home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.ai.neural-nets
- Path: sparky!uunet!mcsun!sunic!dkuug!daimi!buhr
- From: buhr@daimi.aau.dk (Morten Buhr)
- Subject: Re: Meaning of deltas in hidden units
- Message-ID: <1992Dec16.072544.19664@daimi.aau.dk>
- Keywords: BP, delta values
- Sender: news@daimi.aau.dk
- Organization: DAIMI: Computer Science Department, Aarhus University, Denmark
- References: <1992Dec15.162420.22854@Informatik.TU-Muenchen.DE>
- Date: Wed, 16 Dec 92 07:25:44 GMT
- Lines: 30
-
- baginski@Informatik.TU-Muenchen.DE (Boris Baginski) writes:
-
-
- >Everbody knows the meaning of the delta values in the backprop algorithm
- >for output units (difference between target and activation) and the use
- >of the deltas to adjust the internal weights.
-
- >But the question is: what do those values express for the hidden units and
- >for the input units (can be calculated easily, are not used) ?
- >We explored that these deltas take on values up to 50 (+/-) or so, but we only
- >use inputs and targets and activation states in the range of 0-1.
- >So it`s not simply possible to say these values express a realistic target minus
- >activation for the hidden/input units.
-
- >thank you, regards,
- >BORIS BAGINSKI
- >adress: baginski@informatik.tu-muenchen.de
-
- I can't quite see how you calculate the deltas for hidden nodes, since
- you don't have a target to compare your nodes activity with. That is exactly
- what the deltas in backpropagation are used for: to backpropagate the error
- through the network.
-
- /Morten
-
- --
- Morten Buhr / e-mail buhr@daimi.aau.dk
- Computer Science Department
- Aarhus University
- Denmark
-