The question has been raised keeping in mind how training of data points is implementing during epochs or sub-sampling with substitution.
Question is as below
When data is trained during first epoch, does the trained data goes to the second epoch.
What I think about the answer
According to me, during each epoch there is change of accuracy. There cane two possible reasons for this
Reason 1: When a epoch is complete, data is shuffled and hence different accuracy is achieved.
Reason 2: Second may be due to trained data which is coming from previous that leads to change in accuracy in the current approach.
Can you please throw some light to this question.