Current makePDFmap implemented only base statistics -mean, median, rms
The analytical fits with function like (gauss, gauss+pol1) should be also used as in the original C++ implementation on the NDpipeline which we still use.
The performance of the analytical fits vers ML inference should be compared.
If faster will be used on the server and also the client (once the ONNX client is ready).
Current makePDFmap implemented only base statistics -mean, median, rms The analytical fits with function like (gauss, gauss+pol1) should be also used as in the original C++ implementation on the NDpipeline which we still use.
The performance of the analytical fits vers ML inference should be compared. If faster will be used on the server and also the client (once the ONNX client is ready).
Benchmark:
target O(8-128) CPU cores, GPU