Optimize input as fourier-wavelet with a.e. part and discrete part then apply them into original.
We bet the using principal axis of viewpoint as continuity concludes Wavelet(Wavelet(Fourier+Discrete)+Discrete)+Discrete can cover any of the f. So we optimize and use the series of oritinal input to such form then, we apply then into the next steps. This is the analogy to find the pattern in randtools meaning, they causes the map from combination of the combination to combination. So if the transform is wide enough to cover such combinations, the transform can causes (2^omega)^(2^omega) to omega order.
Because we have partial p0, p1, p2 implementation, so this implementation can causes to exclude the noises from input streams, this can be done by catgp partially, so we don't need this now.
Like bitsofcotton/p3:README.md:Tips on another solution, the p3 another solution also causes another solution for this. This is because total countup 6 layers on p7 also each layer have the form tan<a,x> , so this is compatible to p7 with some of the transformation. So we should implement them first.
Nor, for simplify and non compatible reform this problem, we can use Decompose class mother wavelet predictions.