2 papers
math.OC2026
Adaptive Gradient-Based Methods for a Broader Class of Optimization Problems under Performative Prediction
Hiroki Hamaguchi, Yuya Hikima, Hiroshi Sawada +1
We study optimization under performative prediction, where deploying a model affects the future data distribution. For this setting, several gradient-based approaches have been pro…
cs.LG2025
FedPM: Federated Learning Using Second-order Optimization with Preconditioned Mixing of Local Parameters
Hiro Ishii, Kenta Niwa, Hiroshi Sawada +3
We propose Federated Preconditioned Mixing (FedPM), a novel Federated Learning (FL) method that leverages second-order optimization. Prior methods--such as LocalNewton, LTDA, and F…