At the OpenMx Developers meeting on 12/16/11 we discussed the following:
- The content for a grant(s) to fund research related to OpenMx.
- The group discussed progress on debugging the parallel ordinal FIML performance regression. It does not appear as if the regression is due to a race condition which is good in terms of ease of debugging. The current plan is to attack the problem by turning on the
verboseflag in OpenMx. One hypothesis is that there is a situation occuring where the optimization is bi-stabling on the gradient or the hessian when work is divided up that does not occur when the workload is not divided. If this hypothesis turns out to be correct several work-around solutions involving tuning the optimizer precision up and down were discussed.