The evidence framework applied to classification networks
- Creators
- MacKay, David J. C.
Abstract
Three Bayesian ideas are presented for supervised adaptive classifiers. First, it is argued that the output of a classifier should be obtained by marginalizing over the posterior distribution of the parameters; a simple approximation to this integral is proposed and demonstrated. This involves a "moderation" of the most probable classifier's outputs, and yields improved performance. Second, it is demonstrated that the Bayesian framework for model comparison described for regression models in MacKay (1992a,b) can also be applied to classification problems. This framework successfully chooses the magnitude of weight decay terms, and ranks solutions found using different numbers of hidden units. Third, an information-based data selection criterion is derived and demonstrated within this framework.
Additional Information
© 1992 Massachusetts Institute of Technology. Received 20 November 1991; accepted 18 February 1992. Posted Online March 13, 2008. This work was supported by a Caltech Fellowship and a Studentship from SERC, UK.Attached Files
Published - MACnc92d.pdf
Files
Name | Size | Download all |
---|---|---|
md5:8331d1de5909e9356995a3e8291f0617
|
916.9 kB | Preview Download |
Additional details
- Eprint ID
- 13796
- Resolver ID
- CaltechAUTHORS:MACnc92d
- Caltech Fellowship
- Studentship from SERC, UK
- Created
-
2009-06-25Created from EPrint's datestamp field
- Updated
-
2021-11-08Created from EPrint's last_modified field