The residual_chaser_factory allows you to use a skater f2 to predict the errors of f1, thus creating a potentially more accurate skater than f1. It also illustrates composition.
You can use it like this:
from timemachines.skaters.proph.prophskaterssingular import fbprophet_univariate as f1
from timemachines.skaters.simple.hypocratic import quickly_hypocratic as f2
from timemachines.skatertools.composition.residualcomposition import residual_chaser_factory
def f3(y,s,k,a,t,e,r):
return residual_chaser_factory(y=y, s=s, k=k, a=a, t=t, e=e, f1=f1,f2=f2,r2=r)
Then use f3 just as you would any skater:
import numpy as np
y = np.random.randn(100)
for yi in y:
xi, x_std, s = f3(y=yi, s=s, k=3)
See skaters/proph/prophetskaterscomposed.py or skaters/simple/hypocratic.
-+-
Documentation map