Table 2.

Model statistics of biomass prediction in batch bioprocess with values of hidden nodes between 1 and 15 and three different training engines Scaled conjugate gradient, Levenberg-Marquardt and Bayesian Regularization. Model performance was evaluated by RMSE and R². The best performing model is highlighted in bold.

Statistics on model fit for three different training engines	Scaled Conjugate Gradient		Levenberg-Marquardt		Bayesian Regularization
Hidden nodes	RMSE	R2	RMSE	R2	RMSE	R2
1	2.23	0.95	1.29	0.93	1.29	0.93
2	1.81	0.97	0.83	0.96	1.83	0.89
3	3.18	0.96	1.22	0.93	1.07	0.95
4	1.76	0.96	0.94	0.96	1.32	0.93
5	3.17	0.96	1.35	0.93	1.89	0.90
6	1.01	0.96	1.11	0.95	1.63	0.90
7	1.41	0.95	2.17	0.98	1.39	0.92
8	1.02	0.97	0.74	0.98	1.32	0.92
9	2.32	0.96	1.21	0.94	0.86	0.96
10	0.86	0.98	1.18	0.95	1.05	0.95
11	1.76	0.96	1.10	0.95	1.00	0.95
12	1.46	0.96	1.16	0.94	1.32	0.93
13	1.19	0.97	1.32	0.92	1.03	0.95
14	1.14	0.94	1.00	0.96	0.98	0.96
15	2.30	0.97	1.33	0.94	1.14	0.94