home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!trwacs!erwin
- From: erwin@trwacs.fp.trw.com (Harry Erwin)
- Newsgroups: comp.ai.neural-nets
- Subject: Training Networks on Chaotic Time Series
- Keywords: chaos backpropagation
- Message-ID: <670@trwacs.fp.trw.com>
- Date: 23 Jul 92 11:21:46 GMT
- Organization: TRW Systems Division, Fairfax VA
- Lines: 20
-
- I've noticed a lot of interest over the last year in using neural networks
- to predict chaotic time series. I'd like to remind people of a result
- posted here about a year ago. During the summer of 1991, my son, Jeremy
- Erwin, conducted a parametric study under my direction where he trained a
- standard backpropagation network on chaotic time series to see whether the
- network was learning the training sets or generalizing. The chaotic time
- series were generated using the logistic map with the multiplicative
- coefficient parametrically varied over the chaotic region. He discovered
- that the network was clearly learning the training set and not
- generalizing. This was particularly clear for the more highly chaotic
- cases since the effectiveness of the network was significantly reduced for
- a fixed training period. It was also clear that the network "hadn't a
- clue" for regions of the chaotic process that were not represented in the
- training set.
-
- Cave canem,
- --
- Harry Erwin
- Internet: erwin@trwacs.fp.trw.com
-
-