检索结果-内蒙古大学图书馆

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

作者： Leonardo Defilippis Bruno Loureiro Theodor Misiakiewicz Département d'Informatique École Normale Supérieure - PSL & CNRS Department of Statistics and Data Science Yale University

ISBN: (纸本)9798331314385

In this work we investigate the generalization performance of random feature ridge regression (RFRR). Our main contribution is a general deterministic equivalent for the test error of RFRR. Specifically, under a certain concentration property, we show that the test error is well approximated by a closed-form expression that only depends on the feature map eigenvalues. Notably, our approximation guarantee is non-asymptotic, multiplicative, and independent of the feature map dimension— allowing for infinite-dimensional features. We expect this deterministic equivalent to hold broadly beyond our theoretical analysis, and we empirically validate its predictions on various real and synthetic datasets. As an application, we derive sharp excess error rates under standard power-law assumptions of the spectrum and target decay. In particular, we provide a tight result for the smallest number of features achieving optimal minimax error rate.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Probabilistic Framework for Pruning Transformers Via a Finite Admixture of Keys 48

A Probabilistic Framework for Pruning Transformers Via a Fin...

引用

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

作者： Nguyen, Tan M. Nguyen, Tam Bui, Long Do, Hai Nguyen, Duy Khuong Le, Dung D. Tran-The, Hung Ho, Nhat Osher, Stan J. Baraniuk, Richard G. University of California Department of Mathematics Los Angeles United States Rice University Department of Electrical and Computer Engineering Houston United States Fpt Software Ai Center Ha Noi Viet Nam Vin University College of Engineering and Computer Science Viet Nam Deakin University Applied Artificial Intelligence Institute Geelong Australia The University of Texas Department of Statistics and Data Sciences Austin United States

ISBN: (纸本)9781728163277

Pairwise dot product-based self-attention is key to the success of transformers which achieve state-of-the-art performance across a variety of applications in language and vision, but are costly to compute. It has been shown that most attention scores and keys in transformers are redundant and can be removed without loss of accuracy. In this paper, we develop a novel probabilistic framework for pruning attention scores and keys in transformers. We first formulate an admixture model of attention keys whose input data to be clustered are attention queries. We show that attention scores in self-attention correspond to the posterior distribution of this model when attention keys admit a uniform prior distribution. We then relax this uniform prior constraint and let the model learn these priors from data, resulting in a new Finite Admixture of Keys (FiAK). The learned priors are used for pruning away redundant attention scores and keys in the baseline transformers, improving the diversity of attention patterns that the models capture. We corroborate the efficiency of transformers pruned with FiAK on the ImageNet object classification and WikiText-103 language modeling tasks. Our experiments demonstrate that transformers pruned with FiAK yield similar or better accuracy than the baseline dense transformers while being much more efficient in terms of memory and computational cost. © 2023 IEEE.

关键词： admixture models pruning Transformers

来源：评论

学校读者我要写书评

暂无评论

Improving Cybersecurity Measures Through the Revelation of Malignant URLs using Machine Learning Techniques

Improving Cybersecurity Measures Through the Revelation of M...

引用

data Intelligence and Cognitive Informatics (ICDICI), International Conference on

作者： T. M. Saravanan A. Muthusamy C. P. Thamil Selvi M. Senthil Kumar D. Maheshwari Department of Computer Technology Kongu Engineering College Erode India Department of Artificial Intelligence and Data Science Rathinam Technical Campus Coimbatore India Department of Artificial Intelligence and Data Science Erode Sengunthar Engineering College Erode India Department of Computer Technology KPR College of Arts Science and Research Coimbatore India

ISBN: (数字)9798350389609

ISBN: (纸本)9798350389616

In today's world the internet has ingrained itself deeply into our lives, and web searching has become essential for people of all ages, locations, and occupations. However, due to the rise in internet usage, there has been a rise in spoofing attacks through malicious websites. Online purchases, reservations, recharges, and various other transactions are now commonly conducted online. Internet users are increasing; it has become crucial to develop automatic URL detection systems to protect users. This proposed research aims to compare different categorization models like Decision Tree, Random Forest and Logistic Regression to determine which one achieves finest accuracy in distinguishing among genuine websites and False websites. Objective is to identify attacking sites effectively and determine the artificial intelligence algorithm that provides the best accuracy for this use.

关键词： Uniform resource locators Logistic regression Accuracy Machine learning algorithms Feature extraction Object recognition Random forests Web search Smart phones Regression tree analysis

来源：评论

学校读者我要写书评

暂无评论

Scarce data Selection for Semi-Supervised Few-Shot Learning

SSRN

引用

SSRN 2024年

作者： Wu, Xiaotong Zhou, Jinsong Wang, Chao Department of Statistics and Data Science Southern University of Science and Technology Shenzhen518055 China National Centre for Applied Mathematics Shenzhen Shenzhen518055 China

Semi-supervised learning (SSL) offers an effective approach by leveraging unlabeled data to alleviate the excessive reliance on labeled data. Despite demonstrating promising performance, the issue of selecting the optimal data for annotation has not been fully explored, especially when the label budget is scarce. To address the issue, we propose a framework that selectively labels a small number of samples for manual annotation from an unlabeled dataset. Specifically, we first select samples using unsupervised representation learning and clustering techniques under the principles of representativeness and balancedness. Additionally, we incorporate intra-cluster and inter-cluster regularizations to enhance the effectiveness of selected samples. To broaden its applicability to larger datasets with intricate semantics, we propose a hybrid method that integrates $k$-means into the previous framework. Experiments show that our approaches outperform the current state-of-the-art sample selection methods. © 2024, The Authors. All rights reserved.

关键词： Self-supervised learning

来源：评论

学校读者我要写书评

暂无评论

When are ensembles really effective? 23

When are ensembles really effective?

引用

Proceedings of the 37th International Conference on Neural Information Processing Systems

作者： Ryan Theisen Hyunsuk Kim Yaoqing Yang Liam Hodgkinson Michael W. Mahoney Department of Statistics University of California Berkeley Department of Computer Science Dartmouth College School of Mathematics and Statistics University of Melbourne Australia International Computer Science Institute Lawrence Berkeley National Laboratory and Department of Statistics University of California Berkeley

Ensembling has a long history in statistical data analysis, with many impactful applications. However, in many modern machine learning settings, the benefits of ensembling are less ubiquitous and less obvious. We study, both theoretically and empirically, the fundamental question of when ensembling yields significant performance improvements in classification tasks. Theoretically, we prove new results relating the ensemble improvement rate (a measure of how much ensembling decreases the error rate versus a single model, on a relative scale) to the disagreement-error ratio. We show that ensembling improves performance significantly whenever the disagreement rate is large relative to the average error rate; and that, conversely, one classifier is often enough whenever the disagreement rate is low relative to the average error rate. On the way to proving these results, we derive, under a mild condition called competence, improved upper and lower bounds on the average test error rate of the majority vote classifier. To complement this theory, we study ensembling empirically in a variety of settings, verifying the predictions made by our theory, and identifying practical scenarios where ensembling does and does not result in large performance improvements. Perhaps most notably, we demonstrate a distinct difference in behavior between interpolating models (popular in current practice) and non-interpolating models (such as tree-based methods, where ensembling is popular), demonstrating that ensembling helps considerably more in the latter case than in the former.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Exploring Design Optimisation Techniques of a Radio Telescope Using Fixed Costing Constraints 2

Exploring Design Optimisation Techniques of a Radio Telescop...

引用

2nd International Conference on Artificial Intelligence, Computational Electronics and Communication System, AICECS 2023

作者： Iyer, Aditya Arun Prabhu, Gautham Manuru Gupta, Tanay Deshmukh, Shrey Rivankar, Rushit Department of Computer Science and Engineering Manipal Institute of Technology Manipal Academy of Higher Education Karnataka Manipal576104 India Department of Data Science and Computer Applications Manipal Institute of Technology Manipal Academy of Higher Education Karnataka Manipal576104 India Department of Information Communication Technology Manipal Institute of Technology Manipal Academy of Higher Education Karnataka Manipal576104 India

Cost optimization is a common problem encountered in the design of telescopes. This paper comprehensively discusses various radio telescope designs worldwide, focusing on their design and utilities. It contextualizes the Pulsar data challenge and subsequently discusses and develops a mathematical model for designing radio telescopes. The model assumes a fixed-cost budget. This paper expands current ideas of modeling the system using figure-of-merit equations and optimizing them based on a fixed budget to obtain optimal and affordable radio telescope designs. © Published under licence by IOP Publishing Ltd.

关键词： Radio telescopes

来源：评论

学校读者我要写书评

暂无评论

VIKAS: A Multimodal Framework to Aid in Effective Disaster Management 13th

VIKAS: A Multimodal Framework to Aid in Effective Disaster M...

引用

13th International Conference on Applications and Techniques in Information Security, ATIS 2022

作者： Prabhu, Gautham Manuru Gupta, Tanay Srujan, Metta Venkata Soumya, A.R. Palorkar, Anshita Chowdhury, Anurag Department of Computer Science and Engineering Manipal Institute of Technology Manipal Academy of Higher Education Karnataka Manipal576104 India Department of Data Science and Computer Applications Manipal Institute of Technology Manipal Academy of Higher Education Karnataka Manipal576104 India Department of Information and Communication Technology Manipal Institute of Technology Manipal Academy of Higher Education Karnataka Manipal576104 India

ISBN: (纸本)9789819922635

In the event of a disaster, social media is often used to draw attention to affected areas and distressed people. The massive population and diversity in Indian languages warrant a novel real-time, big-data solution that can increase situational awareness, reduce special forces’ response time, and expedite decision-making. The proposed solution, VIKAS, streams text, images, videos, and audio from posts on microblogging platforms using keywords. It then uses the Google Translate and Transliteration APIs to handle multilingual and macaronic hybrid text, including Hinglish. An Apache Kafka event pipeline processes the sheer volume of posts asynchronously. Duplicate, uninformative, or bot-posted data (checked using the Botometer machine learning algorithm) is discarded. Scraped data is also verified through Google’s FactCheck Explorer API. Audio and video clips are processed leveraging speech-to-text methods. The solution incorporates a BERT pre-trained model and word embeddings for natural language processing tasks, including sentiment analysis and classification of textual data. Image classification and object identification are implemented using a ResNet deep learning model. This multimodal approach pinpoints locations using nearby landmarks, severity, and type of support needed, ranging from humanitarian aid and rescue relief to infrastructure damage. Easy-to-interpret visualizations on an accessible dashboard consolidate many details that can streamline resource distribution and personnel deployment. VIKAS was presented to the National Disaster Response Force at the national-level finals of the Smart India Hackathon 2022. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Sentiment analysis

来源：评论

学校读者我要写书评

暂无评论

Link Prediction based on bipartite graph for recommendation system using optimized SVD++

Link Prediction based on bipartite graph for recommendation ...

引用

2022 International Conference on Machine Learning and data Engineering, ICMLDE 2022

作者： Gupta, Anshul Shrinath, Pravin Department of Computer Engineerig MPSTME Narsee Monjee Institute of Management Studies India Department of Data Science SP Jain School of Global Management India

To alleviate the big data difficulties that have created a potential problem for many Internet users, it is necessary to filter, rank, and efficiently communicate the relevant information on the Web, where the diversity of possibilities is overwhelming. Recommender systems, which sift through enormous amounts of dynamically generated data to give consumers with personalised information and services, are able to overcome this problem. Bipartite graphs are currently generally used to store and understand this data due to its sparse nature. data are mapped to a bipartite user-item interaction network where the graph topology captures detailed information about user-item associations, transforming a recommendation issue into a link prediction problem. Earlier, approaches for link prediction in bipartite graphs for various recommendation systems were developed, but the efficacy of the prediction methodology was not close to the standards required by real-time recommenders. So, the primary goal of this research is to offer an effective link prediction-based recommendation system that takes advantage of SVD++ along with K-nearest neighbors and reduces the system's error rate, leading to better outcomes. The proposed system is sorely tested on the MovieLens dataset and compared to some traditional recommendation methods. The results demonstrate that the suggested strategy exceeds all traditional approaches in terms of accuracy, and the actual suggestions are equally encouraging. © 2023 The Authors. Published by Elsevier B.V.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

One-class systems seamlessly fit in the forward-forward algorithm

arXiv

引用

arXiv 2023年

作者： Hopwood, Michael Department of Statistics and Data Science University of Central Florida OrlandoFL United States

The forward-forward algorithm [Hinton, 2022] presents a new method of training neural networks by updating weights during an inference, performing parameter updates for each layer individually. This immediately reduces memory requirements during training and may lead to many more benefits, like seamless online training. This method relies on a loss ("goodness") function that can be evaluated on the activations of each layer, of which can have a varied parameter size, depending on the hyperparamaterization of the network. In the seminal paper, a goodness function was proposed to fill this need;however, if placed in a one-class problem context, one need not pioneer a new loss because these functions can innately handle dynamic network sizes. In this paper, we investigate the performance of deep one-class objective functions when trained in a forward-forward fashion. The code is available at https://***/MichaelHopwood/ForwardForwardOneclass. © 2023, CC BY.

关键词： Inference engines

来源：评论

学校读者我要写书评

暂无评论

AsT: An Asymmetric-Sensitive Transformer for Osteonecrosis of the Femoral Head Detection 37

AsT: An Asymmetric-Sensitive Transformer for Osteonecrosis o...

引用

37th AAAI Conference on Artificial Intelligence, AAAI 2023

作者： Chen, Haoyang Liu, Shuai Lu, Feng Li, Wei Sheng, Bin Li, Mi Jin, Hai Zomaya, Albert Y. National Engineering Research Center for Big Data Technology and System Services Computing Technology and System Lab Cluster and Grid Computing Lab School of Computer Science and Technology Huazhong University of Science and Technology China Centre for Distributed and High Performance Computing School of Computer Science The University of Sydney Australia Department of Computer Science and Engineering Shanghai Jiao Tong University China Tongji Hospital Tongji Medical College Huazhong University of Science and Technology China

ISBN: (纸本)9781577358800

Early diagnosis of osteonecrosis of the femoral head (ONFH) can inhibit the progression and improve femoral head preservation. The radiograph difference between early ONFH and healthy ones is not apparent to the naked eye. It is also hard to produce a large dataset to train the classification model. In this paper, we propose Asymmetric-Sensitive Transformer (AsT) to capture the uneven development of the bilateral femoral head to enable robust ONFH detection. Our ONFH detection is realized using the self-attention mechanism to femoral head regions while conferring sensitivity to the uneven development by the attention-shared transformer. The real-world experiment studies show that AsT achieves the best performance of AUC 0.9313 in the early diagnosis of ONFH and can find out misdiagnosis cases firmly. Copyright © 2023, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Large dataset

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：