Figure 7.

This Figure shows the implementation of the algorithm initially proposed by Yazdani et al. [2020] in the context of this thesis. The neural network takes the input time of a measurement as an input and outputs the seven different state variables. These state variables, together with the inferred parameters, the computational model, and the observed data are used to compute the different loss functions. The gradient of these loss functions is then used to optimize the inferred parameter and the neural network. AD stands for automatic differentiation.