Open NightWinkle opened 4 months ago
Yes, but notice I initialize at $y = 1$. The mínima is unique and at $0$.
Le jeu. 29 févr. 2024, 20:37, Jason Rader @.***> a écrit :
Hmm, does this persist for functions whose derivative is not $0$ at $0$? $y^Ty
- N$ has derivative $0$ at $y = 0$. It's unsurprising if BFGS does not move when initialised at a critical point.
— Reply to this email directly, view it on GitHub https://github.com/patrick-kidger/optimistix/issues/47#issuecomment-1971830099, or unsubscribe https://github.com/notifications/unsubscribe-auth/AB2T2XDAUO6QVOPLTUQHGLTYV6BOFAVCNFSM6AAAAABEAAKIICVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSNZRHAZTAMBZHE . You are receiving this because you authored the thread.Message ID: @.***>
Ah, I think I agree. In fact I think we might be making the same mistake in other solvers as well, e.g.
I'd be happy to take a pull request (+ some tests) adjusting this!
Hi,
I'm not sure if this is the desired behavior, but BFGS directly stops if
f(y0) == 0
. As an example :This is because then the termination condition is already validated at first step. It's not really a big problem since you just need to offset the cost by a constant but it took me a while to figure out where the problem was coming from. Maybe the initial value for
f_info
should be set to infinity, or maybe this should be mentioned in the doc ?