|
|
On-Line Learning
| Name |
On-Line Learning |
| Description |
The learning task here is how to combine advice from several experts.
For instance, several experts predict the weather of the next day by giving a prediction and a probability,
e.g. expert1 says, the probability for rain is 0.8.
The prediction of different experts may be conflicting, e.g. expert2 could say, the probability
for rain is 0.2.
The learner then, at the next day, perceives the true outcome, e.g. it actually rains. Hence, the belief
in expert2 decreases, the one in expert1 increases.
The learner again receives predictions of the experts, makes a combination of these advices, in order to decide
(e.g. whether or not to take an umbrella), compares its prediction with reality and updates the weights for the
experts.
|
| Methods |
WEIGHTED-MAJORITY
|
|
|
