Large scale high energy neutrino detectors measure through going muons, which are reconstructed
to infer the arrival direction of the originating neutrinos on the sky. In addition to this
directional information, it is important to infer the energy of these muons, needed for discriminating
astrophysical signals from atmospheric backgrounds in a diffuse analysis, or for increasing
the ability to distinguish a point source of neutrinos from the backgrounds. Various methods exist
to infer the muon energy, based on the fact that the light produced by each muon is on average
proportional to its energy. Here, we describe the Edepillim algorithm, which interprets the entire
pattern of stochastic muon energy losses along each track. The probability of each loss is used
in a likelihood approach, taking account of the decreasing muon energy along the track, leading
to a one-parameter fit for the initial muon energy. In this presentation, we will review the basic
performance of this algorithm using idealised simulated muons, and discuss how the method
might perform under real-world circumstances where the losses cannot be known precisely, but
must be inferred by an unfolding of the observed light signals across a large array of detectors in
a medium such as ice or water.