Source code for sklvq.solvers._limited_memory_bfgs

from . import ScipyBaseSolver
from ..objectives import ObjectiveBaseClass


[docs]class LimitedMemoryBfgs(ScipyBaseSolver): r"""Limited memory variant of BFGS (L-BFGS) See the documentation of scipy for the parameter list and description. Parameters ---------- jac: None Is set automatically to objective gradient method. However, if no gradient function is available, e.g., for a custom distance function, then jac can be set to None. callback: callable Differently from the non-scipy solvers the signature is callback(xk) with xk the current set of variables, which are the model parameters flattened to one 1D array. """ def __init__(self, objective: ObjectiveBaseClass, **kwargs): super(LimitedMemoryBfgs, self).__init__(objective, method="L-BFGS-B", **kwargs)