Fig. 2: Comparison of radial distribution functions (RDF) resulting from simulations with an atomistic (AA) or coarse-grained (CG) model and corresponding 2-body relevance. | Nature Communications

Fig. 2: Comparison of radial distribution functions (RDF) resulting from simulations with an atomistic (AA) or coarse-grained (CG) model and corresponding 2-body relevance.

From: Peering inside the black box by learning the relevance of many-body functions in neural network potentials

Fig. 2

Panels a, c correspond to water and b, d to methane models. Panels a, b show the results for PaiNN-based, and c, d for SO3Net-based models. The relevance in arbitrary units, shown in red, is normalized by the absolute total relevance over the number of walks of the respective model and rescaled for each model type. A negative value implies a stabilizing interaction as the model output is the energy. Relevance values are averaged in 75 bins across the distance range, the corresponding average is shown in a solid line and shaded regions correspond to the standard deviation in each bin.

Back to article page