Method
- Divide the available data into training and validation sets.
- Use a large number of hidden units.
- Use very small random initial values.
- Use a slow learning rate.
- Compute the validation error rate periodically during training.
- Stop training when the validation error rate "starts to go up".
It is crucial to realize that the validation error is not a good estimate of the generalization error. One method for getting an unbiased estimate of the generalization error is to run the net on a third set of data, the test set, that is not used at all during the training process. The error on the test set gives estimate on generalization; to have the outputs of the net approximate target values given inputs that are not in the training set.
Read more about this topic: Early Stopping
Famous quotes containing the word method:
“It is to be lamented that the principle of national has had very little nourishment in our country, and, instead, has given place to sectional or state partialities. What more promising method for remedying this defect than by uniting American women of every state and every section in a common effort for our whole country.”
—Catherine E. Beecher (1800–1878)
“Women are denied masturbation even more severely than men and that’s another method of control—they’re not taught to please themselves.... Most women—it takes them a while to warm up to the “situation” but once they get into it, I’m sure they’re going to get just as hooked as—well, everyone I know is!”
—Lydia Lunch (b. 1959)
“The most passionate, consistent, extreme and implacable enemy of the Enlightenment and ... all forms of rationalism ... was Johann Georg Hamann. His influence, direct and indirect, upon the romantic revolt against universalism and scientific method ... was considerable and perhaps crucial.”
—Isaiah Berlin (b. 1909)