technique
Recovery Subspace Dimensionality
techniqueactiveprovisional
recovery-subspace-dimensionality-bbc803c6·1 events·first seen 5d agoAliases: Recovery Subspace Dimensionality
Co-occurring entities
More like this (12)
Subspace Projectionlow-rank subspace projectionUnstable Features, Reproducible Subspaces: Understanding Seed Dependence in Sparse AutoencodersLatent World Recovery for Multimodal Learning with Missing ModalitiesThe Stable Recovery Manifold: Geometric Principles Governing Recoverability in Continual LearningLatent World Recoverytarget-space recovery profilesRecoverable but Not Stationary: Local Linear Structures in Weights and ActivationsRank-Constrained Subspace Learning (RCSL)Sparse Subspace-to-Expert Sharing for Task-Agnostic Continual LearningDynamical Systems ReconstructionRecovering the Zipfian Distribution in Unsupervised Term Discovery
Recent events (1)
Stable Recovery Manifold hypothesis: catastrophic forgetting as accessibility problem, not information destruction
A new arXiv preprint investigates the geometric structure of recoverability in continual learning using Split CIFAR-100 and a sequentially trained ResNet-18. The authors introduce Recovery Subspace Dimensionality (k_t) and find that recovery dimensionality remains stable across tasks (mean k_t = 8.0) despite substantial representational drift, with principal-angle drift strongly predicting recoverability (r = -0.862). The findings support the Stable Recovery Manifold hypothesis: forgotten knowledge remains compactly decodable, reframing catastrophic forgetting as a manifold-alignment and accessibility problem rather than true information loss.