Abstract
We show that in supervised learning from a particular data set Bayesian model selection, based on the evidence, does not optimise generalization performance even for a learnable linear problem. This is achieved by examining the finite size effects in hyperparameter assignment from the evidence procedure and its effect on generalisation. Using simulations we corroborate our analytic results and examine an alternative model selection criterion, namely cross-validation. This numerical study shows that in the learnable linear case for finite sized systems leave one out cross-validation estimates correlate more strongly with optimal performance than do those of the evidence.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Bruce A D and Saad D, Statistical mechanics of hypothesis evaluation. J. of Phys. A: Math. Gen. Vol. 27 (1994), pp 3355–3363
MacKay D J C, Bayesian interpolation, Neural Comp. Vol. 4 (1992), pp415–447.
Marion G and Saad D, A statistical mechanical analysis of a Bayesian inference scheme for an unrealizable rule. J. of Phys. A: Math. Gen. Vol. 28 (1995), pp2159–2171.
Marion G and Saad D, Finite size effects in Bayesian model selection and generalization. J. of Phys. A: Math. Gen. Vol. 29 (1996), pp5387–5404.
Sollich P, Finite-size effects in learning and generalization in linear perceptrons. J. of Phys. A: Math. Gen. Vol. 27 (1994), pp7771–7784.
Stone M, An asymptotic equivalence of choice of model by cross-validation and Akaikes criterion, J. Roy. Statist. Soc. Ser. B Vol. 39 (1977), pp44–47.
Watkin T L H, Rau A, Biehl M, The statistical mechanics of learning a rule. Reviews of Modern Physics Vol. 65 (1993), pp499–556.
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 1997 Springer Science+Business Media New York
About this chapter
Cite this chapter
Marion, G., Saad, D. (1997). Data Dependent Hyperparameter Assignment. In: Ellacott, S.W., Mason, J.C., Anderson, I.J. (eds) Mathematics of Neural Networks. Operations Research/Computer Science Interfaces Series, vol 8. Springer, Boston, MA. https://doi.org/10.1007/978-1-4615-6099-9_44
Download citation
DOI: https://doi.org/10.1007/978-1-4615-6099-9_44
Publisher Name: Springer, Boston, MA
Print ISBN: 978-1-4613-7794-8
Online ISBN: 978-1-4615-6099-9
eBook Packages: Springer Book Archive