A specific artificial neural network based model for the identification of pollution sources
MetadataShow full item record
CitationAcheli, D., Kouadri, A. ve Namoune, A. (2009). A specific artificial neural network based model for the identification of pollution sources. Maltepe Üniversitesi. s. 136.
The aim of the present work is to suggest an original approach based on the concept of training of an artificial neural network input or IT-Net in order to predict the possible sources of pollution. The IT-Network is composed of three layers of neurons of which one is hidden. The output layer contains m neurons corresponding to the dimension of pollutant concentration noted x. The input layer is composed of one neuron corresponding to the distance between the sources of pollution and the pollutant concentration sensor. Instead of proceeding to a phase of training with five layers of neurons, it appears more interesting to only apply the same phase of training to a part of the network involving three layers. This approach is promising as it extends the back-propagation algorithm as long as the error function is well-defined. The difference between this form in the input training network and multi-layers perceptron is that the input is not necessarily known because it represents sources of pollution searched in a river for example. Therefore during the phase of training, it becomes necessary to adjust not only the external parameters of the network but also the values of the input by minimizing the error of the network output. The training of an IT-Net enables determining the existing relationship between sources of pollution and data of pollutant concentration.
SourceInternational Conference of Mathematical Sciences
- Makale Koleksiyonu 
The following license files are associated with this item: