检索结果-内蒙古大学图书馆

Multitask Transformer for Cross-Corpus Speech Emotion Recognition

IEEE Transactions on Affective computing 2025年

作者： Ahn, Chung-Soo Rana, Rajib Busso, Carlos Rajapakse, Jagath C. College of Computing and Data Science Singapore University of Southern Queensland Australia Carnegie Mellon University Language Technologies Institute School of Computer Science PittsburghPA15213 United States

Deep learning has significantly advanced the field of Speech Emotion Recognition (SER), yet its efficacy in cross-corpus scenarios remains a challenge. To overcome this limitation, recent studies demonstrate the success of multitask learning, which uses auxiliary tasks to reduce difference between source and target dataset (or transfer knowledge from source to target datasets). Despite the efforts, the overall accuracy for cross-corpus SER is still relatively low and needs attention. To improve performance, we propose a multitask framework with SER as the primary task and contrastive learning and information maximization as auxiliary tasks. We design the auxiliary tasks innovatively to use the target data without emotional labels to develop a better understanding of the target data. The core of our multitask framework is a pre-Trained transformer. While transformers have gained attention in SER, their application to cross-corpus scenarios is still limited. Multimodal approaches for cross-corpus scenario is substantially limited as well. We use text as the second modality, developing separate multitask transformers for audio and text and conduct decision-level fusion during inference. We use publicly available and widely used speech corpora, including the IEMOCAP, MSP-IMPROV and EMO-DB databases. The results demonstrate the benefits of the proposed approach, achieving improved performance on the benchmark databases in cross-corpus settings. © 2010-2012 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Knowledge-Enhanced Self-Supervised Prototypical Network for Few-Shot Event Detection

Knowledge-Enhanced Self-Supervised Prototypical Network for ...

引用

2022 Findings of the Association for Computational Linguistics: EMNLP 2022

作者： Zhao, Kailin Jin, Xiaolong Bai, Long Guo, Jiafeng Cheng, Xueqi CAS Key Laboratory of Network Data Science and Technology Institute of Computing Technology Chinese Academy of Sciences School of Computer Science and Technology University of Chinese Academy of Sciences China

Prototypical network based joint methods have attracted much attention in few-shot event detection, which carry out event detection in a unified sequence tagging framework. However, these methods suffer from the inaccurate prototype representation problem, due to two main reasons: the number of instances for calculating prototypes is limited;And, they do not well capture the relationships among event prototypes. To deal with this problem, we propose a Knowledge-Enhanced self-supervised Prototypical Network, called KE-PN, for few-shot event detection. KE-PN adopts hybrid rules, which can automatically align event types to an external knowledge base, i.e., FrameNet, to obtain more instances. It proposes a self-supervised learning method to filter out noisy data from enhanced instances. KE-PN is further equipped with an auxiliary event type relationship classification module, which injects the relationship information into representations of event prototypes. Extensive experiments on three benchmark datasets, i.e., FewEvent, MAVEN, and ACE2005 demonstrate the state-of-the-art performance of KE-PN. © 2022 Association for Computational Linguistics.

关键词： Classification (of information)

来源：评论

学校读者我要写书评

暂无评论

Probabilistic Reduced-Dimensional Vector Autoregressive Modeling for Dynamics Prediction and Reconstruction with Oblique Projections

arXiv

引用

arXiv 2023年

作者： Mo, Yanfang Yu, Jiaxin Qin, S. Joe The Hong Kong Institute for Data Science City University of Hong Kong Hong Kong The School of Data Science Hong Kong Institute for Data Science City University of Hong Kong Hong Kong The Department of Computing and Decision Science Institute of Data Science Lingnan University Hong Kong

In this paper, we propose a probabilistic reduced-dimensional vector autoregressive (PredVAR) model with oblique projections. This model partitions the measurement space into a dynamic subspace and a static subspace that do not need to be orthogonal. The partition allows us to apply an oblique projection to extract dynamic latent variables (DLVs) from high-dimensional data with maximized predictability. We develop an alternating iterative PredVAR algorithm that exploits the interaction between updating the latent VAR dynamics and estimating the oblique projection, using expectation maximization (EM) and a statistical constraint. In addition, the noise covariance matrices are estimated as a natural outcome of the EM method. A simulation case study of the nonlinear Lorenz oscillation system illustrates the advantages of the proposed approach over two alternatives. Copyright © 2023, The Authors. All rights reserved.

关键词： Iterative methods

来源：评论

学校读者我要写书评

暂无评论

Disentangling linear quadratic control with untrusted ML predictions 24

Disentangling linear quadratic control with untrusted ML pre...

引用

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

作者： Tongxin Li Hao Liu Yisong Yue School of Data Science The Chinese University of Hong Kong Shenzhen China Computing + Mathematical Sciences California Institute of Technology Pasadena

ISBN: (纸本)9798331314385

Uncertain perturbations in dynamical systems often arise from diverse resources, represented by latent components. The predictions for these components, typically generated by "black-box" machine learning tools, are prone to inaccuracies. To tackle this challenge, we introduce DISC, a novel policy that learns a confidence parameter online to harness the potential of accurate predictions while also mitigating the impact of erroneous forecasts. When predictions are precise, DISC leverages this information to achieve near-optimal performance. Conversely, in the case of significant prediction errors, it still has a worst-case competitive ratio guarantee. We provide competitive ratio bounds for DISC under both linear mixing of latent variables as well as a broader class of mixing functions. Our results highlight a first-of-its-kind "best-of-both-worlds" integration of machine-learned predictions, thus lead to a near-optimal consistency and robustness tradeoff, which provably improves what can be obtained without learning the confidence parameter. We validate the applicability of DISC across a spectrum of practical scenarios.

关键词：

来源：评论

学校读者我要写书评

暂无评论

CiRLExplainer: Causality-Inspired Explainer for Graph Neural Networks via Reinforcement Learning

引用

IEEE Transactions on Neural Networks and Learning Systems 2025年第6期36卷 9970-9984页

作者： Wenya Hu Jia Wu Quan Qian School of Computer Engineering and Science Shanghai University Shanghai China School of Computing Macquarie University Sydney NSW Australia School of Computer Engineering and Science the Center of Materials Informatics and Data Science Materials Genome Institute the Key Laboratory of Silicate Cultural Relics Conservation Ministry of Education and Shanghai Institute for Advanced Communication and Data Science Shanghai University Shanghai China

In this article, we propose a new graph neural network (GNN) explainability model, CiRLExplainer, which elucidates GNN predictions from a causal attribution perspective. Initially, a causal graph is constructed to analyze the causal relationships between the graph structure and GNN predicted values, identifying node attributes as confounding factors between the two. Subsequently, a backdoor adjustment strategy is employed to circumvent these confounders. Additionally, since the edges within the graph structure are not independent, reinforcement learning is incorporated. Through a sequential selection process, each step evaluates the combined effects of an edge and the previous structure to generate an explanatory subgraph. Specifically, a policy network predicts the probability of each candidate edge being selected and adds a new edge through sampling. The causal effect of this action is quantified as a reward, reflecting the interactivity among edges. By maximizing the policy gradient during training, the reward stream of the edge sequence is optimized. The CiRLExplainer is versatile and can be applied to any GNN model. A series of experiments was conducted, including accuracy (ACC) analysis of the explanation results, visualization of the explanatory subgraph, and ablation studies considering node attributes as confounding factors. The experimental results demonstrate that our model not only outperforms current state-of-the-art explanation techniques, but also provides precise semantic explanations from a causal perspective. Additionally, the experiments validate the rationale for considering node attributes as confounding factors, thereby enhancing the explanatory power and ACC of the model. Notably, across different datasets, our explainer achieved improvements over the best baseline models in the ACC-area under the curve (AUC) metrics by 5.89%, 5.69%, and 4.87%, respectively.

关键词： Predictive models Graph neural networks Reinforcement learning Accuracy Training Prediction algorithms Analytical models Sustainable development Focusing data science

来源：评论

学校读者我要写书评

暂无评论

A Survey on Privacy-Preserving data Publishing Methods and Models in Relational Electronic Health Records 1

引用

3rd International Conference on Sustainable Advanced computing, ICSAC 2021

作者： Jayapradha, J. Prakash, M. Department of Computing Technologies SRM Institute of Science and Technology Tamil Nadu Kattankulathur Chengalpattu603203 India Department of Data Science and Business Systems SRM Institute of Science and Technology Tamil Nadu Kattankulathur Chengalpattu603203 India

ISBN: (数字)9789811690129

ISBN: (纸本)9789811690112

The collections of various digital information from corporations, public sectors, private sectors, and individuals have formed remarkable opportunities for analysis and decision-making. The electronic health record is highly constructive for various study purposes;however, it should be carried in a way that preserves the identity and privacy of the patients. Due to the digital transformation, the flow of data increases that lead to the need for big data. Protecting privacy cannot be straightforward as there are several privacy threats such as membership disclosure, identity disclosure, and attribute disclosure. The privacy methods such as generalization, suppression, pseudonymization, bucketization, slicing are illustrated with examples. The study focused on various privacy models such as k-anonymity, l-diversity, t-closeness, differential privacy, and cryptographic technique. The models are discussed with the sample dataset and workflow. The paper has discussed the security and privacy concerns on data publishing, the need for big data scope, an overview of threats, samples, privacy methods, models, and challenges in privacy-preserving data publishing. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Records management

来源：评论

学校读者我要写书评

暂无评论

HealthGuardian: Artificial Intelligence System for Early Diagnosis and Clinical Support

HealthGuardian: Artificial Intelligence System for Early Dia...

引用

International Conference on Computer and Communications Technologies (ICCCT)

作者： V Radhamani D. Sudha Devi E Chandra Blessie P J Isabella Priya R Tanushree Department of Computing (Decision and Computing Sciences) Coimbatore Institute of Technology Coimbatore Tamil Nadu India Department of Computing (Data Science) Coimbatore Institute of Technology Coimbatore Tamil Nadu India Department of Computing (Artificial Intelligence and Machine Learning) Coimbatore Institute of Technology Coimbatore Tamil Nadu India

ISBN: (数字)9798331537579

ISBN: (纸本)9798331537586

In today's fast-paced world, access to timely and accurate healthcare information is critical, yet many individuals face challenges in assessing their symptoms, finding appropriate care, and understanding potential treatments. To address these issues, we propose HealthGuardian, an AI-powered clinical decision support system designed to assist users with symptom checking, personalized treatment recommendations, healthcare provider suggestions, online consultations, and early-stage diabetes prediction. HealthGuardian allows users to input their symptoms through natural language processing (NLP) and a recur-rent neural network (RNN) model, providing possible diagnoses and advising when to seek medical attention. The system offers personalized treatment plans based on medical history, suggests nearby healthcare providers using the Overpass API, and facilitates online consultations with doctors. Additionally, it features a predictive tool for early-stage diabetes risk assessment using various machine learning algorithms. The system leverages machine learning algorithms such as logistic regression, decision tree classifier, Naive Bayes, and random forest classifier to enhance diagnostic accuracy and prediction reliability. Our results demonstrate that HealthGuardian improves early diagnosis accuracy, provides actionable treatment guidance, and increases healthcare accessibility, helping users make informed decisions about their health.

关键词： Logistic regression Accuracy Machine learning algorithms Neural networks Medical services Diabetes Reliability Risk management Medical diagnostic imaging Random forests

来源：评论

学校读者我要写书评

暂无评论

Move Match: Live Dance Motion Monitoring and Feedback System

Move Match: Live Dance Motion Monitoring and Feedback System

引用

Computational, Communication and Information Technology (ICCCIT), International Conference on

作者： G Baby Saral Subasri. S Ramya Shree R K. Regin Bose Dept. Artificial Intelligence & Data Science Rajalakshmi Institute of Technology Chennai India Artificial Intelligence and Data Science Rajalakshmi Institute of Technology Chennai India Artificial Intelligence and Data Science Student Rajalakshmi Institute of Technology Chennai India Dept. AI&DS & Cloud Computing Rajalakshmi Institute of Technology Chennai India

ISBN: (数字)9798331512965

ISBN: (纸本)9798331512972

Dance movement accuracy prediction is a critical component of modern dance training, yet traditional methods often lack objectivity due to the subjective nature of evaluations and the complexity of human movement. To address these challenges this paper introduces MoveMatch, which is a web-based platform built for real-time dance movement prediction. By leveraging deep learning techniques, the platform performs posture estimation in real-time and detects key points. The system compares the detected poses using Euclidean distance to calculate the alignment between corresponding key points which provides a quantitative measure of the user's accuracy. Based on this comparison, MoveMatch delivers real-time feedback, offering an accuracy score and personalized suggestions for improvement. This immediate feedback fosters a more efficient and engaging learning experience for dancers. Extensive testing of MoveMatch in various dance styles demonstrates its effectiveness in objectively assessing performance, enhancing both training quality and accessibility. The integration of pose estimation, deep learning and real-time comparison offers a novel, user-friendly solution for dancers seeking to improve their skills, transforming traditional dance education with AIdriven insights.

关键词： Training Deep learning Humanities Accuracy Three-dimensional displays Tracking Pose estimation Euclidean distance Streaming media Real-time systems

来源：评论

学校读者我要写书评

暂无评论

A New Interactive Gaming Approach for Enhancing Math Learning Among Marginalized Communities 1

A New Interactive Gaming Approach for Enhancing Math Learnin...

引用

1st IEEE International Conference on Smart Technology, ICE-SMARTec 2023

作者： Awang, Hapini Mansor, Nur Suhaili Harun, Nor Hazlyna Bakar, Juhaida Abu Al-Mashhadani, Abdulrazak F. Shahatha Hassan, Mohamad Ghozali Institute for Advanced and Smart Digital Opportunities School of Computing Universiti Utara Malaysia Sintok Malaysia School of Computing Universiti Utara Malaysia Data Science Research Lab Sintok Malaysia Sohar University Faculty of Business Sohar Oman Disaster Management Institute School of Technology Management and Logistics Universiti Utara Malaysia Sintok Malaysia

ISBN: (纸本)9798350306156

Mathematics is often considered a boring and challenging subject, which makes it difficult for educators to attract marginalized children to learn and embrace it. Therefore, this article discusses the development and evaluation of a mobile game application called Math-Hero, designed to enhance marginalized children's mathematical skills through an interactive, engaging, and heartfelt learning environment. To examine the effectiveness of Math-Hero, an online survey method was employed to collect data from 32 respondents. The evaluation model was developed following the Technology Acceptance Model (TAM). Three constructs - Perceived Usefulness (PU), Perceived Ease of Use (PEOU), and Intention to Adopt (ITA) - were examined quantitatively using descriptive statistics and correlation analysis. Additionally, qualitative data were analyzed using thematic and sentiment analyses. The results of quantitative analyses indicate that Math-Hero is easy to use, beneficial, and the users intend to use it in the future. Its ease of use has a positive correlation with usefulness and intention to adopt it. The sentiment analysis also supports the findings, as most of the users have a positive and neutral perception of Math-Hero. Although Math-Hero has shown promise in improving marginalized children's mathematical skills, there is still room for improvement to make it even more robust and effective in the future. Moreover, further research is necessary to empirically demonstrate the effectiveness of Math-Hero among diverse marginalized communities, beyond just Semai youngsters. © 2023 IEEE.

关键词： Sentiment analysis

来源：评论

学校读者我要写书评

暂无评论

CausalDiff: causality-inspired disentanglement via diffusion model for adversarial defense 24

CausalDiff: causality-inspired disentanglement via diffusion...

引用

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

作者： Mingkun Zhang Keping Bi Wei Chen Quanrun Chen Jiafeng Guo Xueqi Cheng CAS Key Laboratory of AI Safety Institute of Computing Technology CAS Key Laboratory of Network Data Science and Technology Institute of Computing Technology CAS School of Statistics University of International Business and Economics

ISBN: (纸本)9798331314385

Despite ongoing efforts to defend neural classifiers from adversarial attacks, they remain vulnerable, especially to unseen attacks. In contrast, humans are difficult to be cheated by subtle manipulations, since we make judgments only based on essential factors. Inspired by this observation, we attempt to model label generation with essential label-causative factors and incorporate label-non-causative factors to assist data generation. For an adversarial example, we aim to discriminate the perturbations as non-causative factors and make predictions only based on the label-causative factors. Concretely, we propose a casual diffusion model (CausalDiff) that adapts diffusion models for conditional data generation and disentangles the two types of casual factors by learning towards a novel casual information bottleneck objective. Empirically, CausalDiff has significantly outperformed state-of-the-art defense methods on various unseen attacks, achieving an average robustness of 86.39% (+4.01%) on CIFAR-10, 56.25% (+3.13%) on CIFAR-100, and 82.62% (+4.93%) on GTSRB (German Traffic Sign Recognition Benchmark). The code is available at https://***/CAS-AISafetyBasicResearchGroup/CausalDiff.

关键词：

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：