1 paper
Ran Shadmi, Jonathan Laserson, Gil Elbaz
In order to analyze a trained model performance and identify its weak spots, one has to set aside a portion of the data for testing. The test set has to be large enough to detect s…