Feature extraction for tensor data serves as an important step in many tasks such as anomaly detection, process monitoring, image classification, and quality control. Although many methods have been proposed for tensor feature extraction, there are still two challenges that need to be addressed: (i) how to reduce the computation cost for high dimensional and large volume tensor data; (ii) how to interpret the output features and evaluate their significance. The most recent methods in deep learning, such as Convolutional Neural Network, have shown outstanding performance in analyzing tensor data, but their wide adoption is still hindered by model complexity and lack of interpretability. To fill this research gap, we propose to use CP-decomposition to approximately compress the convolutional layer (CPAC-Conv layer) in deep learning. The contributions of our work include three aspects: (i) we adapt CP-decomposition to compress convolutional kernels and derive the expressions of forward and backward propagations for our proposed CPAC-Conv layer; (ii) compared with the original convolutional layer, the proposed CPAC-Conv layer can reduce the number of parameters without decaying prediction performance. It can combine with other layers to build novel Deep Neural Networks; (iii) the value of decomposed kernels indicates the significance of the corresponding feature map, which provides us with insights to guide feature selection.

Yinan Wang

Weihong “Grace” Guo

Xiaowei Yue

IISE Transactions

IISE TRANSACTIONS 目前在以下服务中被提取/索引：CSA/ASCE 土木工程文摘； CSA-计算机与信息系统文摘； CSA-腐蚀文摘； CSA-电子与通信文摘； CSA-工程材料文摘；带有 METADEX 的 CSA-材料研究数据库； CSA-机械与交通工程文摘； CSA-固态和超导文摘； INSPEC INFORMATION SERVICES AND SCIENCE CITATION INDEX.INSTITUTE OF INDUSTRIAL AND SYSTEMS ENGINEERS 和我们的出版商 TAYLOR & FRANCIS 尽一切努力确保我们出版物中包含的所有信息（“内容”）的准确性。但是，工业和系统工程师协会和我们的出版商 TAYLOR & FRANCIS、我们的代理和我们的许可方不对内容的准确性、完整性或对任何目的的适用性作出任何陈述或保证。本出版物中表达的任何观点和观点均为作者的观点和观点，并非工业与系统工程师协会和我们的出版商 TAYLOR & FRANCIS 的观点或认可。不应依赖内容的准确性，而应通过主要信息来源进行独立验证。工业和系统工程师协会和我们的出版商 TAYLOR & FRANCIS 不对任何直接或间接与与使用内容有关或因使用内容而产生的。访问和使用条款和条件可在 HTTP://WWW.TANDFONLINE.COM/PAGE/TERMS-AND-CONDITIONS 找到。

IISE Transactions is currently indexed in the following services: CSA/ASCE Civil Engineering Abstracts, CSA Computer and Information Systems Abstracts, CSA Corrosion Abstracts, CSA Electronics and Communication Abstracts, CSA Engineering Materials Abstracts, CSA Materials Research Database with METADEX, CSA Mechanical and Transportation Engineering Abstracts, CSA Solid State and Superconductivity Abstracts, and INSPEC Information Services and Science Citation Index. The Institute of Industrial and Systems Engineers and our publisher, Taylor & Francis, strive to ensure the accuracy of the information contained in our publications. However, neither the Institute of Industrial and Systems Engineers nor our publisher, Taylor & Francis, nor our agents or licensors make any representations or warranties regarding the accuracy, completeness, or suitability of the content for any purpose. The views and opinions expressed in this publication are those of the authors and do not necessarily reflect the views or endorsements of the Institute of Industrial and Systems Engineers or our publisher, Taylor & Francis. The accuracy of the content should not be relied upon and should be independently verified through primary sources. The Institute of Industrial and Systems Engineers and our publisher, Taylor & Francis, are not responsible for any direct or indirect consequences arising from the use of the content. Terms and conditions for access and use can be found at http://www.tandfonline.com/page/terms-and-conditions.

IISE Transactions 目前在以下服务中被提取和索引：CSA/ASCE土木工程文摘、CSA计算机与信息系统文摘、CSA腐蚀文摘、CSA电子与通信文摘、CSA工程材料文摘、带有METADEX的CSA材料研究数据库、CSA机械与交通工程文摘、CSA固态和超导文摘，以及INSPEC信息服务和科学引文索引。工业与系统工程师协会及我们的出版商Taylor & Francis尽力确保本出版物中包含的信息的准确性。然而，工业与系统工程师协会及我们的出版商Taylor & Francis、我们的代理和许可方不对内容的准确性、完整性或适用性作出任何声明或保证。本出版物中表达的观点和看法仅代表作者个人，并不代表工业与系统工程师协会或我们的出版商Taylor & Francis的观点或认可。内容的准确性不应被依赖，建议通过主要信息来源进行独立验证。工业与系统工程师协会及我们的出版商Taylor & Francis不对因使用内容而直接或间接产生的任何后果负责。访问和使用条款和条件可在http://www.tandfonline.com/page/terms-and-conditions找到。

Tensor decomposition to compress convolutional layers in deep learning

This work proposes to use CP-decomposition to approximately compress the convolutional layer (CPAC-Conv layer) in deep learning, which can combine with other layers to build novel Deep Neural Networks and indicate the significance of the corresponding feature map.

Category	Quartile
工程技术	3
工程技术, 工程工业	4
工程技术, 运筹学与管理科学	4

Tensor decomposition to compress convolutional layers in deep learning

ivySCI AI Smartly Parses PDF, Answers Researchers' Questions, and Helps You Understand Papers in Seconds

Journal Info

Category	Quartile
OPERATIONS RESEARCH & MANAGEMENT SCIENCE	2