Optimal signal denoising algorithms for photon-counting experiments

Photon-counting experiments are ubiquitous in science and engineering, particularly in

astronomy and health informatics. Unknown signal extraction from a noisy background is an

outstanding problem that is extensively discussed in these fields, but due to its inherently

degenerate nature, there is no definite solution. We present a computational and accuracy

benchmark of multiple classes of solutions proposed for this problem, specifically kernel,

nearest neighbor, and change-point density estimators, and study the advantages and

limitations of each methodology.