Model parameter: Difference between revisions

Latest revision as of 16:31, 7 June 2014

Definition

In machine learning, the term model parameter or parameter is used for initially unknown numerical values that are part of a general functional form used to predict outputs from inputs. For a specific choice of the value of each parameter, we get a specific function that predicts outputs from inputs. The goal of machine learning problems is to determine a good choice of the set of parameter values to be able to predict the outputs well from the inputs.

Related notions

Model hyperparameter is used for a (typically discrete) parameter that selects between different choices of model.

@@ Line 1: / Line 1: @@
 ==Definition==
-In machine learning, the term '''parameter''' is used for initially unknown numerical values that are part of a general functional form used to predict outputs from inputs. For a specific choice of the value of each parameter, we get a specific function that predicts outputs from inputs. The goal of machine learning problems is to determine a good choice of the set of parameter values to be able to predict the outputs well from the inputs.
+In machine learning, the term '''model parameter''' or '''parameter''' is used for initially unknown numerical values that are part of a general functional form used to predict outputs from inputs. For a specific choice of the value of each parameter, we get a specific function that predicts outputs from inputs. The goal of machine learning problems is to determine a good choice of the set of parameter values to be able to predict the outputs well from the inputs.
 ==Related notions==
-* [[Hyperparameter]] is a constant that controls the [[learning algorithm]] itself. The choice of hyperparameter affects how quickly the algorithm finds the parameters, and might also affect the choices of parameter values themselves.
+* [[Model hyperparameter]] is used for a (typically discrete) parameter that selects between different choices of model.