The Evidence Framework Applied to Classification Networks
- 1 September 1992
- journal article
- Published by MIT Press in Neural Computation
- Vol. 4 (5) , 720-736
- https://doi.org/10.1162/neco.1992.4.5.720
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.Keywords
This publication has 8 references indexed in Scilit:
- Information-Based Objective Functions for Active Data SelectionNeural Computation, 1992
- Exact Calculation of the Hessian Matrix for the Multilayer PerceptronNeural Computation, 1992
- A Practical Bayesian Framework for Backpropagation NetworksNeural Computation, 1992
- Bayesian InterpolationNeural Computation, 1992
- Query-based learning applied to partially trained multilayer perceptronsIEEE Transactions on Neural Networks, 1991
- Neural net algorithms that learn in polynomial time from examples and queriesIEEE Transactions on Neural Networks, 1991
- Sequential updating of conditional probabilities on directed graphical structuresNetworks, 1990
- Learning algorithms and probability distributions in feed-forward and feed-back networksProceedings of the National Academy of Sciences, 1987