On Block g-Circulant Matrices with Discrete Cosine and Sine Transforms for Transformer-Based Translation Machine

Author:

Asriani Euis1,Muchtadi-Alamsyah Intan23ORCID,Purwarianti Ayu34

Affiliation:

1. Doctoral Program of Mathematics, Faculty of Mathematics and Natural Sciences, Institut Teknologi Bandung, Bandung 40132, Indonesia

2. Algebra Research Group, Faculty of Mathematics and Natural Sciences, Institut Teknologi Bandung, Bandung 40132, Indonesia

3. University Center of Excellence Artificial Intelligence on Vision, Natural Language Processing and Big Data Analytics (U-CoE AI-VLB), Institut Teknologi Bandung, Bandung 40132, Indonesia

4. Informatics Research Group, School of Electrical Engineering and Informatics, Institut Teknologi Bandung, Bandung 40132, Indonesia

Abstract

Transformer has emerged as one of the modern neural networks that has been applied in numerous applications. However, transformers’ large and deep architecture makes them computationally and memory-intensive. In this paper, we propose the block g-circulant matrices to replace the dense weight matrices in the feedforward layers of the transformer and leverage the DCT-DST algorithm to multiply these matrices with the input vector. Our test using Portuguese-English datasets shows that the suggested method improves model memory efficiency compared to the dense transformer but at the cost of a slight drop in accuracy. We found that the model Dense-block 1-circulant DCT-DST of 128 dimensions achieved the highest model memory efficiency at 22.14%. We further show that the same model achieved a BLEU score of 26.47%.

Funder

Hibah PDD Kementerian Pendidikan, Kebudayaan, Riset, dan Teknologi 2023

Publisher

MDPI AG

Reference37 articles.

1. Mitsuda, K., Higashinaka, R., Sugiyama, H., Mizukami, M., Kinebuchi, T., Nakamura, R., Adachi, N., and Kawabata, H. (2022). Conversational AI for Natural Human-Centric Interaction: Proceedings of the 12th International Workshop on Spoken Dialogue System Technology, Singapore, IWSDS 2021, Springer Nature.

2. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A.N., Kaiser, L., and dan Polosukhin, I. (2017). Attention is all you need. arXiv.

3. Ranganathan, J., and Abuka, G. (December, January 29). Text summarization using transformer model. Proceedings of the 2022 Ninth International Conference on Social Networks Analysis, Management and Security (SNAMS), Milan, Italy.

4. Arnab, A., Dehghani, M., Heigold, G., Sun, C., Lucic, M., and Schmid, C. (2021, January 10–17). Vivit: A video vision transformer. Proceedings of the IEEE/CVF International Conference on Computer Vision, Montreal, QC, Canada.

5. Zeng, P., Zhang, H., Song, J., and Gao, L. (2022, January 23–29). S2 transformer for image captioning. Proceedings of the International Joint Conferences on Artificial Intelligence, Vienna, Austria.

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3