Architecture and loss functions used in this paper. The input is the total mass and optionally X-ray maps that is then compressed using a convolutional NN (blue encoder) into a 7D latent space. From this latent space, we can get the similarity cluster and distance losses, Equations 1 and 3, or further transform it using a fully connected NN (red classifier) to obtain the classification loss, Equation 2. All losses are then weighted summed using the weights λ_CCLP, λ_dist, and λ_class, Equation 4. See Figure B.1 for the full encoder and classifier architecture.