Let x 2 RD beadatapoint.

Hierarchical multi-label classification (HMC) is a challenging classification task extending standard multi-label classification problems by imposing a hierarchy constraint on the classes. In this paper, we propose C-HMCNN(h), a novel approach for HMC problems, which, given a network h for the underlying multi-label classification problem, exploits the hierarchy information in order to produce predictions coherent with the constraint and improve performance. We conduct an extensive experimental analysis showing the superior performance of C-HMCNN(h) when compared to state-of-the-art models.

Additional Feedback: I liked the toy examples, which illustrate in which situations the algorithm might work well. I would argue that the setting where a child label corresponds to a region of the space that is fully included by the bigger region of its parent label is most logical. For hierarchical multi-label classification problems, child labels can often semantically be seen as specializations of their parent labels. This can indeed be translated to the feature space, but I would argue that in practice this happens in a slightly different way. Child labels often have the same features active as their parents, e.g.

The submission introduces a new loss function for hierarchical multi-label classification. The justification of the loss function is purely empirical given in a form of results obtained on an illustrative synthetic example. The learning under this loss can be efficiently performed using GPUs. The introduced algorithm obtains the state-of-the-art results. The reviewers agreed that the paper is clearly written, the loss function well-motivated and interesting, and the results worth publishing.

Hierarchical multi-label classification (HMC) is a challenging classification task extending standard multi-label classification problems by imposing a hierarchy constraint on the classes. In this paper, we propose C-HMCNN(h), a novel approach for HMC problems, which, given a network h for the underlying multi-label classification problem, exploits the hierarchy information in order to produce predictions coherent with the constraint and improve performance. We conduct an extensive experimental analysis showing the superior performance of C-HMCNN(h) when compared to state-of-the-art models.