?url_ver=Z39.88-2004&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Adc&rft.title=Why+feed-forward+networks+are+in+a+bad+shape&rft.creator=Smagt%2C+P.+van+der&rft.creator=Hirzinger%2C+G.&rft.subject=Neural+Nets&rft.subject=Statistical+Models&rft.description=It+has+often+been+noted+that+the+learning+problem+in+feed-forward+neural+networks+is+very+badly+conditioned.+Although%2C+generally%2C+the+special+form+of+the+transfer+function+is+usually+taken+to+be+the+cause+of+this+condition%2C+we+show+that+it+is+caused+by+the+manner+in+which+neurons+are+connected.+By+analyzing+the+expected+values+of+the+Hessian+in+a+feed-forward+network+it+is+shown+that%2C+even+in+a+network+where+all+the+learning+samples+are+well+chosen+and+the+transfer+function+is+not+in+its+saturated+state%2C+the+system+has+a+non-optimal+condition.+We+subsequently+propose+a+change+in+the+feed-forward+network+structure+which+alleviates+this+problem.+We+finally+demonstrate+the+positive+influence+of+this+approach.&rft.publisher=Springer+Verlag%2C+Berlin&rft.date=1998&rft.type=Conference+Paper&rft.type=NonPeerReviewed&rft.format=application%2Fpostscript&rft.identifier=http%3A%2F%2Fcogprints.org%2F495%2F2%2FSmaHir98b.ps&rft.identifier=++Smagt%2C+P.+van+der+and+Hirzinger%2C+G.++(1998)+Why+feed-forward+networks+are+in+a+bad+shape.++%5BConference+Paper%5D++++(In+Press)++&rft.relation=http%3A%2F%2Fcogprints.org%2F495%2F