1. Why Do We Use Epochs and Batch Size?
Training dataset sometimes may be very large for the computers to deal with, so we have to divide the dataset into smaller size to help the computers update the weights of the neural networks each time. That is why we use epochs and batch size.
2. Epochs
Definition: One Epoch is when an ENTIRE dataset is passed forward and backward through the neural network only ONCE.
The number or Epoch gives no clear conclusions because datasets and the diversity of them are different.
3. Batch Size
Definition: Total number of training examples present in a single batch.
NOTE: Batch Size and Number or batches are two different definitions.
4. Batch
Definition: Batch is the number that you devide your all data into several parts.
5. Example
We can divide the dataset of 2000 examples into batches of 500 then it will take 4 iterations to complete 1 epoch.
Batch Size: 500
Itertaion: 4