Reproduce “Prediction and Generalisation over Directed Actions by Grid Cells” 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
Changmin Yu, Timothy E. J. Behrens, Neil Burgess
The authors develop a method for predicting state occupancy under different directed actions. It reuses Fourier modes for translation-invariant transition structures with periodic boundaries, updating action-specific eigenvalues rather than recomputing the eigenbasis. The method also supports simulated navigation under wind and obstacle conditions. In CartPole, Fourier-based agents learn faster than baseline agents, though this preliminary result does not establish generalization.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.