Fig. 6: Training logs for the feedforward network solving the inverse problem.
From: Inferring topological transitions in pattern-forming processes with self-supervised learning
The mean value baseline corresponds to ΔA and the theoretical minimum to \(\bar{\Delta }\). Here, we used a network with two hidden layers, having 1024 and 512 neurons, respectively, \({{\mathtt{Dense}}}_{512}^{{{{\rm{ReLU}}}}}\times {{\mathtt{Dropout}}}_{0.25}\times {{\mathtt{Dense}}}_{1024}^{{{{\rm{ReLU}}}}}\times {{\mathtt{Dropout}}}_{0.25}\times {{\mathtt{Dense}}}_{2}^{{{{\rm{linear}}}}}\).
