A model w1 that is more similar to w2 predicts the unknown data (red) better, which results in a better generalization of the model.
The performance improvement of the proposed system over existing systems can be qualified in terms of better generalization ability, preventing weight shifts, fast convergence and robustness against system disturbance.
As is displayed in two formulas above that a positive value (frac{text{I}}{text{C}}) is added to the diagonal of (FF^{P}) or (F^{P} F) in order for better generalization performance.
We conclude that a simpler ANN architecture leads to better generalization in the case of multilevel choice.
In this work, a better generalization capacity were obtained using the back-propagation through time network, which presented an average relative error of 2.2% between the calculated and predicted F values.
In order to have better generalization, the model was designed to realize a supervised learning such that after learning, output for any input vector is the same as the output of a higher order network that implements the same input data set using M̃ number of higher order terms.
In other words, the instances inside the margin band are penalized together with the misclassified instances to reach a better generalization in testing.
It is common to repeat steps (4) to (7) with different partitioning of the training and test sets in order to achieve better generalization with the recognition models. .
All learning curves confirm the fact that the soft decision tree structure is able to provide better generalization in contrast to the canonical hard decision tree structure.
