Multi-task learning is a process used to learn domain-specific bias. It consists in simultaneously training models on different tasks derived from the same domain and forcing them to exchange domain information. This transfer of knowledge is performed by imposing constraints on the parameters...