Reproduce “Forgetting to Improve: Principled Data Removal in Active Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Manuel Wendl, Erik Englesson, Andreas Krause, Carl Henrik Ek
The authors introduce F2I, a method that removes observations from a Gaussian process when estimated to increase predictive uncertainty in a task-specific target region. It combines epistemic and aleatoric influence estimates, updating the model after removals. In Cartpole Balance, Forgetful RL with a linear-GP dynamics model learns to balance, while GP-MBPO retaining all samples does not solve that task.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.