The Multi Layer Perceptron Network consists of an input layer, one or more hidden layers and an output layer. The input layer neurons are trained and modelled under potential parameters of significant influence for the event of interest. For instance, for the analysis of soil erosion, input layers comprised of spatial datasets such as the digital elevation model and its derivatives, land cover, soil composition and regional climate data. The exposure of a particular landscape to soil erosion was analyzed and it was possible to then present the modelled results as a probability map or as a map that illustrated quantitative attribute (e.g. kilograms of soil relocated per hectare).

  MLP structure Figure 1: Network structure
MLP for soil erosion  
The results are validated and the weights are corrected for each training epoch to match the training pattern.

For more information, documents on the various fields of application are available upon request.