Hyperparameter

Description

• Knobs that ML practitioner has to tweak during successive runs of training a model
• Unlike parameters, hyperparameters are not automatically learned
• Example - learning rate

Common hyperparameters

• Learning rate | Learning rate decay | Epoch | Batch size
• Number of Layers | Activation functions of layers

Difference between parameter and hyperparameter

Area	Hyperparameter	Parameter
Who decides?	Trainer	Training process
Example	Number of layers Learning rate Activation functions	Weight Bias

FAQ

• What is the difference between a parameter and hyperparameter?
  • Parameters are automatically learned during the training process.
  • Hyperparameters have to be decided by the trainer

See also

• Parameter
• Learning rate | Learning rate decay
• Epoch | Batch size
• Layer | Activation function