| foceiGradPooledDirect_ | R Documentation |
The same C++ core the fit's own gradient uses ('gradPooledCore'), but with the point passed in rather than read out of 'op_focei'. 'est="vae"' with 'nonMuTheta="grad"' needs exactly this: its M-step evaluates the gradient at a theta and an encoder eta matrix that are not the inner problem's, and at an omega that changes every step. Requires 'foceiGradPooledSetupLoad_()' first, and a live FOCEi inner problem (the shared pool, 'rxVaeOuter', and the theta/eta par_ptr maps all come from it).
foceiGradPooledDirect_(thVals, ebes, Oi, dOiEst, tr28, cores)
thVals |
natural-scale theta, in ntheta order |
ebes |
nsub x neta matrix of etas to take the gradient at |
Oi |
inverse of the current Omega |
dOiEst |
list of d(Omega^-1)/d(estimation-scale omega element) |
tr28 |
the matching trace terms |
cores |
thread count |
natural-scale gradient (thetas, sigmas, omegas), or NULL if it declined
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.