Diagram of the model behind AICE: artificial neural networks. For each target molecule (H₂O, CO, CO₂, CH₃OH, NH₃, and CH₄), a multilayer perceptron is used to predict the corresponding molecular fraction. The IR ice spectrum in absorbance is fed to the model, which transforms the input through a series of steps or layers. In each one, a linear combination of the previous values (x_i) is followed by the application of an activation function (ϕ), obtaining a_j = ϕ(∑_i w_ji x_i + b_j), with w_ji and b_j being the weights and bias of the layer j, respectively. We note that the actual neural networks used in AICE are larger than depicted in this scheme (see Section 2.2 and Fig. 2 for more details).