It attempts to avoid steps that are too large and can help improve robustness 
of convergence in some cases.


> On Oct 17, 2016, at 4:53 AM, André Carvalho da Silva 
> <> wrote:
> Hello everyone,
> I’ve been wondering about what does the “optional step-control” part of the 
> MIPS algorithm serves for.
> Ok, I understand it must be associated with the MIPS alphaD and alphaP steps, 
> but I wonder for what kind of particular conditions it is useful. 
> Does it try to accelerate convergence? Does it solve convergence problems?
> Thank you in advance.
> Best regards,
> André

Reply via email to