Revisiting Flatness-Aware Optimization in Continual Learning With Orthogonal Gradient Projection
Sign Up to like & getrecommendations! Published in 2025 at "IEEE Transactions on Pattern Analysis and Machine Intelligence"
DOI: 10.1109/tpami.2025.3539019
Abstract: The goal of continual learning (CL) is to learn from a series of continuously arriving new tasks without forgetting previously learned old tasks. To avoid catastrophic forgetting of old tasks, orthogonal gradient projection (OGP) based… read more here.
Keywords: projection; loss; flatness aware; new tasks ... See more keywords
Adversarial Feature Alignment: Avoid Catastrophic Forgetting in Incremental Task Lifelong Learning
Sign Up to like & getrecommendations! Published in 2019 at "Neural Computation"
DOI: 10.1162/neco_a_01232
Abstract: Humans are able to master a variety of knowledge and skills with ongoing learning. By contrast, dramatic performance degradation is observed when new tasks are added to an existing neural network model. This phenomenon, termed… read more here.
Keywords: adversarial feature; catastrophic forgetting; feature alignment; new tasks ... See more keywords