Press to Close

Table 3. Configurations of the best neural networks for each of the five different random partitions of the Californian and Mediterranean data sets. The output layer always contains a single neuron. 'Li' indicates linear transfer functions, 'Lo' logistic transfer functions and 'T' hyperbolic tangent transfer functions.

California data set
Partition First hidden layer Second hidden layer Output layer
1 12Lo, 3T, 13Li 2Lo, 16T 1Lo
2 9Lo, 12T, 8Li 5Lo, 12T, 4Li 1Lo
3 11Lo, 1Li -- 1Lo
4 7Lo, 8T -- 1Lo
5 4Lo, 5T, 6Li -- 1Lo
Mediterranean data set
Partition First hidden layer Second hidden layer Output Layer
1 12Lo, 3T, 13Li -- 1Li
2 13Lo, 5T -- 1Lo
3 14Lo, 7T -- 1Lo
4 3Lo, 7T -- 1Lo
5 31Lo -- 1Lo