An Empirical Investigation of the Use of a Neural Network Committee for Identifying the Streptococcus Pneumoniae Growth Phases in Batch Cultivations

摘要

Streptococcus pneumoniae is a bacterial pathogen that causes many life-threatening diseases and an effective vaccine against this pathogen is still subject of research. These bacteria grow with low carbon dioxide production, which hinders the application of exhaust gas composition for on-line process monitoring. This work investigates the proposal of a committee of neural networks for identifying Streptococcus pneumoniae growth phases, to be used for on-line state inference. The committee results as well as the accuracy for predicting the culture phases are compared to the results of a unique neural network, for different input variables. The best configuration for the software was: a committee of three NN trained with two input attributes (optical density and mass of alkali solution), 200 epochs of training and log sigmoid as the activation function in the hidden layer as well as in the output layer.