Piñeiro, Roberto C.

Profile Picture

Filters

20041
Reset filters

Settings

Citations

Publication Search Results

Now showing 1 - 1 of 1
  • Thumbnail Image
    Publication
    Evolutionary learning methods for multilayer morphological perceptron
    (2004) Piñeiro, Roberto C.; Ortiz-Álvarez, Jorge; College of Engineering; Rodríguez, Néstor; Rodríguez, Domingo; Department of Electrical and Computer Engineering; Acuña, Edgar
    This thesis describes three compressive learning algorithms for multilayer morphological perceptrons. The three algorithms are based on evolutionary algorithms: direct encoding method, indirect encoding method, and catesian genetic programming method. The direct encoding method uses adaptive mutation as the genetic algorithm approaches convergence to fine tune network parameters to reach optimal values. In addition, the algorithms use a special fitness function which penalize those networks with redundant neurons. The training of the neural network using the indirect encoding method is done by finding the solution without considering the exact connectivity of the network. Looking for the set of connection weights and network architecture in a reduced search space, this simple, but powerful, training algorithm is able to evolve to a feasible solution using up to three layers suficient to perform most pattern classification. The last method uses Cartesian genetic programming to evolve network architecture and connection weights simultaneously. The resulting program consists of the multilayer morphological perceptron, which is able to classify patterns received as the inputs. The algorithm introduces the use of the morphological neuron computational model as the function used by the generated programs. Prototypes were implemented using Matlab, and tested using data sets used previously by other researchers.