Your next step
One training step at a time
Predict, measure, adjust — and which numbers are which.
Start learningWhat you’ll learn
Start with an imperfect prediction
Identify the forward pass in training.
About 4 minutes · Open activity
Measure the mismatch
Explain loss as a chosen measure of prediction error.
About 4 minutes · Open activity
Find a direction to improve
Interpret a gradient as a local direction signal.
About 4 minutes · Open activity
Send the error signal backward
Explain backpropagation's role in assigning parameter-change information.
About 4 minutes · Open activity
Take a learning step
Identify the parameter update performed by an optimizer.
About 4 minutes · Open activity
Choose the step size
Explain the learning rate as a training setting.
About 4 minutes · Open activity
A batch is a group of examples
Recognize which examples contribute to one update.
About 4 minutes · Open activity
An epoch is one pass through the examples
Identify an epoch in a finite training dataset.
About 4 minutes · Open activity
A checkpoint saves a model state
Recognize a saved training snapshot.
About 4 minutes · Open activity
Animate a training step
Order prediction, loss, backward gradient calculation and update; identify the learning-rate control and the saved settings.
About 6 minutes · Open activity