Taming the Sigmoid Bottleneck: Provably Argmaxable Sparse Multi-Label Classification

Open in new window