on these tokenizers, we convert speech/text sequences into discrete speech/text token sequences consisting of similar language units, thus mitigating the domain mismatch problem and length mismatch problem, which are caused by the distinct characteristics between speech and text. Finally, we feed the discrete speech and text tokens into a modality-agnostic Transformer encoder and pre-train with token-level masking language modeling (tMLM). Experiments show that token2vec is significantly superior to various speech-only pre-training baselines, with up to 17.7% relative WER reduction. Token2vec model is also validated on a non-ASR task, i.e., spoken intent classification, and shows good transferability.

关键词： Analytical models Signal processing Benchmark testing Transformers Acoustics Feeds Task analysis

Q&A: Query-Based Representation Learning for Multi-Track Symbolic Music re-Arrangement

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Zhao, Jingwei Xia, Gus Wang, Ye School of Computing NUS Singapore Institute of Data Science NUS Singapore Integrative Sciences and Engineering Programme NUS Graduate School Singapore Music X Lab NYU Shanghai China MBZUAI United Arab Emirates

Music rearrangement is a common music practice of reconstructing and reconceptualizing a piece using new composition or instrumentation styles, which is also an important task of automatic music generation. Existing studies typically model the mapping from a source piece to a target piece via supervised learning. In this paper, we tackle rearrangement problems via self-supervised learning, in which the mapping styles can be regarded as conditions and controlled in a flexible way. Specifically, we are inspired by the representation disentanglement idea and propose Q&A, a query-based algorithm for multi-track music rearrangement under an encoder-decoder framework. Q&A learns both a content representation from the mixture and function (style) representations from each individual track, while the latter queries the former in order to rearrange a new piece. Our current model focuses on popular music and provides a controllable pathway to four scenarios: 1) re-instrumentation, 2) piano cover generation, 3) orchestration, and 4) voice separation. Experiments show that our query system achieves high-quality rearrangement results with delicate multi-track structures, significantly outperforming the baselines. © 2023, CC BY.

关键词： Music

SSD: A State-based Stealthy Backdoor Attack For IMU/GNSS Navigation System in UAV Route Planning

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Wang, Zhaoxuan Li, Yang Zhang, Jie Han, Xingshuo Liu, Kangbo Lyu, Yang Zhou, Yuan Zhang, Tianwei Pan, Quan School of Cybersecurity Northwestern Polytechnical University Xi’an710129 China School of Automation Northwestern Polytechnical University Xi’an710129 China CFAR and IHPC Agency for Science Technology and Research Singapore College of Computing and Data Science Nanyang Technological University Singapore639798 Singapore School of Computer Science and Technology Zhejiang Sci-Tech University Zhejiang 310018 China

Unmanned aerial vehicles (UAVs) are increasingly employed to perform high-risk tasks that require minimal human intervention. However, they face escalating cybersecurity threats, particularly from GNSS spoofing attacks. While previous studies have extensively investigated the impacts of GNSS spoofing on UAVs, few have focused on its effects on specific tasks. Moreover, the influence of UAV motion states on the assessment of cybersecurity risks is often overlooked. To address these gaps, we first provide a detailed evaluation of how motion states affect the effectiveness of network attacks. We demonstrate that nonlinear motion states not only enhance the effectiveness of position spoofing in GNSS spoofing attacks but also reduce the probability of detecting speed-related attacks. Building upon this, we propose a state-triggered backdoor attack method (SSD) to deceive GNSS systems and assess its risk to trajectory planning tasks. Extensive validation of SSD’s effectiveness and stealthiness is conducted. Experimental results show that, with appropriately tuned hyperparameters, SSD significantly increases positioning errors and the risk of task failure, while maintaining high stealthy rates across three state-of-the-art detectors. © 2025, CC BY.

关键词： Risk assessment

Research on Domain Event Extraction Methods Based on Pre-Trained Model

学校读者我要写书评

暂无评论

Research on Domain Event Extraction Methods Based on Pre-Tra...

2022 International Conference on Signal Processing, Computer Networks, and Communications, SPCNC 2022

作者： Meng, Fanshen Zhang, Yonggang Liu, Xuhong Liu, Xiulei Yu, Junyang Institute of Data Science and Information Analysis Beijing Information Science & Technology University Beijing100029 China Center for Information Research Academy of Military Sciences Beijing100142 China Beijing Key Laboratory of Network Culture and Digital Communication Beijing Information Science & Technology University Beijing100029 China Beijing Advanced Innovation Center of Materials Genome Engineering Beijing Information Science & Technology University Beijing100029 China School of Software Henan University Henan 475001 China

ISBN: (纸本)9781510664616

One of the important research directions in information extraction is event extraction(EE). It aims at recognizing event types and event arguments from natural language texts, which is an important technical basis for artificial intelligence application that serves for the information work in business, science and technology, military and other fields. Currently, data annotation samples of the relevant field based on encyclopedia and news data are relatively rare and lack relevant datasets. Therefore, there are only a few public research on the event extraction in the relevant intelligence field. By integrating universal information extraction, the event extraction method which use the pre-trained model for the relevant intelligence field can handle the problem of rare data annotation samples for the event extraction in the relevant field. By expanding training samples automatically, the context information of encyclopedia and news data is learnt effectively to extract relevant events from encyclopedia and news data. Compared with other event extraction methods, using precision, recall and F1 value as assessment indicators, in event recognition tasks, the value of F1 enhances 1.26%, and in argument recognition tasks, the value of F1 enhances 1.58%. The method can significantly boost the extraction performance in small samples shown in the experimental results. © 2023 SPIE.

关键词： Information retrieval

Backdoor Attack with Sparse and Invisible Trigger

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Gao, Yinghua Li, Yiming Gong, Xueluan Li, Zhifeng Xia, Shu-Tao Wang, Qian The Tsinghua Shenzhen International Graduate School Tsinghua University Shenzhen518055 China The State Key Laboratory of Blockchain and Data Security Zhejiang University Hangzhou311200 China College of Computing and Data Science Nanyang Technological University 639798 Singapore The School of Computer Science Wuhan University China Tencent Data Platform Shenzhen518057 China The Research Center of Artificial Intelligence Peng Cheng Laboratory Shenzhen518000 China The School of Cyber Science and Engineering Wuhan University China

Deep neural networks (DNNs) are vulnerable to backdoor attacks, where the adversary manipulates a small portion of training data such that the victim model predicts normally on the benign samples but classifies the triggered samples as the target class. The backdoor attack is an emerging yet threatening training-phase threat, leading to serious risks in DNN-based applications. In this paper, we revisit the trigger patterns of existing backdoor attacks. We reveal that they are either visible or not sparse and therefore are not stealthy enough. More importantly, it is not feasible to simply combine existing methods to design an effective sparse and invisible backdoor attack. To address this problem, we formulate the trigger generation as a bi-level optimization problem with sparsity and invisibility constraints and propose an effective method to solve it. The proposed method is dubbed sparse and invisible backdoor attack (SIBA). We conduct extensive experiments on benchmark datasets under different settings, which verify the effectiveness of our attack and its resistance to existing backdoor defenses. The codes for reproducing main experiments are available at https://***/YinghuaGao/SIBA. © 2023, CC BY.

关键词： Deep neural networks

Impact of E-Waste on Pregnant Lady, Newborn Baby, and Fetus 4

学校读者我要写书评

暂无评论

Impact of E-Waste on Pregnant Lady, Newborn Baby, and Fetus

4th International Conference on Technological Advancements in Computational sciences, ICTACS 2024

作者： Tiwari, Ankita Babu, N. Ramesh Rakesh, Nitin Kumar, Mandeep Dhattarwal, Jagjit Singh Bhola, Abhishek College of Engineering Koneru Lakshmaih Education Foundation Department of Engineering Mathematics Guntur India School of Advnced Sciences Vellore Institute of Technology Department of Mathematucs Chennai India Department of Computer Science and Engineering Nagpur Campus Pune India School of Law Sharda University Greater Noida India Koneru Lakshmaiah Education Foundation Department of Artificial Intelligence & Data Science Andhra Pradesh Vaddeswaram India Chaudhary Charan Singh Haryana Agricultural University College of Agriculture Bawal India

ISBN: (纸本)9798350387490

Hazardous materials included in electronic trash (e-waste) can affect fetuses, neonates, and expectant mothers. Preterm birth, birth irregularities, and miscarriage are among the risks associated with pregnant women who are exposed to e-waste chemicals. Exposure to e-waste puts newborns at risk for respiratory and neurodevelopmental disorders. Developmental irregularities and growth restrictions are possible in fetuses. To reduce these risks and protect mother and child health, effective methods including appropriate e-waste discarding and public health programmes are important. © 2024 IEEE.

关键词： E-waste Environmental health Fetal development Neonatal health Neuro developmental disorders maternal pregnancy Public health policy recycling technologies Toxic chemicals

Continual Diffuser (CoD): Mastering Continual Offline Reinforcement Learning with Experience Rehearsal

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Hu, Jifeng Shen, Li Huang, Sili Yang, Zhejian Chen, Hechang Sun, Lichao Chang, Yi Tao, Dacheng School of Artificial Intelligence Jilin University Changchun China Sun Yat-sen University Zhongshan China Lehigh University BethlehemPA United States College of Computing and Data Science NTU Singapore

Artificial neural networks, especially recent diffusion-based models, have shown remarkable superiority in gaming, control, and QA systems, where the training tasks’ datasets are usually static. However, in real-world applications, such as robotic control of reinforcement learning (RL), the tasks are changing, and new tasks arise in a sequential order. This situation poses the new challenge of plasticity-stability trade-off for training an agent who can adapt to task changes and retain acquired knowledge. In view of this, we propose a rehearsal-based continual diffusion model, called Continual Diffuser (CoD), to endow the diffuser with the capabilities of quick adaptation (plasticity) and lasting retention (stability). Specifically, we first construct an offline benchmark that contains 90 tasks from multiple domains. Then, we train the CoD on each task with sequential modeling and conditional generation for making decisions. Next, we preserve a small portion of previous datasets as the rehearsal buffer and replay it to retain the acquired knowledge. Extensive experiments on a series of tasks show CoD can achieve a promising plasticity-stability trade-off and outperform existing diffusion-based methods and other representative baselines on most tasks. Source code is available at here. © 2024, CC BY.

关键词： Reinforcement learning