NeuralTools: Testing Results in Training Set Versus Testing Set

PN/GRN nets test very well on the training set due to interpolation and possible overfitting; testing set results show real predictive accuracy.

Jan 22, 2026
Knowledge

Title

Answer

Applies to: NeuralTools, all releases

Why do PN/GRN nets show much better testing results on the training set than on the testing set?

PN/GRN nets interpolate using the training data. Testing on the training set often shows very low error because the net “sees” those points, while new testing cases reveal true predictive accuracy. Overfitting also makes training-set results misleading.

Overfitting During Training

URL Name

testing-results-in-training-set-versus-testing-set