AI RESEARCH

Self-Scaled Broyden Family of Quasi-Newton Methods in JAX

arXiv CS.LG

ArXi:2603.10599v1 Announce Type: cross We present a JAX implementation of the Self-Scaled Broyden family of quasi-Newton methods, fully compatible with JAX and building on the Optimistix~\cite{rader_optimistix_2024} optimisation library. The implementation includes BFGS, DFP, Broyden and their Self-Scaled variants(SSBFGS, SSDFP, SSBroyden), together with a Zoom line search satisfying the strong Wolfe conditions.