SPORF
SPORF copied to clipboard
neurodata
This is the implementation of Sparse Projection Oblique Randomer Forest
The current sampling scheme just randomly places -1s and +1s in the projection matrix. Therefore it is possible to get redundant columns. Evaluating the same split directions multiple times is...
like this: https://www.rdocumentation.org/packages/grf/versions/0.10.1/topics/tune_causal_forest for example, this is the key paper on the topic: https://dl.acm.org/citation.cfm?id=2177404 though there are others more recently.
The following line should be added to the roxygen comments `#' @family Random matrix generation functions`
@jovo @randalburns Below is a tentative list of things we may want to implement that currently haven't been, along with *conservative* estimates of the amount of time (assuming 100% effort)...
I came across a problem when I gave `Predict` data from only one class when the trained forest had 2 classes.
https://github.com/neurodata/R-RerF/blob/staging/R/BuildTree.R#L138