BFGS for the unit ball
BFGS for the unit ball
Let be an initial unit vector and let be an initial positive-definite matrix. Repeatedly define
and then update and .
BFGS for the unit ball. The trial step converges to zero.
This conjecture concerns the behavior of the BFGS update when the underlying compact convex set is the unit ball. The preceding discussion establishes determinant decrease for the update in the general membership algorithm, while the convergence of the trial steps in this special case is suggested only by numerical experiments.
Sources & referencesView supporting material
Primary source
Jiayi Guo and Adrian S. Lewis, “Rescaling nonsmooth optimization using BFGS and Shor updates”, arXiv:1802.06453 (2018).
Progress summary
Nothing recorded yet. Refresh searches the literature and the public web for attempts on this problem, and writes the first summary here.
Solutions 0
Sign in to submit a solution.
No solutions have been posted yet.