After continuing a training, accuracy on validation set suddenly jumps up

A model was initially trained and gave and validation accuracy at best of 67%

when the training is continued in a new training session, the validation accuracy suddenly jumps to above 95%


Before the training was started, I reloaded the model & param with Restore Model.

Somehow that jump does not sound right. Could you check if there is a problem.

The project is EFE_Leaf_AutoML, Run5 and Run15

We will investigate and get back on this.

OK thanks. AFAIR, I saw the same behavior on another dataset.