Double-Criteria Active Learning for Multiclass Brain-Computer Interfaces

Algorithm 1

The double-criteria active learning with the ELM algorithm.

Inputs: with labeled samples, with unlabeled samples , the tradeoff parameter (), the number of   samples selected on basis of their uncertainty (), the batch size (), and the terminating condition.
Output: The final learned ELM classifier.
(1)	Train the ELM classifier using labeled set .
(2)	Repeat
(3)	Calculate the estimated probability for the samples in with the pretrained ELM classifier according to equation (5) or (6).
(4)	Calculate the uncertainty level of each sample in using equation (7).
(5)	Include the most uncertain samples into the set .
(6)	Select samples from using equation (11).
(7)	Label the selected samples.
(8)	Update the labeled set and unlabeled set .
(9)	Use the extended set to train a new ELM classifier.
(10)	Until the terminating condition is satisfied.
(11)	Return the output the final learned ELM classifier.