At developers meeting on 11/4 we discussed the following:

- The group discussed Dan Hackett's latest progress with PPML. Dan has achieved some dramatic speedups (up to 30x speedup) for models with fixed structures that allow an analytical solution to be computed for the error matrices and in some cases the covariance matrices.
- The group discussed the changes that would need to be made to the current implementation to adapt PPML to handle ordinal data. Parallelizing several aspects of PPML was also discussed.
- The enormous speedup PPML could possibly offer Brain Image Data Models was discussed. These models would need to be re-written as RAM models but could experience enormous speedup improvements under PPML.
- It was decided that Dan will integrate his current progress with PPML into the OpenMx trunk. First he will incorporate the front PPML code. Then he will address moving some of the PPML code to the back end of OpenMx.
- The group also discussed speedup in computing the expected covariance that could be gained from the implementing the algorithm described by Boker et al. here. [5]
- Mike Neale discussed his progress with understanding SADMVN and adapting the current FORTRAN implementation to a thread-safe version. Ultimately, it was decided that the group either needed to: (1) find a FORTRAN expert to decipher and implement a thread-safe SADMVN or (2) find an qualified C Programmer who also understand the statistical routines in SADMVN to create a thread-safe C version of SADMVN that could be deployed onto a GPU to gain additional speedup. A graduate student in need of a Master's Thesis was discussed for this role.
- The group discussed implementing Weighted Least Squares and Full Information Weighted Least Squares in OpenMx and the possible tricks and pitfalls that could arise during implementation. Fortunately, we will be able to confirm with Mplus and Classic Mx to ensure a correct implementations.
- The group discussed funding opportunities. In particular the group discussed the relationship of computing likelihoods from NIH and Dave Evans work with Efficient Privacy-Preserving Biometric Identification. [6]