This poor performance is due to the nearest neighbor classification which does not allow semantically dissimilar classes to be predicted confidently. It is meant to be "an evaluation dataset notably meant to be used to evaluate models trained on the HowTo100M dataset" [38]. Table 3: Tenclasses which PS-ZSAR performs worstonintheUCF-101 dataset. A shared multi-attention framework for multi-label zeroshotlearning.

Since the gradient ofthe conventional sign function is almost zero everywhere which cannot be used for back-propagation, several attempts have been proposed to alleviate the optimization difficulty by using approximate gradient.

The model combines aprobabilistic encoder withamulti-task model suchthatitcan generate inexpensive and realistic tasks of the class of problems of interest.

Appendix C 4. Ifyouareusingexistingassets (e.g., code, data, models) orcurating/releasingnewassets... (a) Ifyourworkusesexistingassets, didyoucitethecreators?[Yes]See