Comparison of the distribution of normalised prediction errors to a standard normal distribution. The histograms represent the distributions of (y^pred − y^sim)/σ^pred for M_dust (top panel), and T_dust (bottom panel), for a test data set predicted by the trained neural network with the entire suite of JWST filters, in S3. The dotted curves represent the standard normal distributions (i.e. N(0,1)). The solid curves are the normal distributions fitted to each of the histograms with µ = 0.04 and σ = 0.94 for M_dust, and µ = −0.001 and σ = 0.76 for T_dust.