Jason K. Moore
Jason K. Moore
If you are making use of the results of "info" in some way, I would reconsider that. Ideally we save only a single txt file per example (or per solution).
> So should not at least one simple example show how the info dict is used? Sure that's fine. There are some examples that run super fast and we don't...
But keep in mind that `info` dict is something from cyipopt (not opty), so we don't technically need to explain it here.
@HuaweiWang, if you have a moment, could you confirm that I'm doing the same thing you do in your paper? I've made a small modification to this example: https://opty.readthedocs.io/latest/examples/plot_non_contiguous_parameter_identification.html by...
I saw your bias PR, but I'm not sure having biased measurements is so common (at least not large bias relative to the measurements).
Do you think your bias addition reduces the chance the optimizer finds a local minima?
> Do you think, this simple spring_damper_friction mass system is too simple a system to test these methods? If we can't get these things to work well with a simple...
If you read the paper, do you think I'm implementing it correctly? It seems too simple to be true.
> I just thought, that maybe this simple system only has one minimum for (c, k, friction), so Wang's method cannot help to jump out of a local minimum to...
Here is Huawei's code: https://github.com/Huawei-ClevelandStateUniv/Stochastic_Paper/blob/master/controller_identification/StandingModel_PD_Eps_Noise_SameInit.py#L306 It seems like the noise is added just like we are doing.