Closed JulianBiesheuvel closed 4 months ago
Done. But now .appy and .groupby are used. Especially apply can still be a performance bottleneck. It is worth looking into using Numba or similar in the future. A new issue will be made for this with a low priority, see issue #39.
In the custom loss function, for example, the for-loop should be rewritten/removed by more efficient code. Options are vectorizing, or adding
@jit
for Numba usage.