Tag: uncertainty based loss function weighting for multi task learning