Reproduce “Provable Representation Learning for Imitation Learning via Bi-level Optimization” 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
Sanjeev Arora, Simon S. Du, Sham Kakade, Yuping Luo, Nikunj Saunshi
The authors study transferring representations learned from experts on related tasks to new imitation tasks. They learn a shared representation across source tasks, then fit a task-specific policy for a new task. Theoretical bounds show benefits under shared-representation assumptions; simulations generally show higher returns, but scratch policies can win with plentiful target data, and observation-only guarantees retain a Bellman-error term.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.