next up previous contents index
Next: sigmoidalUnit.m, sigmoidalUnitDeriv.m, valueUnit.m, valueUnit2Deriv.m, Up: Training functions and WeightUpdateGlob.m Previous: batchMode.m and batchModeNewton.m

incrementalMode.m and incrementalModeNewton.m

Modules that do the weight update in incremental mode (see Section 3.1.2 for explanation). All the statements in these modules operate on a single input pattern at a time. Therefore although the contents of the file seem to be not very different from the batch mode version, the functionality is different.

Here, when an input pattern is represented to the network, the corresponding error is immediately used to calculate the weight changes and update the network. The weight update operations are contained in a loop iterating for all of the patterns in an epoch. Since only a single pattern is of interest at any particular time, the calculations work on the corresponding columns of the matrices, as opposed to working with the whole matrix in batch mode, mentioned in the previous section.



Cengiz Gunay
2000-06-25