Modeling tool using neural networks for l(+)-lactic acid production by pellet-form Rhizopus oryzae NRRL 395 on biodiesel crude glycerol

Table 1 Comparison of results

Training method	Number of layers	Number of neurons on each layer	Mean squared error
Levenberg–Marquardt	2	15	0.15
Levenberg–Marquardt	3	15	0.51
Levenberg–Marquardt	4	15	0.157
Levenberg–Marquardt	2	20	1.7
Levenberg–Marquardt	3	20	0.36
Levenberg–Marquardt	4	20	0.08
Levenberg–Marquardt	2	25	0.79
Levenberg–Marquardt	3	25	0.04
Levenberg–Marquardt	4	25	184.5
Quasi-Newton	2	15	244.23
Quasi-Newton	3	15	781.88
Quasi-Newton	4	15	482.86
Quasi-Newton	2	20	351.43
Quasi-Newton	3	20	499.8
Quasi-Newton	4	20	217.64
Quasi-Newton	2	25	431.66
Quasi-Newton	3	25	172.11
Quasi-Newton	4	25	898.75
Scaled Conjugate Gradient	2	15	244.23
Scaled Conjugate Gradient	3	15	781.88
Scaled Conjugate Gradient	4	15	482.86
Scaled Conjugate Gradient	2	20	351.43
Scaled Conjugate Gradient	3	20	499.8
Scaled Conjugate Gradient	4	20	217.64
Scaled Conjugate Gradient	2	25	431.66
Scaled Conjugate Gradient	3	25	172.11
Scaled Conjugate Gradient	4	25	898.75
Fletcher–Powell	2	15	244.23
Fletcher–Powell	3	15	781.88
Fletcher–Powell	4	15	482.86
Fletcher–Powell	2	20	351.43
Fletcher–Powell	3	20	499.8
Fletcher–Powell	4	20	217.64
Fletcher–Powell	2	25	431.66
Fletcher–Powell	3	25	172.11
Fletcher–Powell	4	25	898.75
Polak–Ribiere	2	15	244.23
Polak–Ribiere	3	15	781.88
Polak–Ribiere	4	15	482.86
Polak–Ribiere	2	20	351.43
Polak–Ribiere	3	20	499.8
Polak–Ribiere	4	20	217.64
Polak–Ribiere	2	25	431.66
Polak–Ribiere	3	25	172.11
Polak–Ribiere	4	25	898.75

ISSN: 2661-801X