Active learning: iteratively expanding the training setΒΆ