Graph Few-shot Class-incremental Learning

被引:18
|
作者
Tan, Zhen [1 ]
Ding, Kaize [1 ]
Guo, Ruocheng [2 ]
Liu, Huan [1 ]
机构
[1] Arizona State Univ, Tempe, AZ 85287 USA
[2] City Univ Hong Kong, Hong Kong, Peoples R China
关键词
Graph Neural Networks; Incremental Learning; Few-shot Learning;
D O I
10.1145/3488560.3498455
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
The ability to incrementally learn new classes is vital to all real-world artificial intelligence systems. A large portion of high-impact applications like social media, recommendation systems, E-commerce platforms, etc. can be represented by graph models. In this paper, we investigate the challenging yet practical problem, Graph Few-shot Class-incremental (Graph FCL) problem, where the graph model is tasked to classify both newly encountered classes and previously learned classes. Towards that purpose, we put forward a Graph Pseudo Incremental Learning paradigm by sampling tasks recurrently from the base classes, so as to produce an arbitrary number of training episodes for our model to practice the incremental learning skill. Furthermore, we design a Hierarchical-Attention-based Graph Meta-learning framework, HAG-Meta from an optimization perspective. We present a task-sensitive regularizer calculated from task-level attention and node class prototypes to mitigate overfitting onto either novel or base classes. To employ the topological knowledge, we add a node-level attention module to adjust the prototype representation. Our model not only achieves greater stability of old knowledge consolidation, but also acquires advantageous adaptability to new knowledge with very limited data samples. Extensive experiments on three real-world datasets, including Amazon-clothing, Reddit, and DBLP, show that our framework demonstrates remarkable advantages in comparison with the baseline and other related state-of-the-art methods.
引用
收藏
页码:987 / 996
页数:10
相关论文
共 50 条
  • [31] Overcomplete-to-sparse representation learning for few-shot class-incremental learning
    Fu, Mengying
    Liu, Binghao
    Ma, Tianren
    Ye, Qixiang
    [J]. MULTIMEDIA SYSTEMS, 2024, 30 (02)
  • [32] Learning to complement: Relation complementation network for few-shot class-incremental learning
    Wang, Ye
    Wang, Yaxiong
    Zhao, Guoshuai
    Qian, Xueming
    [J]. KNOWLEDGE-BASED SYSTEMS, 2023, 282
  • [33] Overcomplete-to-sparse representation learning for few-shot class-incremental learning
    Fu Mengying
    Liu Binghao
    Ma Tianren
    Ye Qixiang
    [J]. Multimedia Systems, 2024, 30
  • [34] Grassmann Graph Embedding for Few-Shot Class Incremental Learning
    Gu, Ziqi
    Xu, Chunyan
    Cui, Zhen
    [J]. PATTERN RECOGNITION AND COMPUTER VISION, PRCV 2023, PT VIII, 2024, 14432 : 179 - 191
  • [35] Few-Shot Class-Incremental Learning via Class-Aware Bilateral Distillation
    Zhao, Linglan
    Lu, Jing
    Xu, Yunlu
    Cheng, Zhanzhan
    Guo, Dashan
    Niu, Yi
    Fang, Xiangzhong
    [J]. 2023 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2023, : 11838 - 11847
  • [36] Semantic-visual Guided Transformer for Few-shot Class-incremental Learning
    Qiu, Wenhao
    Fu, Sichao
    Zhang, Jingyi
    Lei, Chengxiang
    Peng, Qinmu
    [J]. 2023 IEEE INTERNATIONAL CONFERENCE ON MULTIMEDIA AND EXPO, ICME, 2023, : 2885 - 2890
  • [37] Few-shot class-incremental audio classification via discriminative prototype learning
    Xie, Wei
    Li, Yanxiong
    He, Qianhua
    Cao, Wenchang
    [J]. EXPERT SYSTEMS WITH APPLICATIONS, 2023, 225
  • [38] Synthesized Feature based Few-Shot Class-Incremental Learning on a Mixture of Subspaces
    Cheraghian, Ali
    Rahman, Shafin
    Ramasinghe, Sameera
    Fang, Pengfei
    Simon, Christian
    Petersson, Lars
    Harandi, Mehrtash
    [J]. 2021 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2021), 2021, : 8641 - 8650
  • [39] Self-Promoted Prototype Refinement for Few-Shot Class-Incremental Learning
    Zhu, Kai
    Cao, Yang
    Zhai, Wei
    Cheng, Jie
    Zha, Zheng-Jun
    [J]. 2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021, 2021, : 6797 - 6806
  • [40] Semantic-aware Knowledge Distillation for Few-Shot Class-Incremental Learning
    Cheraghian, Ali
    Rahman, Shafin
    Fang, Pengfei
    Roy, Soumava Kumar
    Petersson, Lars
    Harandi, Mehrtash
    [J]. 2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021, 2021, : 2534 - 2543