Incremental Learning with Differentiable Architecture and Forgetting Search

被引:1
|
作者
Smith, James Seale [1 ]
Seymour, Zachary [2 ]
Chiu, Han-Pang [2 ]
机构
[1] Georgia Inst Technol, Atlanta, GA 30332 USA
[2] SRI Int, Princeton, NJ USA
关键词
continual learning; incremental learning; neural architecture search;
D O I
10.1109/IJCNN55064.2022.9892400
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
As progress is made on training machine learning models on incrementally expanding classification tasks (i.e., incremental learning), a next step is to translate this progress to industry expectations. One technique missing from incremental learning is automatic architecture design via Neural Architecture Search (NAS). In this paper, we show that leveraging NAS for incremental learning results in strong performance gains for classification tasks. Specifically, we contribute the following: first, we create a strong baseline approach for incremental learning based on Differentiable Architecture Search (DARTS) and state-of-the-art incremental learning strategies, outperforming many existing strategies trained with similar-sized popular architectures; second, we extend the idea of architecture search to regularize architecture forgetting, boosting performance past our proposed baseline. We evaluate our method on both RF signal and image classification tasks, and demonstrate we can achieve up to a 10% performance increase over state-of-the-art methods. Most importantly, our contribution enables learning from continuous distributions on real-world application data for which the complexity of the data distribution is unknown, or the modality less explored (such as RF signal classification).
引用
收藏
页数:8
相关论文
共 50 条
  • [1] Differentiable Architecture Search for Reinforcement Learning
    Miao, Yingjie
    Song, Xingyou
    Co-Reyes, John D.
    Peng, Daiyi
    Yue, Summer
    Brevdo, Eugene
    Faust, Aleksandra
    INTERNATIONAL CONFERENCE ON AUTOMATED MACHINE LEARNING, VOL 188, 2022, 188
  • [2] Graph Differentiable Architecture Search with Structure Learning
    Qin, Yijian
    Wang, Xin
    Zhang, Zeyang
    Zhu, Wenwu
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 34 (NEURIPS 2021), 2021, 34
  • [3] BaLeNAS: Differentiable Architecture Search via the Bayesian Learning Rule
    Zhang, Miao
    Pan, Shirui
    Chang, Xiaojun
    Su, Steven
    Hu, Jilin
    Haffari, Gholamreza
    Yang, Bin
    2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2022, : 11861 - 11870
  • [4] Cyclic Differentiable Architecture Search
    Yu, Hongyuan
    Peng, Houwen
    Huang, Yan
    Fu, Jianlong
    Du, Hao
    Wang, Liang
    Ling, Haibin
    IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2023, 45 (01) : 211 - 228
  • [5] Differentiable quantum architecture search
    Zhang, Shi-Xin
    Hsieh, Chang-Yu
    Zhang, Shengyu
    Yao, Hong
    QUANTUM SCIENCE AND TECHNOLOGY, 2022, 7 (04)
  • [6] Regularized Differentiable Architecture Search
    Wang, Lanfei
    Xie, Lingxi
    Zhao, Kaili
    Guo, Jun
    Tian, Qi
    IEEE EMBEDDED SYSTEMS LETTERS, 2023, 15 (03) : 129 - 132
  • [7] The limitations of differentiable architecture search
    Guillaume, Lacharme
    Hubert, Cardot
    Christophe, Lente
    Nicolas, Monmarche
    PATTERN ANALYSIS AND APPLICATIONS, 2024, 27 (02)
  • [8] Group Differentiable Architecture Search
    Shen, Chaoyuan
    Xu, Jinhua
    IEEE ACCESS, 2021, 9 : 76585 - 76591
  • [9] ATL-DAS: Automatic Topology Learning for Differentiable Architecture Search
    Liu, Qinglin
    Lin, Jingbo
    Lv, Xiaoqian
    Yu, Wei
    Li, Zonglin
    Zhang, Shengping
    DISPLAYS, 2023, 80
  • [10] Differentiable Architecture Search with Random Features
    Zhang, Xuanyang
    Li, Yonggang
    Zhang, Xiangyu
    Wang, Yongtao
    Sun, Jian
    2023 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2023, : 16060 - 16069