{"id":"arxiv:2610.30460","kind":"preprint","title":"Equivariant Experience Replay for Continual Learning in Imaging Inverse Problems","authors":["Guixian Xu","Jin Zhu","Jinglai Li","Junqi Tang"],"author_count":4,"year":2026,"published":"2026-10-21","venue":null,"cited_by":null,"doi":null,"arxiv_id":"2610.30460","nct_id":null,"source_url":"https://arxiv.org/abs/2610.30460","archive_url":"https://space-kitty.com/papers/arxiv:2610.30460","updated":"2026-10-21","pmid":null,"pmcid":null,"openalex_id":null,"pdf_url":"https://arxiv.org/pdf/2610.30460","abstract":"In this work, we propose Equivariant Experience Replay, a new paradigm for achieving continual learning of deep networks for solving ill-posed imaging inverse problems. Inspired by recent advances on unsupervised learning methods such as equivariant imaging/splitting, we find the exploitation of group-symmetry structure is particularly relevant and effective in continual learning setting for ill-posed inversion tasks. We therefore propose this new experience replay scheme, which is tailored for such inverse problems. Our numerical study demonstrates the superiority of our method compared to existing continual learning methods for inverse problems in medical imaging and image processing.","abstract_license":"CC0-1.0 (arXiv descriptive metadata)","license":null,"is_open_access":false,"retracted":false,"peer_reviewed":false,"categories":["eess.IV"],"details":null,"provenance":"arxiv","attribution":"Abstract: arXiv descriptive metadata (CC0). Thank you to arXiv for use of its open access interoperability. This service was not reviewed or approved by, nor does it necessarily express or reflect the policies or opinions of, arXiv.","topics":[],"cited_in_dossiers":[]}