From Softmax to Sparsemax: A Sparse Model of Attention and Multi-Label Classification
This provides a sparse activation function for improving efficiency and interpretability in multi-label classification and attention mechanisms, though it is incremental as it builds on softmax.
The authors introduced sparsemax, a sparse alternative to softmax for outputting sparse probabilities, and applied it to multi-label classification and attention-based neural networks, achieving competitive performance with more compact attention focus.
We propose sparsemax, a new activation function similar to the traditional softmax, but able to output sparse probabilities. After deriving its properties, we show how its Jacobian can be efficiently computed, enabling its use in a network trained with backpropagation. Then, we propose a new smooth and convex loss function which is the sparsemax analogue of the logistic loss. We reveal an unexpected connection between this new loss and the Huber classification loss. We obtain promising empirical results in multi-label classification problems and in attention-based neural networks for natural language inference. For the latter, we achieve a similar performance as the traditional softmax, but with a selective, more compact, attention focus.