The mlp-bgd-1 method does regression and classification using a
multilayer perceptron neural network with one hidden layer, trained by
batch gradient descent. The training data is divided into four equal
parts, and four training runs are done, each on 3/4 of the data. The
1/4 remaining in each run is used as a validation set to determine the
best stopping point. Predictions are made with the resulting ensemble
of four networks.
See the notes for more details.
This method uses the software for flexible Bayesian modeling written
by Radford Neal (release of 1997-06-22), available from
Radford Neal's home page.
Directory listing of the results (and source
files) available for the mlp-bgd-1 method. Put the desired files
in the appropriate methods directory in your delve hierarchy and
uncompress them with using the "gunzip *.gz" command and
untar them using "tar -xvf *.tar".
Neal, R. M. (1998) ``Assessing relevance determination methods using DELVE'',
to appear in C. M. Bishop (ed) Generalization in Neural Networks and
Machine Learning, Springer-Verlag.