Steven G. Johnson

@jpritikin this will only work if it's not an active constraint at the optimum, I think.

It should be enough to just increase rho by a factor of 10 and run another inner iteration.

That's odd; with a larger rho, it should take smaller steps, from what I recall; I wonder why it is giving the same `g`? Can you print out where you...

Most of the algorithms don't construct a Hessian approximation… and most of the ones that do are low-storage variants that construct the Hessian (or, more commonly, its inverse) only implicitly...

It's tricky to design a good API to expose information that is algorithm-dependent, since the *type* of information (e.g. how the Hessian is represented) will be completely different for different...

> Or some offer the hessian, others the inv_hessian As I said, it's even worse than that. Many of the algorithms that compute a Hessian do so in a rather...

> if an uncertainty-like quantity can be returned, that would be nice already. Feasible? It is certainly easier to have a uniform API to fetch an array of uncertainties for...

f2c offsets all of the array pointers by `-1` so that it can use the original 1-based indexing in C. If the compiler is complaining about subtracting 1 from the...

The ORIG_DIRECT algorithm actually lets you return NaN to indicate infeasible points: https://github.com/stevengj/nlopt/blob/master/src/algs/direct/README

Note that the catastrophic cancellation happens way sooner than |x|=20 … by `x = 4` the `1 - tanh(x)^2` formula has already lost 2 digits, by `x = 6` it...