Reproduce “Inner Product Regularization for Imitation 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
Benjamin A. Christie, Tianyu Qiu, Amy H. Choi, David Fridovich-Keil, Dylan P. Losey
InneR adds a training-time regularizer to imitation learning that encourages nearby robot states to share directions their policy ignores while preserving task-relevant responses. It aligns local policy response matrices and penalizes small eigenvalues of neighborhood averages. Across three real-robot tasks, it averaged 56.7% success with three distractors versus behavior cloning’s 18.3%; without distractors, averages were similar, in structured state-vector tests.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.