Trace Norm Regularised Deep Multi-Task Learning Yongxin Yang and Timothy Hospedales https://arxiv.org/abs/1606.04038