More Than Memory Savings: Zeroth-Order Optimization Mitigates Forgetting in Continual Learning
Anzeige
Ähnliche Artikel
arXiv – cs.LG
•
DecoHD: Hyperdimensional Klassifikation mit extremen Speicherersparnissen
arXiv – cs.AI
•
Memory-Free Continual Learning with Null Space Adaptation for Zero-Shot Vision-Language Models
arXiv – cs.LG
•
An Ensembled Penalized Federated Learning Framework for Falling People Detection
arXiv – cs.LG
•
Local Timescale Gates for Timescale-Robust Continual Spiking Neural Networks
arXiv – cs.LG
•
Self-Evolving LLMs via Continual Instruction Tuning
arXiv – cs.AI
•
Forget What's Sensitive, Remember What Matters: Token-Level Differential Privacy in Memory Sculpting for Continual Learning