Deep Learning for direct Dark Matter search with nuclear emulsions

We propose a new method for the discrimination of sub-micron nuclear recoil tracks from an instrumental background in fine-grain emulsions used directional dark matter search. The proposed uses 3D Convolutional Neural Network, whose parameters are optimised by Bayesian Unlike previous studies focused on extracting information, we focus signal/background separation exploiting polarisation dependence Localised Surface Plasmon Resonance phenomenon. Comparing with conventional cut-based approach shows significant boost reduction factor given signal efficiency.