Reproduce “GSP-KalmanNet: Tracking Graph Signals via Neural-Aided Kalman Filtering” 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
Itay Buchnik, Guy Sagi, Nimrod Leinwand, Yuval Loya, Nir Shlezinger, Tirza Routtenberg
GSP-KalmanNet combines state-space prediction with a learned graph-filter update. The RNN produces a gain diagonal in graph-frequency coordinates rather than an unrestricted Kalman gain. In power-system state-estimation experiments with exponential measurement noise, it outperformed other learnable benchmarks, while learnable methods showed a greater-than-10-dB gap over model-based methods. These results apply to tested conditions and stationary graphs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.