Table 2.

Results for different combinations of neural network architectures and output layers.

Architecture	Output	RMS [cm] (R < 60 cm)	RMS [cm] (R > 60 cm)
4-layer MLP	Linear	0.2363	0.3558
4-layer MLP	Constrained	0.2331	0.3406
DiNN	Linear	0.2314	0.3470
DiNN	Constrained	0.2280	0.3419

4-layer MLP with linear output is the same as the MLP model discussed in Section 5 and DiNN with constrained output is the same as the model shown in Figure 5. The RMS is calculated on models' prediction of simulated 100 electron S2s and averaged over x,y dimension.