English

Continual Learning by Three-Phase Consolidation

Machine Learning 2024-03-25 v1 Computer Vision and Pattern Recognition

Abstract

TPC (Three-Phase Consolidation) is here introduced as a simple but effective approach to continually learn new classes (and/or instances of known classes) while controlling forgetting of previous knowledge. Each experience (a.k.a. task) is learned in three phases characterized by different rules and learning dynamics, aimed at removing the class-bias problem (due to class unbalancing) and limiting gradient-based corrections to prevent forgetting of underrepresented classes. Several experiments on complex datasets demonstrate its accuracy and efficiency advantages over competitive existing approaches. The algorithm and all the results presented in this paper are fully reproducible thanks to its publication on the Avalanche open framework for continual learning.

Keywords

Cite

@article{arxiv.2403.14679,
  title  = {Continual Learning by Three-Phase Consolidation},
  author = {Davide Maltoni and Lorenzo Pellegrini},
  journal= {arXiv preprint arXiv:2403.14679},
  year   = {2024}
}

Comments

13 pages, 2 figures, 8 tables. Preprint under review

R2 v1 2026-06-28T15:29:03.547Z