. Author manuscript; available in PMC: 2017 Aug 24.

Published in final edited form as: J Comput Aided Mol Des. 2016 Aug 24;30(8):595–608. doi: 10.1007/s10822-016-9938-8

Table 1.

Graph convolution model hyperparameters

Group	Hyperparameter	Default value
Input	Maximum number of atoms per molecule	60
Input	Maximum atom pair graph distance	2
Weave	Number of Weave modules	1
	(A → A)₀ convolution depth	50
	(A → P)₀ convolution depth	50
	(P → P)₀ convolution depth	50
	(P → A)₀ convolution depth	50
	(A → A)₁ convolution depth	50
	(P → P)₁ convolution depth	50
Reduction	Final atom layer convolution depth	128
Reduction	Reduction to molecule-level features	Gaussian histogram
Post-reduction	Fully-connected layers (number of units per layer)	2000, 100
Training	Batch size	96
	Learning rate	0.003
	Optimization method	Adagrad