Epoch

An epoch in machine learning is a single pass through the entire training dataset. The number of epochs, denoted as $N$, determines how many times the data is applied to the model.

Why Use Multiple Epochs?

• Repetition for Learning: The data is applied to the model $N$ times to improve learning and accuracy. For example, if $N=10$, the model will see the entire dataset 10 times.

Example

Epoch 1/10
6250/6250 [==============================] - 6s 910us/step - loss: 0.1782

• Epoch 1/10: Indicates the model is currently on the first epoch out of a total of 10.
• Batches: For efficiency, the dataset is divided into smaller groups called ‘batches’. In TensorFlow, the default batch size is 32. With 200,000 examples, this results in 6,250 batches.
• Batch Execution: The notation 6250/6250 shows the progress of batch execution within the current epoch.