Research map / Continual learning / Online and replay-based
Meta Continual Learning Revisited: Implicitly Enhancing Online Hessian Approximation via Variance Reduction
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Revisits meta continual learning and shows it implicitly performs online Hessian approximation for regularization-based continual learning, using variance reduction to improve it.
Problem
Regularization-based continual learning methods suffer suboptimal trade-offs between knowledge transfer and forgetting because their Hessian approximations of model weights are fixed and unchanging during training.
Approach
Connects meta continual learning to regularization-based methods and argues that meta-learning implicitly yields an online Hessian approximation, which is further enhanced via variance reduction so the Hessian estimate adapts during training.
Experimental setup
Builds on: meta continual learning regularization-based continual learning
Closest papers by meaning
- Learning to Continually Learn with the Bayesian Principle (ICML 2024 · similarity 0.91)
- A Statistical Theory of Regularization-Based Continual Learning (ICML 2024 · similarity 0.90)
- Optimal Rates in Continual Linear Regression via Increasing Regularization (NeurIPS 2025 · similarity 0.89)
- Train with Perturbation, Infer after Merging: A Two-Stage Framework for Continual Learning (NeurIPS 2025 · similarity 0.89)
- Trust Region Continual Learning as an Implicit Meta-Learner (NeurIPS 2026 · similarity 0.88)
- An Effective Dynamic Gradient Calibration Method for Continual Learning (ICML 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.