Table 3.

Position of nodes in structured training sets.

Node	r (Å)	Theta (degrees)	Phi (degrees)
1	2.65	90	0
2	2.65	90	105
3	2.65	17	−128
4	2.65	163	−128
5	3.18	90	172
6	3.18	90	−68
7	3.18	30	53
8	3.18	150	53

Nodes 1–4 were placed on the vertices of a tetrahedron centered on the oxygen of the central water molecule, at a distance from the center that approximately corresponds to the first peak in the water radial distribution function (see Fig. 3c to see how nodes 1–4 are positioned relative to the nearest four oxygens to the oxygen of the central water molecule). Nodes 5–8 were positioned similarly, at a slightly greater distance, as the structure they are positioned to account for occurs at a slightly greater distance (see Fig. 5c to see how nodes 5–8 are positioned relative to the fifth to eighth nearest oxygens to the oxygen of the central water molecule). Nodes 1–4 were used in the creation of Structured 1a and Structured 2a training sets (and their hydrogen‐ centered equivalents), whereas all eight nodes were used in the construction of Structured 1b and Structured 2b training sets (and their hydrogen‐ centered equivalents).