Winnow (algorithm): Difference between revisions

Revision as of 11:33, 10 February 2017

Revision as of 01:21, 19 February 2017

Line 14:

* If an example is correctly classified, do nothing.

* If an example is predicted incorrectly and the correct result was 0, for each feature <math>x_{i}=1</math>, the corresponding weight <math>w_{i}</math> is setmultiplied toby 0<math>\alpha</math> (demotionpromotion step).

<math>\forall x_{i} = 1, w_{i} = 0\alpha w_{i}</math>

* If an example is predicted incorrectly and the correct result was 1, for each feature <math>x_{i}=1</math>, the corresponding weight <math>w_{i}</math> is multipliedset by <math>\alpha</math>to 0 (promotiondemotion step).

<math>\forall x_{i} = 1, w_{i} = \alpha w_{i}0</math>

A typical value for