Suomeksi
Comments on the IJCNN'98 article
There is a Bayesian interpretation for the algorithm presented in the paper:
- G. E. Hinton and
D. van Camp. (1993)
Keeping neural networks simple by minimizing the description length
of the weights. In Proceedings of COLT-93.
- D. MacKay. (1995)
Probable networks and plausible predictions - a review of practical
Bayesian methods for supervised neural networks. Network 6(3),
pp. 469-505.
- D. MacKay. (1995)
Ensemble learning and evidence maximization.
Available only as a Post Script version.
- D. MacKay. (1995?)
Developments in Probabilistic Modelling with Neural Networks - Ensemble Learning.
Available only as a Post Script version.
According to the Bayesian interpretation, my paper presents an algorithm for
approximating the posterior density. If the approximation were a diagonal
Gaussian density, the term 1/2 ln 12 v_i in equation 15 should read
1/2 ln 2 pi e v_i. I recommend to use the value 2 pi e (about 17.08) instead
of 12.