Energy & materialsResearch findings
A Pseudo-Feature-Based Approach for Federated Class Incremental Learning
Original published 2026.11.11 13:41 KSTAdded to Blackdeck 2026.11.15 20:11 KST
What has changed?
Pseudo features generated from prototypes effectively reduce catastrophic forgetting without past data
Key points
- Addresses catastrophic forgetting in federated learning with continuously arriving new classes
- Eliminates reliance on past data by using pseudo features derived from prototypes
- Reduces communication cost and improves learning efficiency
Connect it to your work
No need for historical data or heavy models like GANs, reducing communication overhead
What remains unproven
This research note is based on an abstract; check the original for test conditions and application scope.
Originals used for this note
Links open the original publisher in a new tab.