Dependency graph and topological sorting could give better performance for multiple feature calculation.
"Optional" feature dependencies can be implemented, for example Median can mark the sorted magnitude array as an optional dependency which will be used if it is required by some other feature (for example AndersonDarling) only
Another useful abstraction which can be introduced is dependency-only features, for example sorted magnitude array or an approximate FFT of light curve, which can be used by Periodogram and by autocorrelation function #19.
Could be related to enum-dispatch, see light-curve/light-curve#60. Or to (de)serialisation, see light-curve/light-curve#66
hombit
commented
2 years ago
Dependency graph and topological sorting could give better performance for multiple feature calculation.
"Optional" feature dependencies can be implemented, for example
Mediancan mark the sorted magnitude array as an optional dependency which will be used if it is required by some other feature (for exampleAndersonDarling) onlyAnother useful abstraction which can be introduced is dependency-only features, for example sorted magnitude array or an approximate FFT of light curve, which can be used by
Periodogramand by autocorrelation function #19.Could be related to enum-dispatch, see light-curve/light-curve#60. Or to (de)serialisation, see light-curve/light-curve#66