Home → Troubleshooting → NeuralTools → Different Results from Training Same Data
Applies to: NeuralTools 5.x–7.x
I trained a net and saved the training report. Then I experimented with changing settings and retraining, but I didn't like the results. So I changed the settings back to the original and retrained, but the training got different results even though the data were exactly the same.
By default, the testing subset is selected differently during each training session, so the results will not be the same. The fewer cases you have, the more variation in results you should expect, all else being equal. See Recommended Amount of Data.
You can freeze the selection of the testing subset by using the setting "Select Same Cases as Long as This Number is the Same" in the middle of the Train tab on the Training dialog.
Last edited: 2015-10-15