Hyperparameter: Difference between revisions

From Machinelearning
(Created page with "==Definition== The term '''hyperparameter''' is used for something that is not a model parameter but broadly affects the final result of the learning algorithm and it...")
 
No edit summary
 
(One intermediate revision by the same user not shown)
Line 10: Line 10:
| [[model hyperparameter]] || [[model selection]]
| [[model hyperparameter]] || [[model selection]]
|-
|-
| [[regularization hyperparameter]] || regularization
| [[cost function hyperparameter]] || [[cost function selection]]
|-
|-
| [[learning algorithm hyperparameter]] ||
| [[regularization hyperparameter]] || [[regularization]]
|-
| [[learning algorithm hyperparameter]] || [[learning algorithm]]
|}
|}

Latest revision as of 16:42, 7 June 2014

Definition

The term hyperparameter is used for something that is not a model parameter but broadly affects the final result of the learning algorithm and its ability to perform well on new test data.

Types of hyperparameters

Hyperparameter type Stage at which it appears
model hyperparameter model selection
cost function hyperparameter cost function selection
regularization hyperparameter regularization
learning algorithm hyperparameter learning algorithm