Figure 2 | Scientific Reports

Figure 2

From: Parsimonious neural networks learn interpretable physical laws

Figure 2

(a) PNN model 1 RMSEs on the training/validation/test sets compared to the feed forward network (b) We see that energy conservation between PNN1 and the verlet integrator is comparable (TE: total energy) (c) Forward and reverse trajectories generated by PNN1 show good reversibility (d) A visualization of PNN model 1 found by the genetic algorithm, attempting to predict positions and velocities one step ahead.

Back to article page