fitCyclopsModel() passes a control object to the final Cyclops fit, but the later fold refits in getCV() do not.
This means CV predictions may ignore settings such as threads, seed, tolerance, and maxIterations.
The final trained model should be unaffected. The issue is limited to PLP-generated CV predictions / CV performance summaries for Cyclops logistic models.
Likely fix: pass an optional control argument through createCyclopsModel() to getCV(), and then into the fold-level Cyclops::fitCyclopsModel() call.
fitCyclopsModel()passes acontrolobject to the final Cyclops fit, but the later fold refits ingetCV()do not.This means CV predictions may ignore settings such as
threads,seed,tolerance, andmaxIterations.The final trained model should be unaffected. The issue is limited to PLP-generated CV predictions / CV performance summaries for Cyclops logistic models.
Likely fix: pass an optional
controlargument throughcreateCyclopsModel()togetCV(), and then into the fold-levelCyclops::fitCyclopsModel()call.