Machine learning terminology

This page lists standard terminology in machine learning and corresponding terminology we might see in other related areas.

Most standard term	Other terms that may be used	Meaning
Example	Data point, instance, case, specific feature values, specific input values	The machine learning problem generally involves predicting an output (label) from input information (i.e., the values of particular features). Each particular choice of input values (feature values) constitutes one example. The example is called labeled if the corresponding observed output is included, and unlabeled otherwise.
Label	Output, observed value, specific output value	For any individual example, the corresponding label is the output as actually observed.
Model	Function, Predictive model	This is the function that we are trying to find, that does a good job of predicting the output from the inputs (features).
Feature	Input coordinate, spanning function, (corresponds to) parameter	For each example, specifying the example involves specifying the values of all features. Each feature corresponds to one coordinate of the input domain. We can distinguish between "elementary features" that need to be explicitly specified, and "derived features" that can be deduced from the elementary features.
Model type	Functional form	This is the general type of function that we are trying to fit, with potentially unknown parameters, that are to be determined using the learning algorithm.