Table 1. Results of the Trained Models on the Validation and Test Data Sets for Various Architectures^a.

	Validation			Test
Model	↑Acc.	↓Loss	↑r2	↑Acc.	↓Loss	↑r2
GCSConv	0.44	0.23	0.76	0.46	0.23	0.77
GCSConv (no edges)b	0.60	0.56	0.44	0.60	0.54	0.47
GCSConv (extra nodes)c	0.75	0.72	0.31	0.75	0.66	0.34
GCNConv	0.59	0.46	0.57	0.59	0.45	0.55
Dense	0.00	1.00	–1.0	0.00	1.00	–1.00

^aData Set is .

^bAll edges were simply set to 1.

^cNumber of nodes was increased from 23 to 291. Evaluation time increases to 14.1 ms per variant.