liveBook · Manning

Epoch

As explained in Timeless Algorithms, The Foundational Ideas

An epoch is one complete pass through a training dataset during model training.

An epoch is one complete pass through the training data. Because training processes data in mini-batches, an epoch contains multiple parameter updates; training usually continues across many epochs. section 11.2.5

θnew=θold-η∂L∂θ

Each mini-batch update adjusts parameters using the loss gradient; many such updates can occur within one epoch. section 11.2.5

Batch size of one

When training processes one observation per update, processing every observation once completes one epoch. section 11.3.5

How Timeless Algorithms, The Foundational Ideas covers it

How deep it goes
Defines an epoch and relates it to mini-batches, repeated parameter updates, and gradual learning.
Examples
Uses a batch-size-one update to illustrate when processing the training data completes an epoch.
What you can do after
Understand how an epoch counts data passes while multiple mini-batch updates accumulate across training.
How it connects
Connects epochs to mini-batches, stochastic gradient descent, backpropagation, gradient descent, and loss.

Also explained in

Compare how 14 books explain epoch →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Timeless Algorithms, The Foundational Ideas. The section links point to the parts of the book it draws on.