检索结果-内蒙古大学图书馆

41st International Conference on Machine Learning, ICML 2024

作者： Wei, Yongxian Hu, Zixuan Shen, Li Wang, Zhenyi Li, Yu Yuan, Chun Tao, Dacheng Tsinghua Shenzhen International Graduate School Tsinghua University China College of Computing & Data Science Nanyang Technological University Singapore School of Cyber Science and Technology Sun Yatsen University China University of Maryland College Park United States Department of Computer Science and Engineering The Chinese University of Hong Kong Hong Kong

Data-Free Meta-Learning (DFML) aims to derive knowledge from a collection of pre-trained models without accessing their original data, enabling the rapid adaptation to new unseen tasks. Current methods often overlook the heterogeneity among pre-trained models, which leads to performance degradation due to task conflicts. In this paper, we empirically and theoretically identify and analyze the model heterogeneity in DFML. We find that model heterogeneity introduces a heterogeneity-homogeneity trade-off, where homogeneous models reduce task conflicts but also increase the overfitting risk. Balancing this trade-off is crucial for learning shared representations across tasks. Based on our findings, we propose Task Groupings Regularization, a novel approach that benefits from model heterogeneity by grouping and aligning conflicting tasks. Specifically, we embed pre-trained models into a task space to compute dissimilarity, and group heterogeneous models together based on this measure. Then, we introduce implicit gradient regularization within each group to mitigate potential conflicts. By encouraging a gradient direction suitable for all tasks, the meta-model captures shared representations that generalize across tasks. Comprehensive experiments showcase the superiority of our approach in multiple benchmarks, effectively tackling the model heterogeneity in challenging multi-domain and multi-architecture scenarios. Copyright 2024 by the author(s)

关键词：

来源：评论

学校读者我要写书评

暂无评论

Hyperautomation for Air Quality Evaluations: A Perspective of Evidential Three-way Decision-making

引用

Cognitive Computation 2024年第5期16卷 2437-2453页

作者： Ding, Juanjuan Zhang, Chao Li, Deyu Sangaiah, Arun Kumar School of Computer and Information Technology Shanxi University Shanxi Taiyuan030006 China Key Laboratory of Computational Intelligence and Chinese Information Processing of Ministry of Education Shanxi University Shanxi Taiyuan030006 China International Graduate Institute of Artificial Intelligence National Yunlin University of Science and Technology Douliou64002 Taiwan Department of Electrical and Computer Engineering Lebanese American University Byblos Lebanon

Hyperautomation acts as a real digital transformation with the support of several cutting-edge cognitive computation methods that include robotic process automation, natural language processing, artificial intelligence, and other emerging ones, which is conducive to processing complex industrial processes via extending the range of various data-driven cognitive decision-making algorithms. The study of air quality evaluations (AQE) plays a significant role in ensuring healthy atmospheric environments. In view of the objective existence of uncertainties, AQE can be modeled and addressed by a typical data-driven automated decision-making problem, and hyperautomation can provide a reasonable solution via associating with a variety of techniques. This article explores hyperautomation for AQE via evidential three-way large-scale group decision-making (LSGDM) in an intuitionistic fuzzy (IF) setting. First, the notion of intuitionistic fuzzy sets (IFSs) is incorporated into the paradigm of multi-granularity (MG) three-way decisions (TWD), and the model of adjustable MG IF probabilistic rough sets (PRSs) is developed. Second, an IF clustering analysis with the improved technique for order preference by similarity to ideal solution (TOPSIS) method is conducted to affirm representative members within a decision group. Third, a novel IF LSGDM method is built via adjustable MG IF PRSs and the evidence reasoning (ER) method. Finally, a case study in the setting of AQE is studied by using the presented evidential three-way LSGDM method. Corresponding experimental analyses are carried out for illustrating the efficiency of hyperautomation for AQE. In general, the proposed method improves the performance of information fusion by virtue of adjustable MG IF PRSs, and the TOPSIS method avoids the influence of subjective factors on decision results. Meanwhile, the evaluation information of decision-makers (DMs) is fully analyzed by means of the ER method, which can provide more explaina

关键词： Decision making

来源：评论

学校读者我要写书评

暂无评论

Investigating the Impact of Data Augmentation for Fine-Grained Grocery Product Classification Based on Vision Transformer

Investigating the Impact of Data Augmentation for Fine-Grain...

引用

2023 International Conference on Artificial Intelligence Robotics, Signal and Image Processing, AIRoSIP 2023

作者： Rahmania, Rissa Heryadi, Yaya Lukas, Lukas Suparta, Wayan Sonata, Ilvico Bina Nusantara University Binus Graduate Program Computer Science Computer Science Department Jakarta11480 Indonesia Faculty of Engineering Jakarta12930 Indonesia Institut Teknologi Nasional Yogyakarta Faculty of Industrial Technology Department of Electrical Engineering Yogyakarta55281 Indonesia

ISBN: (纸本)9798350327472

The existence of fine-grained image classification supporting smart retail provides effectiveness in recognizing products with high similarity. However, the generic classification method performs poorly in identifying products from a subordinate category. This paper aims to identify augmentation techniques to leverage the Vision Transformer (ViT) model to classify the fine-grained grocery product, which involves embedding patches and transformer encoders to extract the main features. First, we develop a fine-grained image dataset with ColorJitter, CutOut, and combining both augmentations. Secondly, we perform experiments and analysis of ViT size, embedded patch size and image size in the patch embedding process. Lastly, the ViT model are evaluated according to the image sizes 224, 384, and 512 in accuracy, loss, and confusion matrix. The highest accuracy was obtained at 0.9922. The ColorJitter and CutOut improved the confusion matrix in ViT-B/16 and ViT-L/16 with an image size of 384 and 512. The results show that both augmentations in the ViT model are able to distinguish fine-grained grocery products. © 2023 IEEE.

关键词： Embeddings

来源：评论

学校读者我要写书评

暂无评论

Rethinking Rule-Based Approaches in Session-Based Recommendation 48

Rethinking Rule-Based Approaches in Session-Based Recommenda...

引用

48th IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2023

作者： Wang, Liuyin Li, Mingchao Zheng, Hai-Tao Tsinghua University Department of Computer Science and Technology Beijing China Pengcheng Laboratory Shenzhen China Tsinghua University Shenzhen International Graduate School China

ISBN: (纸本)9781728163277

Existing approaches to session-based recommendation are focused on using advanced deep neural networks (DNNs). Recent studies have shown that some traditional methods can outperform some DNN-based models. However, in recent years, few studies have tried to build traditional models. In this paper, we investigate this issue and propose a concise rule-based method for session-based recommendation. Specifically, we make item adjacent and N-gram-based rules to extract frequenting, sequential and other patterns of the limited historical information to construct item correlation dictionaries. Then, we exploit these dictionaries at the inference stage by constructing candidate item generation and item set fusion rules to acquire candidate items. By this means, we can leverage short-term user-item interaction records to generate candidate items for anonymous sessions. Finally, we sort the candidate items to make a recommendation. Extensive experimental results on three real public datasets show that our method can significantly outperform existing traditional methods and even outperforms several representative DNN-based models. The model just costs a few seconds to train and inference, and occupies less memory space than that. Compared with the previous models, it has a huge lead in time and space. © 2023 IEEE.

关键词： co-occurrence dictionary construction rule-based approaches session-based recommendation

来源：评论

学校读者我要写书评

暂无评论

Applying Burn-in Strategy to Deep Reinforcement Learning with Spiking Neural Networks

Applying Burn-in Strategy to Deep Reinforcement Learning wit...

引用

International Symposium on Soft Computing and Intelligent Systems (SCIS)

作者： Takahiro Iwata Seankein Yoshioka Daisuke Miki Graduate School of Computer Science Chiba Institute of Technology Chiba Japan Department of Computer Science Chiba Institute of Technology Chiba Japan

ISBN: (数字)9798350373332

ISBN: (纸本)9798350373349

In recent years, neural networks in deep reinforce-ment learning (D RL) have achieved success in various tasks, owing to their high expressiveness and flexibility. Neural networks are expected to be increasingly used for robot control. However, large neural networks consume a significant amount of power during training and operation. Given that the power supply that can be installed in robots is limited, challenges remain in the practical application of robot control using DRL, which includes large-scale neural networks. Conversely, spiking neural networks (SNNs) have high power efficiency, which is achieved through dedicated computational elements. Previous studies reported that SNNs can be applied to DRL to learn a robot's walking motion in a virtual environment and reduce energy consumption. Although previous studies have applied SNNs to DRL, they have the drawback of recursive computation, causing output delays. In this study, we propose a method that combines SNNs with a structure that addresses the issues of previous studies using a twin-delayed deep deterministic policy gradient (TD3). In addition, a burn-in strategy was introduced to estimate the action values using the internal states of the SNNs. Experiments in a virtual environment showed that the proposed method achieved higher rewards than the previous methods.

关键词： Legged locomotion Training Power supplies Neural networks Robot control Virtual environments Membrane potentials Spiking neural networks Robustness Delays

来源：评论

学校读者我要写书评

暂无评论

Exploring Shifted Domain Problem within MNIST Dataset

Exploring Shifted Domain Problem within MNIST Dataset

引用

2023 International Conference on computer science, Information technology and Engineering, ICCoSITE 2023

作者： Hidayaturrahman Heryadi, Yaya Lukas Suparta, Wayan Arifin, Yulyani Bina Nusatara University BINUS Graduate Program - Computer Science Computer Science Department Jakarta11480 Indonesia Faculty of Engineering Jakarta12930 Indonesia Institut Teknologi Nasional Yogyakarta Faculty of Industrial Technology Department of Electrical Engineering Yogyakarta55281 Indonesia

ISBN: (纸本)9798350320954

Domain data can be shifted in any direction so it will be shared in different distributions to its original domain. This could be a problem since the model was trained with different distributions. It is found that adversarial domain adaptation using domain adversarial neural networks (DANN) can help to solve this problem on some scale. DANN can minimize the discrepancy between source and target data so the model can work well in both domains. The experiment is done by utilizing MNIST dataset that shifted into some conditions. In a condition when the shifting of distribution is too far, DANN is struggling to maintain the knowledge extracted from source data which leads to underperformance in the source and target domain. In contrast, when the shifting is closer, DANN can easily fit the model so it can perform well in both domains. It proves DANN is one of the good approaches to performing domain adaptation in small discrepancies. © 2023 IEEE.

关键词： Adversarial Domain Adaptation Domain Shifting Problem MNIST Transfer Learning

来源：评论

学校读者我要写书评

暂无评论

Production and characterization of polymeric nanocomposite foams based on starch and cellulose nanofibers from oil palm mesocarp fiber

引用

Applied Nanoscience (Switzerland) 2025年第2期15卷 1-11页

作者： Souza, Mariana Costa Gonçalves, Ana Paula Bispo da Silva Oliveira, Emanoel Igor dos Santos, Juscivaldo Passos Lazzari, Lídia Kunz Zattera, Ademir José José, Nadia Mamede Chemical Engineering Graduate Program Polytechnic School Federal University of Bahia Bahia Salvador Brazil Department of Education Federal Institute of Education Science and Technology Feira de Santana Bahia Brazil Postgraduate Program in Process Engineering and Technologies University of Caxias Do Sul Rio Grande Do Sul Caxias Do Sul Brazil

The demand for sustainable and eco-friendly materials has promoted studies over the years to explore different polymeric materials that meet requirements such as biodegradability and sustainability. In this context, biopolymer materials based on cellulose nanoparticles and starch from different botanical sources have been investigated, aiming to achieve satisfactory performance. The present study aims to develop and characterize cassava starch foams based on density, scanning electron microscopy (SEM), and three-point flexure tests, and to evaluate the effect of cellulose nanofibers obtained from palm mesocarp fibers using the ultrafine friction grinding method after different processing times, characterized by X-ray diffractometry (XRD), microscopy, and chemical composition, as a reinforcing filler. The chemical composition and scanning electron microscopy revealed the effectiveness of the cellulose isolation process, with the analysis of the chemical composition revealing a cellulose content of 55.70% in the fibers after pulping and bleaching, in addition to changes in the visual characteristics of the material after the process aimed at isolating the cellulose. After grinding, XRD showed an increase in crystallinity (76.1% at the maximum grinding time), along with typical microscopy images of cellulose nanofibers. The analysis of the obtained nanocomposites provided insights into the role of these nanostructures in the thermo-expanded starch matrix, indicating that the nanofibers promoted changes such as an increase in mechanical properties and crystallinity, which contributed to improving overall mechanical performance. A 67.48% increase in flexural strength was achieved for the formulation with cellulose nanofibers that underwent 150 min of grinding, without causing major variations in density. © King Abdulaziz City for science and technology 2025.

关键词： Starch

来源：评论

学校读者我要写书评

暂无评论

Benchmarking Performance of Document Level Classification and Topic Modeling

引用

computers, Materials & Continua 2022年第4期71卷 125-141页

作者： Muhammad Shahid Bhatti Azmat Ullah Rohaya Latip Abid Sohail Anum Riaz Rohail Hassan Department of Computer Science COMSATS University IslamabadLahorePakistan Department of Communication Technology and Network Faculty of Computer ScienceUniversiti Putra MalaysiaSelangor43400Malaysia Othman Yeop Abdullah Graduate School of Business(OYAGSB) Universiti Utara Malaysia(UUM)Kuala Lumpur50300Malaysia

Text classification of low resource language is always a trivial and challenging *** paper discusses the process of Urdu news classification and Urdu documents *** is one of the most famous spoken languages in *** implementation of computational methodologies for text classification has increased over ***,Urdu language has not much experimented with research,it does not have readily available datasets,which turn out to be the primary reason behind limited research and applying the latest methodologies to the *** overcome these obstacles,a mediumsized dataset having six categories is collected from authentic Pakistani news *** is a rich but complex *** processing can be challenging for Urdu due to its complex features as compared to other *** frequency-inverse document frequency(TFIDF)based term weighting scheme for extracting features,chi-2 for selecting essential features,and Linear discriminant analysis(LDA)for dimensionality reduction have been *** matrix and cosine similarity measure have been used to identify similar documents in a collection and find the semantic meaning of words in a document FastText model has been *** training-test split evaluation methodology is used for this experimentation,which includes 70%for training data and 30%for testing ***-of-the-art machine learning and deep dense neural network approaches for Urdu news classification have been ***,we trained Multinomial Naïve Bayes,XGBoost,Bagging,and Deep dense neural *** and deep dense neural network outperformed the other *** experimental results show that deep dense achieves 92.0%mean f1 score,and Bagging 95.0%f1 score.

关键词： Deep neural network machine learning natural language processing TFIDF sparse matrix cosine similarity classification linear discriminant analysis gradient boosting

来源：评论

学校读者我要写书评

暂无评论

Framework for Integration of Generative AI into Metaverse Asset Creation

Framework for Integration of Generative AI into Metaverse As...

引用

Intelligent Metaverse Technologies & Applications (iMETA), International Conference on

作者： Rio Mukhtarom Paweroi Mario Köppen Department of Computer Science and Networks Graduate School of Computer Science and Systems Engineering Kyushu Institute of Technology Iizuka Japan

ISBN: (数字)9798350351514

ISBN: (纸本)9798350351521

Metaverse, a virtual world, is developing rapidly and is widely used in multi-sector. The number of users is projected to increase year over year. Due to the development of the metaverse platform, digital asset creation is demanding. Creating high-quality and diverse 3D digital objects is challenging. This study proposes the frameworks for integrating generative AI to create diverse 3D assets into the metaverse. We study different approaches for asset creation, i.e., generative 3D model-based, generative image projection-based, and generative language script-based. Creators can use this workflow to optimize the creation of 3D assets. Moreover, this study compares the results of generative AI and procedural generation on generating diverse 3D objects. The result shows that generative AI can simplify 3D creation and generate more diverse objects.

关键词： Solid modeling Three-dimensional displays Procedural generation Generative AI Metaverse Complexity theory Usability

来源：评论

学校读者我要写书评

暂无评论

Dialog Breakdown Detection using Multimodal Features for Non-task-oriented Dialog Systems 11

Dialog Breakdown Detection using Multimodal Features for Non...

引用

11th IEEE Global Conference on Consumer Electronics, GCCE 2022

作者： Tsubokura, Kazuya Iribe, Yurie Kitaoka, Norihide Aichi Prefectural University Graduate School of Information Science and Technology Nagakute Japan Toyohashi University of Technology Department of Computer Science and Engineering Toyohashi Japan

ISBN: (纸本)9781665492324

In this paper, we propose a dialog breakdown detection method for non-task-oriented dialog systems, using multimodal information such as language information, acoustic information, facial expressions, and gaze behavior. Existing automated dialog systems do not always respond appropriately to the utterances of their users, and these systems lack the ability to restore the flow of the dialog by noticing when they have made a serious error. To address this problem, in this study we propose a method of detecting dialog breakdowns using multimodal information collected from users. Data from interactions between human users and the dialog system, such as the content and acoustic features of the user's speech, as well as their facial expressions and gaze direction, are collected using the' Wizard of Oz' (WOZ) method. We then focus on exchanges where breakdowns in the dialog occurred. Features are extracted from the data for each modality (language, acoustic and visual). After dimension compression using principal component analysis (PCA), dialog breakdown detection is conducted using a deep neural network (DNN). In our evaluation experiment, the proposed method achieved an 85.7% accuracy rate for dialog breakdown detection. © 2022 IEEE.

关键词： Principal component analysis

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：