Task-Projected Hyperdimensional Computing for Multi-task Learning

被引:1
|
作者
Chang, Cheng-Yang [1 ]
Chuang, Yu-Chuan [1 ]
Wu, An-Yeu [1 ]
机构
[1] Natl Taiwan Univ, Grad Inst Elect Engn, Taipei, Taiwan
来源
ARTIFICIAL INTELLIGENCE APPLICATIONS AND INNOVATIONS, AIAI 2020, PT I | 2020年 / 583卷
关键词
Hyperdimensional Computing; Multi-task learning; Redundant dimensionality;
D O I
10.1007/978-3-030-49161-1_21
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Brain-inspired Hyperdimensional (HD) computing is an emerging technique for cognitive tasks in the field of low-power design. As an energyefficient and fast learning computational paradigm, HD computing has shown great success in many real-world applications. However, an HD model incrementally trained on multiple tasks suffers from the negative impacts of catastrophic forgetting. The model forgets the knowledge learned from previous tasks and only focuses on the current one. To the best of our knowledge, no study has been conducted to investigate the feasibility of applying multi-task learning to HD computing. In this paper, we propose Task-Projected Hyperdimensional Computing (TP-HDC) to make the HD model simultaneously support multiple tasks by exploiting the redundant dimensionality in the hyperspace. To mitigate the interferences between different tasks, we project each task into a separate subspace for learning. Compared with the baselinemethod, our approach efficiently utilizes the unused capacity in the hyperspace and shows a 12.8% improvement in averaged accuracy with negligible memory overhead.
引用
收藏
页码:241 / 251
页数:11
相关论文
共 50 条
  • [21] Additive multi-task learning models and task diagnostics
    Miller, Nikolay
    Zhang, Guoyi
    COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION, 2024, 53 (12) : 6120 - 6137
  • [22] Multi-task deep learning for medical image computing and analysis: A review
    Zhao, Yan
    Wang, Xiuying
    Che, Tongtong
    Bao, Guoqing
    Li, Shuyu
    COMPUTERS IN BIOLOGY AND MEDICINE, 2023, 153
  • [23] Multi-task learning for intelligent data processing in granular computing context
    Liu, Han
    Cocea, Mihaela
    Ding, Weili
    GRANULAR COMPUTING, 2018, 3 (03) : 257 - 273
  • [24] Communication-Efficient Coded Computing for Distributed Multi-Task Learning
    Hu, Haoyang
    Wu, Youlong
    Shi, Yuanming
    Li, Songze
    Jiang, Chunxiao
    Zhang, Wei
    IEEE TRANSACTIONS ON COMMUNICATIONS, 2023, 71 (07) : 3861 - 3875
  • [25] A Comparison of Multi-task Learning and Single-Task Learning Approaches
    Marquet, Thomas
    Oswald, Elisabeth
    APPLIED CRYPTOGRAPHY AND NETWORK SECURITY WORKSHOPS, ACNS 2023 SATELLITE WORKSHOPS, ADSC 2023, AIBLOCK 2023, AIHWS 2023, AIOTS 2023, CIMSS 2023, CLOUD S&P 2023, SCI 2023, SECMT 2023, SIMLA 2023, 2023, 13907 : 121 - 138
  • [26] TASK AWARE MULTI-TASK LEARNING FOR SPEECH TO TEXT TASKS
    Indurthi, Sathish
    Zaidi, Mohd Abbas
    Lakumarapu, Nikhil Kumar
    Lee, Beomseok
    Han, Hyojung
    Ahn, Seokchan
    Kim, Sangha
    Kim, Chanwoo
    Hwang, Inchul
    2021 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP 2021), 2021, : 7723 - 7727
  • [27] Fitting and sharing multi-task learning
    Piao, Chengkai
    Wei, Jinmao
    APPLIED INTELLIGENCE, 2024, 54 (9-10) : 6918 - 6929
  • [28] Convex multi-task feature learning
    Andreas Argyriou
    Theodoros Evgeniou
    Massimiliano Pontil
    Machine Learning, 2008, 73 : 243 - 272
  • [29] Manifold Regularized Multi-Task Learning
    Yang, Peipei
    Zhang, Xu-Yao
    Huang, Kaizhu
    Liu, Cheng-Lin
    NEURAL INFORMATION PROCESSING, ICONIP 2012, PT III, 2012, 7665 : 528 - 536
  • [30] Multi-task learning with deformable convolution
    Li, Jie
    Huang, Lei
    Wei, Zhiqiang
    Zhang, Wenfeng
    Qin, Qibing
    JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION, 2021, 77