Guodong Zheng, Yan Sun, Peng Wang, Enneng Yang, Jiafei Lyu, Xiaochun Cao, Dacheng Tao, Li Shen · 2026 · arXiv
DOI: 10.48550/arXiv.2610.31551 · open-access PDF
Comprender el olvido catastrófico, en el que el aprendizaje de nuevas tareas degrada el rendimiento en tareas previamente aprendidas, sigue siendo un desafío fundamental en el aprendizaje continuo (CL). Los análisis teóricos existentes a menudo se basan en modelos o estructuras de tareas específicas y ofrecen una visión limitada de cómo la optimización iterativa afecta al olvido. Para abordar esta brecha, desarrollamos un marco de estabilidad algorítmica para el olvido esperado de la población bajo SGD sin prescribir una arquitectura de modelo particular.
Machine translation (Meta M2M100); the English original follows.
Understanding catastrophic forgetting, in which learning new tasks degrades performance on previously learned tasks, remains a fundamental challenge in continual learning (CL). Existing theoretical analyses often rely on specific model or task structures and offer limited insight into how iterative optimization affects forgetting. To address this gap, we develop an algorithmic stability framework for expected population forgetting under SGD without prescribing a particular model architecture. Specifically, we decompose forgetting into intra-task and cross-task generalization gaps and empirical
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.31551, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange