Online learning model

In machine learning, an online learning model is a learning model that does not have a separate training and testing phase. The pseudo-code for this model is:

do
 retrieve the next example x
 get prediction from algorithm
 if prediction is wrong then increment the number of mistakes
 feed the correct result to the algorithm so that it can learn from the mistake
until there are no more examples

A common way to evaluate an algorithm using the online learning model is to give the total maximum number of mistakes.

See also

• Online machine learning