.png){kind=icon}
Return
The continuous learning algorithm with null space and eigenface-based orthogonal weight modification
.png){kind=icon}
PRE
AI
DOI：10.1016/j.eswa.2025.127468.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The challenge in continual learning is to enable deep neural networks to learn new knowledge in sequential task learning without forgetting previously acquired task knowledge. Orthogonal Weight Modification (OWM) is a significant continual learning algorithm proposed in recent years. The CEig-OWM algorithm is a continual learning algorithm that employs orthogonal weight modification based on eigenface input space representation. To enhance the stability of the CEig-OWM algorithm in retaining learned knowledge, this study introduces the null space of zero eigenvalue vectors into the calculation of the input space orthogonal matrix within the CEig-OWM algorithm, yielding the Null Space Eigenface-based Orthogonal Weight Modification (NEig-OWM) continual learning algorithm. Extensive comparative experiments on continual learning have been conducted across various datasets, and the experimental results demonstrate that the NEig-OWM algorithm exhibits the best continual learning performance among those with comparable parameter sizes.
Keywords:
Continual learning
Orthogonal weight modification
Null space
Eigenface
Journal
IF:
7.5
Papers:
3.0W
Citations:
10.2W
Organization
No organization information available
Cited Papers
.png){kind=icon}
ShareSave
ShareSave
ShareSaveShareSave no more
no more