Open nicholasloveday opened 5 months ago
It's not immediately clear how this can be done since the output is a dictionary of numpy arrays. This will require a major rewrite to replace the existing function. The alternative is that we create a second, simpler version that just works for targeting the mean.
It's also worth noting that scipy has an implementation which we should compare with the sklearn version.
Currently data is brought into memory in the isotonic regression calculation. We should see if we can avoid this.