.png){kind=icon}
.png){kind=icon}
Return
Continual learning via semantic memory system
.png){kind=icon}
PRE
AI
DOI：10.1016/j.patcog.2026.114689.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
•
A novel SMS framework preserves multi-source representations and provides semantically rich and robust information for continual learning.
•
A novel DGBMO approach optimizes a graph relation matrix that leads to storing a selection of more diverse representations.
•
A novel SRE approach estimates the gradient direction to improve the sample’s quality.
Keywords:
Continual learning
Dynamic expansion models
Memory-based method
Pre-trained models
Journal
IF:
7.6
Papers:
1.3W
Citations:
4.5W