Implicitly and Explicitly ConstrainedOptimization Problems for Training of Recurrent Neural Networks
2014 (English)Conference paper (Refereed)
Training of recurrent neural networks is typically formulated as unconstrained optimization problems. There is, however, an implicit constraint stating that the equations of state must be satisfied at every iteration in the optimization process. Such constraints can make a problem highly non-linear and thus difficult to solve. A potential remedy is to reformulate the problem into one in which the parameters and state are treated as independent variables and all constraints appear explicitly. In this paper we compare an implicitly and an explicitly constrained formulation of the same problem. Reported numerical results suggest that the latter is in some respects superior.
Place, publisher, year, edition, pages
2014. 461-466 p.
IdentifiersURN: urn:nbn:se:liu:diva-125064ISBN: 978-287419095-7OAI: oai:DiVA.org:liu-125064DiVA: diva2:902683
22nd European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning. Bruges (Belgium), 23-25 April 2014,