yeast cells classification by kohonen work参考-论文.pdf

IFAC DECOM-TT 2004 Copyright © IFAC
AutomaticAUTOMATIC Systems for SYSTEMS Building the FOR Infrastion neural
cells, form of cells , etc. The established relations network is that the Kohonen network trained in an
between values of morphophysiological parameters, unsupervised mode. This means that the Kohonen
technological process and intensity of multiplying network is presented with data, but the correct output
and accumulation of biomass impose an automatic corresponding to these data is not specified. Using
computation of these parameters. For this purpose an the Kohonen network the data can be classified into
automatic complex for calculation of the groups.
morphophysiological parameters has been created
(Mitev and Popova, 1989) and a mathematical model
based on these parameters has been proposed in the 2. CALCULATION OF THE PARAMETERSUSED
case of continious yeast cultivation (Mitev and FOR CLASSIFICATION
Popova, 1995). An adaptive control algorithm for the
biomass concentration, based on the model, is The data used for classification have been obtained
developed (Popova, and Patarinska, 2001). Then the by the following procedure (Mitev and Popova,
yeast cells are classified by multilayer perceptron 1989):
with back-propagation learning algorithm in four 1. A sample is taken from the fermentors.
grou