Attenuated Total Reflectance Fourier Transform Infrared Spectroscopy and Artificial Neural Networks Applied to Differentiate Escherichia coli 𝑝𝑎𝑝𝐺⁺/𝑝𝑎𝑝𝐺⁻ Strains

Table 1

The results obtained for ANNs classifying E. coli 𝑝𝑎𝑝𝐺⁺ strains.

ID	Topology	Quality	Quality	Quality	Learning	Error	Activation function	Activation function
		(learning)	(testing)	(validation)	algorithm	function	(hidden neurons)	(output neurons)
1	MLP 19-5-2	97,02%	93,51%	99,07%	BFGS 89	SOS	Tanh	Linear
2	MLP 19-6-2	86,50%	87,04%	82,40%	BFGS 40	SOS	Tanh	Linear
3	MLP 19-7-2	94,04%	90,74%	89,81%	BFGS 55	SOS	Tanh	Linear
4	MLP 19-8-2	96,42%	92,59%	88,89%	BFGS 103	SOS	Tanh	Linear

MLP: multilayer perceptron, BFGS: Broyden-Fletcher-Goldfarb-Shanno algorithm, SOS: sum of squares, Tanh: hyperbolic tangent, Linear: linear function.