Last updated: 2018-07-20

workflowr checks: (Click a bullet for more information)
Expand here to see past versions:


Investigation 1. The FLASH objective function can behave very erratically.

Investigation 2. This problem only occurs when using ebnm_pn, not ebnm_ash.

Investigation 3. The objective can continue to get worse as loadings are repeatedly updated. Nonetheless, convergence takes place (from above!).

Notes 1. Notes on computing the FLASH objective function. I derive an explicit expression for the KL divergence between prior and posterior.

Notes 2. An alternate algorithm for optimizing the FLASH objective, using the explicit expression derived in the previous note.

Investigation 4. The alternate algorithm yields much different results from FLASH. In particular, the fits of the priors on the loadings are visibly different.


This reproducible R Markdown analysis was created with workflowr 1.0.1