Induced-metric optimizer where the inverse metric γ⁻¹ = exp(s)·I is a learnable scalar, updated online each step by ascending a local surrogate v = ξ·exp(s)·‖g‖² with regularization and clipping in log-domain. Plain and log-loss embedding variants. O(1) learnable state (one scalar s).
Issues are discussions and bug reports on this entry.
No issues yet.