b71f5aaf3371c2cdfb7a7c0497f569d4-AuthorFeedback.pdf

However,the proposed approach is not restricted to this particular form of attention. For example, spatially37 global feature-based attentioncan also be studied within the context of neural encoding models as "channel-wise"38 attention-weights instead ofspatial attention masks. Webelievetheobservation thatneural response prediction may39 be a useful supervision goal to study attentional deployment is particularly exciting and can be extended in novel40 ways.