检索结果-内蒙古大学图书馆

41st International Conference on Machine Learning, ICML 2024

作者： Shen, Han Yang, Zhuoran Chen, Tianyi Department of Electrical Computer and Systems Engineering Rensselaer Polytechnic Institute United States Department of Statistics and Data Science Yale University United States

Bilevel optimization has been recently applied to many machine learning tasks. However, their applications have been restricted to the supervised learning setting, where static objective functions with benign structures are considered. But bilevel problems such as incentive design, inverse reinforcement learning (RL), and RL from human feedback (RLHF) are often modeled as dynamic objective functions that go beyond the simple static objective structures, which pose significant challenges of using existing bilevel solutions. To tackle this new class of bilevel problems, we introduce the first principled algorithmic framework for solving bilevel RL problems through the lens of penalty formulation. We provide theoretical studies of the problem landscape and its penalty-based (policy) gradient algorithms. We demonstrate the effectiveness of our algorithms via simulations in the Stackelberg game and RLHF. Copyright 2024 by the author(s)

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

Theoretical Insights for Diffusion Guidance: A Case Study for Gaussian Mixture Models 41

Theoretical Insights for Diffusion Guidance: A Case Study fo...

引用

41st International Conference on Machine Learning, ICML 2024

作者： Wu, Yuchen Chen, Minshuo Li, Zihao Wang, Mengdi Wei, Yuting Department of Statistics and Data Science The Wharton School University of Pennsylvania United States Department of Electrical and Computer Engineering Princeton University United States

Diffusion models benefit from instillation of task-specific information into the score function to steer the sample generation towards desired properties. Such information is coined as guidance. For example, in text-to-image synthesis, text input is encoded as guidance to generate semantically aligned images. Proper guidance inputs are closely tied to the performance of diffusion models. A common observation is that strong guidance promotes a tight alignment to the task-specific information, while reducing the diversity of the generated samples. In this paper, we provide the first theoretical study towards understanding the influence of guidance on diffusion models in the context of Gaussian mixture models. Under mild conditions, we prove that incorporating diffusion guidance not only boosts classification confidence but also diminishes distribution diversity, leading to a reduction in the differential entropy of the output distribution. Our analysis covers the widely adopted sampling schemes including those based on the SDE and ODE reverse processes, and leverages comparison inequalities for differential equations as well as the Fokker-Planck equation that characterizes the evolution of probability density function, which may be of independent theoretical interest. Copyright 2024 by the author(s)

关键词： Fokker Planck equation

来源：评论

学校读者我要写书评

暂无评论

Hierarchical Encoder-Decoder with Addressable Memory Network for Diagnosis Prediction 28th

Hierarchical Encoder-Decoder with Addressable Memory Networ...

引用

28th International Conference on database Systems for Advanced Applications, DASFAA 2023

作者： Wang, Mingxia Xiong, Yun Zhang, Yao Yu, Philip S. Zhu, Yangyong Shanghai Key Laboratory of Data Science School of Computer Science Fudan University Shanghai China Department of Computer Science University of Illinois Chicago Chicago United States

ISBN: (纸本)9783031306778

Deep learning methods have demonstrated success in diagnosis prediction on Electronic Health Records (EHRs). Early attempts utilize sequential models to encode patient historical records, but they lack the ability to identify critical diseases for patient health conditions. Besides, some works focus on the hierarchical structure of EHR data during learning patient representations, but neglect it in diagnosis prediction process, which leads to insufficient utilization of hierarchical information. To tackle these challenges, we propose a new Hierarchical Encoder-Decoder with Addressable Memory Network for Diagnosis Prediction named HAMNet. Specifically, we employ a hierarchical encoder-decoder framework to utilize the hierarchical structure of historical visits both during representation learning and diagnosis prediction. Furthermore, we propose the addressable memory network to distinguish core diseases towards patient health status at next visit through the well-designed addressing mechanism. It employs feed-forward layers to build the memory network, which can automatically update the memory module through gradient backpropagation without explicit read/write operations. Finally, we evaluate the performance of HAMNet on a large real-world EHR dataset MIMIC-III, and it achieves the state-of-the-art performance compared with baselines. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

M3Hop-CoT: Misogynous Meme Identification with Multimodal Multi-hop Chain-of-Thought

M3Hop-CoT: Misogynous Meme Identification with Multimodal Mu...

引用

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

作者： Kumari, Gitanjali Jain, Kirtan Ekbal, Asif Department of Computer Science and Engineering India Indian Institute of Technology Patna India School of AI and Data Science Indian Institute of Technology Jodhpur India

ISBN: (纸本)9798891761643

In recent years, there has been a significant rise in the phenomenon of hate against women on social media platforms, particularly through the use of misogynous memes. These memes often target women with subtle and obscure cues, making their detection a challenging task for automated systems. Recently, Large Language Models (LLMs) have shown promising results in reasoning using Chain-of-Thought (CoT) prompting to generate the intermediate reasoning chains as the rationale to facilitate multimodal tasks, but often neglect cultural diversity and key aspects like emotion and contextual knowledge hidden in the visual modalities. To address this gap, we introduce a Multimodal Multi-hop CoT (M3Hop-CoT) framework for Misogynous meme identification, combining a CLIP-based classifier and a multimodal CoT module with entity-object-relationship integration. M3Hop-CoT employs a three-step multimodal prompting principle to induce emotions, target awareness, and contextual knowledge for meme analysis. Our empirical evaluation, including both qualitative and quantitative analysis, validates the efficacy of the M3HopCoT framework on the SemEval-2022 Task 5 (MAMI task) dataset, highlighting its strong performance in the macro-F1 score. Furthermore, we evaluate the model's generalizability by evaluating it on various benchmark meme datasets, offering a thorough insight into the effectiveness of our approach across different datasets. © 2024 Association for Computational Linguistics.

关键词： Chains

来源：评论

学校读者我要写书评

暂无评论

Drawing Order Diagrams Through Two-Dimension Extension

引用

Journal of Graph Algorithms and Applications 2023年第9期27卷 783-802页

作者： Dürrschnabel, Dominik Hanika, Tom Stumme, Gerd Knowledge & Data Engineering Group Research Center for Information System Design & Department of Electrical Engineering and Computer Science University of Kassel Germany Institute of Computer Science University of Hildesheim Germany Berlin School of Library and Information Science Humboldt-Universität zu Berlin Germany

Ordinal real-world data such as concept hierarchies, ontologies, genealogies, or task dependencies in scheduling often has the property to not only contain pairwise comparable, but also incomparable elements. Order diagrams provide an important tool for understanding and thus discovering knowledge in such data. Easily readable drawings of such order diagrams are hard to come by, even for small ordered sets. Many attempts were made to transfer classical graph drawing approaches to order diagrams. Although these methods produce satisfying results for some ordered sets, they unfortunately perform poorly in general. In this work, we present the novel algorithm DimDraw to decompose an ordered set (e.g., a concept hierarchy) in linear orders and to produce a corresponding order diagram. This algorithm is based on a relation between the dimension of an ordered set and the bipartiteness of its transitive incompatibility graph. To evaluate the quality of the algorithm, a user study was conducted where generated drawings were compared with ones from state-of-the-art drawing algorithms. © 2023 Brown University. All rights reserved.

关键词： Set theory

来源：评论

学校读者我要写书评

暂无评论

The Significance of Utilizing the dependencies among Labels in Multi Label Classification 2

The Significance of Utilizing the dependencies among Labels ...

引用

2nd International Engineering Conference on Electrical, Energy, and Artificial Intelligence, EICEEAI 2023

作者： Alzoubi, Haneen Alzyoud, Mazen Alazaidah, Raed Al-Shanableh, Najah Khafajah, Hayel Almatarneh, Sattam Al-al-Bayu University Department of Computer Science Mafraq Jordan Ai Zarqa University Department of Data Science Zarqa Jordan

ISBN: (纸本)9798350373363

Multi-Label Classification (MLC) is a general type of classification that attracted scholars in the last few years. It imposes a high challenge since the problem search space of MLC is very large and follows an exponential function of growth. Moreover, the accuracy of classification in MLC is still very low when compared to the accuracy of Single Label Classification (SLC). Consequently, many scholars and researchers proposed to utilize and exploit the dependencies among class labels in to minimize the size of the problem search space of MLC, and hence, improve the accuracy of the classification task. Unfortunately, very few studies address this issue and attempt to discover the benefits of utilizing and exploiting the dependencies among labels in the domain of MLC. Therefore, this research attempts to identify the significance of discovering and utilizing these dependencies, with respect to three evaluation metrics designed specifically for MLC, and considering four different multi label datasets. The results revealed the clear significance of discovering and utilizing high order dependencies among labels in MLC, especially with high cardinality datasets. © 2023 IEEE.

关键词： Classification (of information)

来源：评论

学校读者我要写书评

暂无评论

Navigating Complex Multiclass Classification in High-Dimensional Spaces: A Hybrid Approach 7

Navigating Complex Multiclass Classification in High-Dimensi...

引用

7th IEEE International Conference on Computational Systems and Information Technology for Sustainable Solutions, CSITSS 2023

作者： Nemani, Praneeth Sundar Vadali, Venkata Surya Medi, Prathistith Raj Marisetty, Ashish Vollala, Satyanarayana Iiit Naya Raipur Department of Computer Science and Engineering India Iiit Naya Raipur Department of Data Science and Artificial Intelligence India

ISBN: (纸本)9798350343144

Large-scale tabular data classification is a critical task and the complexity arises from the vast amount of structured data generated in these fields, coupled with the challenges of high dimensionality and limited sample sizes. To address these challenges, advanced machine learning algorithms are required to analyze and categorize instances within these datasets effectively. In this work, we propose the usage of a state-of-the-art classifier called XBNet, which combines the strengths of tree-based classifiers and neural networks to tackle large tabular datasets. Our methodology is validated on a dataset with 64 dimensions and 11 classes, showcasing the model's capability to detect patterns and extract relevant features automatically. Furthermore, we employ K-Fold cross-validation to assess the model's performance, achieving an impressive training accuracy of 61.9% and a validation accuracy of 31.7%. These results surpass those of competing algorithms, confirming the superiority of our proposed methodology in handling large-scale tabular data classification tasks. © 2023 IEEE.

关键词： Classification (of information)

来源：评论

学校读者我要写书评

暂无评论

Soft-GNN:towards robust graph neural networks via self-adaptive data utilization

引用

Frontiers of computer science 2025年第4期19卷 1-12页

作者： Yao WU Hong HUANG Yu SONG Hai JIN National Engineering Research Center for Big Data Technology and System Service Computing Technology and System LabCluster and Grid Computing LabSchool of Computer Science and TechnologyHuazhong University of Science and TechnologyWuhan 430074China College of Information and Communication National University of Defense TechnologyWuhan 430019China Department of Computer Science and Operations Research Universitéde MontréalMontreal H3C 3J7Canada

Graph neural networks(GNNs)have gained traction and have been applied to various graph-based data analysis tasks due to their high ***,a major concern is their robustness,particularly when faced with graph data that has been deliberately or accidentally polluted with *** presents a challenge in learning robust GNNs under noisy *** address this issue,we propose a novel framework called Soft-GNN,which mitigates the influence of label noise by adapting the data utilized in *** approach employs a dynamic data utilization strategy that estimates adaptive weights based on prediction deviation,local deviation,and global *** better utilizing significant training samples and reducing the impact of label noise through dynamic data selection,GNNs are trained to be more *** evaluate the performance,robustness,generality,and complexity of our model on five real-world datasets,and our experimental results demonstrate the superiority of our approach over existing methods.

关键词： graph neural networks node classification label noise robustness

来源：评论

学校读者我要写书评

暂无评论

Advancing Anomaly Detection in Time Series data: A Knowledge Distillation Approach with LSTM Model

Advancing Anomaly Detection in Time Series Data: A Knowledge...

引用

2023 Innovations in Intelligent Systems and Applications Conference, ASYU 2023

作者： Kilinc, Sena Camlidere, Bora Yildiz, Ekrem Guler, Ali Kerem Alsan, Huseyin Fuat Arsan, Taner Kadir Has University Department of Computer Engineering Istanbul Turkey Turknet Department of Data Science Istanbul Turkey

ISBN: (纸本)9798350306590

This paper focuses on enhancing anomaly detection in time series data using deep learning techniques. Particularly, it investigates the integration of knowledge distillation with LSTM-based models for improved precision, efficiency, and interpretability. The study outlines objectives such as dataset preprocessing, developing a novel LSTM-knowledge distillation framework, incorporating Grafana, InfluxDB, Flask API with Docker, performance assessment, and practical implications. Results highlight the efficacy of knowledge distillation in enhancing student model performance. The proposed approach enhances anomaly detection, offering a viable solution for real-world applications. © 2023 IEEE.

关键词： Distillation

来源：评论

学校读者我要写书评

暂无评论

Network Traffic Anomaly Detection Using Quantile Regression with Tolerance

Network Traffic Anomaly Detection Using Quantile Regression ...

引用

2023 IEEE International Black Sea Conference on Communications and Networking, BlackSeaCom 2023

作者： Alsan, Huseyin Fuat Guler, Ali Kerem Yildiz, Ekrem Kilinc, Sena Camlidere, Bora Arsan, Taner Kadir Has University Department of Computer Engineering Istanbul Turkey Turknet Department of Data Science Istanbul Turkey

ISBN: (纸本)9798350337822

Network traffic anomaly detection describes a time series anomaly detection problem where a sudden increase or decrease (called spikes) in network traffic is predicted. data is modeled with the trend and heteroscedastic noise component. Traditional autoregressive models struggle to capture data changes effectively, making anomaly detection difficult. Our approach is to generate upper and lower limits by using quantile regression. We use a deep learning based multilayer perceptron model to predict five data quantiles 1, 25, 50, 75, and 99. The upper and lower limits are calculated as differences between the quantile-1 and quantile-99. Any data that is outside these limits are considered as an anomaly. We also add tolerance to these limits to add flexibility to anomaly detection. Anomalies and non-anomalies are labeled to get a binary classification task. Anomaly detection is class imbalanced by nature;therefore, precision, recall, and F-1 score are computed to evaluate the proposed anomaly detection method. We conclude that choosing tolerance is a tradeoff between false alarms and missing anomaly detections. © 2023 IEEE.

关键词： Time series

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：