* test.stats.R
is used to produce test statistics for dmt package;
* test.stats.txt
test statistics; correlations between true model parameters and
estimates given by dmt.
============================================
DESCRIPTION OF COMPLETED TEST CASES
The tests are run with the different covariance structures and
latent variable dimensionalities.
tests/ppca.R # probabilistic PCA (1 and 2 data sets)
tests/pfa.R # probabilistic factor analysis (1 and 2 data sets)
tests/pcca.R # probabilistic CCA (2 data sets)
tests/depmod.R # prob. CCA
tests/depmod.simcca.R # prob. SimCCA (Wx = Wy)
tests/pcca.nnW.R # prob. CCA with non-negative W
tests/psimcca.nnW.R # prob. SimCCA (Wx = Wy) with non-negative W
tests/drcca.R # drCCA functions
OK regularization: yes -> W >=0 (alpha/beta?)
OK regularization: no
OK matched: yes
OK matched: no
==============================================
TODO TESTS
- currently measuring correlation, add accuracy measures
- soft regularization on W (tuning priors$Nm.wx.wy.sigma)
- pcca.with.isotropic.margins
- speedup tests
==============================================
TODO FEATURES
- Sparsity regularization for W and/or z (Gaussian / elastic net / IBP / ...)
- T priors when dx != dy
- Wy fixed, only optimize Wx (regression-type models)