Was wondering if I can get some clarity on how to the “Fraction of FCs to use for alignment” parameter impacts hyperparameter optimization. I found a couple of posts by @hsnyder in response to other RBMC issues which explains the parameter clearly, with suggestions to decrease that parameter from 0.5 to 0.3 in cases where (1) where reducing VRAM usage is desirable or (2) not enough signal in the “top half of the resolution range.”
Are there other scenarios that merit tuning this parameter such as improving trajectory computation? Are there any risks to increasing this parameter, for e.g. noisier cross-validation?