We present a model of a bidirectional three-layer neural network with sigmoidal units, which can be trained to learn arbitrary mappings. We introduce a bidirectional activation-based learning algorithm (BAL), inspired by O'Reilly's supervised Generalized Recirculation (GeneRec) algorithm that has been designed as a biologically plausible alternative to standard error backpropagation. BAL shares several features with GeneRec, but differs from it by being completely bidirectional regarding the activation propagation and the weight updates. In pilot experiments, we test the learning properties of BAL using three artificial data sets with binary patterns of increasing complexity. © 2013 Springer-Verlag Berlin Heidelberg.
CITATION STYLE
Farkaš, I., & Rebrová, K. (2013). Bidirectional activation-based neural network learning algorithm. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 8131 LNCS, pp. 154–161). https://doi.org/10.1007/978-3-642-40728-4_20
Mendeley helps you to discover research relevant for your work.