Learning Recurrent Neural Networks with Hessian-Free Optimization
JamesMartensJMARTENS@CS.TORONTO.EDUIlyaSutskeverILYA@CS.UTORONTO.CAUniversityofToronto,CanadaAbstractInthisworkweresolvethelong-outstandingproblemofhowtoeffectivelytrainrecurrentneu-ralnetworks(RNNs)oncomplexanddifficultsequencemodelingproblemswhichmaycon-tainlong-termdatadepende