_spg_optim returns history of objective function values #1856

Merged
merged 1 commit into from Aug 1, 2014

Projects

None yet

2 participants

@kshedden
Contributor

This PR attaches the sequence of all objective function values visited by _spg_optim to the results that it returns. It also adds tests for the convergence status of cor_nearest_factor.

@kshedden kshedden _spg_optim returns history of objective function values
555028e
@josef-pkt
Member

I was also trying out different things today, but didn't see anything that would be a likely cause.
ARPACK svds on Ubuntu 32 might need to be checking.
I was trying out different starting values and different lam_min and lam_max.

One thing I also added in my version is to check the number of iterations:
adding 'iterations': itr to the return Bunch in _spg_optim.

I was a bit surprised that it takes around 1000 iterations.

@josef-pkt josef-pkt added the PR label Aug 1, 2014
@josef-pkt
Member

I'm merging this as another trial to figure out what's going on.
(I didn't have time to add my changes.)

possible problems
start value: uses ARPACK svds which changed across scipy versions
convergence criterium: When I was fiddling with lam_min and lam_max, the iterations stopped too early. (not sure whether it was supposed to or not)

@josef-pkt josef-pkt merged commit b257c7c into statsmodels:master Aug 1, 2014

2 checks passed

continuous-integration/appveyor AppVeyor build succeeded
Details
continuous-integration/travis-ci The Travis CI build passed
Details
@josef-pkt josef-pkt added the type-test label Aug 24, 2014
@josef-pkt josef-pkt added this to the 0.6 milestone Aug 24, 2014
@kshedden kshedden deleted the kshedden:cor_nearest_maint2 branch Sep 27, 2014
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment