Doubt regarding epoch and input size`

#1

If I have 100 images divided in 5 batches of 20 images each, then in 1 epoch(Forward+backward propogation), 20 images are precessed or 100 before updating weights?

#2

Hey @preetishvij. there are total of 5 batches, so in every epoch 5 times forward propagation and backward propagation is done, First of all first 20 images are forward propagated, than backward propagation is done for weight updation. then again 20 images forward propagated, and then backward propagated, and so on. for all batches.

HOpe this resolved your doubt.
Plz mark it as resolved in my doubts section. :blush:

#3

But isn’t 1 time forward+backward = 1 epoch? Or batch a way of dividing my main epoch into smaller epochs?

#4

Hey @preetishvij, it holds when you don’t divide your data in batches. But when you divide your data in batches, than it means we want to apply mini batch gradient descend, and weights are updated as many times as there are batches.

HOpe this resolved your doubt.
Plz mark it as resolved in my doubts section. :blush: