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:
“Traditional scientific method has always been at the very best 20-20 hindsight. It’s good for seeing where you’ve been. It’s good for testing the truth of what you think you know, but it can’t tell you where you ought to go.”
—Robert M. Pirsig (b. 1928)
“There is assuredly no more effectual method of clearing up one’s own mind on any subject than by talking it over, so to speak, with men of real power and grasp, who have considered it from a totally different point of view.”
—Thomas Henry Huxley (1825–95)
“If all feeling for grace and beauty were not extinguished in the mass of mankind at the actual moment, such a method of locomotion as cycling could never have found acceptance; no man or woman with the slightest aesthetic sense could assume the ludicrous position necessary for it.”
—Ouida [Marie Louise De La Ramée] (1839–1908)