Problems about the proposed approch in the paper.

[LONGRYUU]

I got some puzzles while reading your paper.

1.How to get the attribute vector z? More specifically, how to transfer the image features into vector z? In the paper, z is obtained from a feed-forward layer, what functions in pytorch did you use to combine this layer? Linear functions or convolutional layers? There could be several strategies to compress the 3-dimension image features into a vector.

2.In formulation 8, the subscript 1/n is put out of the brackets, is it a typo error? Does it mean β P(1) √(P(2)) or β √(P(1) P(2))？

[xuewyang]

1. I used average pooling, then a linear function followed by a sigmoid layer. See table 1 of this paper
2. That is not a typo. I used the latter one.

[LONGRYUU]

So the output of sigmoid layer is a probability distribution over the 990 attributions, is it? And z is the feature vector before the sigmoid layer with size of 990?

[xuewyang]

So the output of sigmoid layer is a probability distribution over the 990 attributions, is it? And z is the feature vector before the sigmoid layer with size of 990?

Yes.

[LONGRYUU]

Are you planning to update your paper on arxiv? I'd like to try some ideas but in lack of the correct scores of your approach. Specifically, the results in Table 2 are not updated so I can not make fair comparison with your approach and other baselines you employed.

[xuewyang]

Are you planning to update your paper on arxiv? I'd like to try some ideas but in lack of the correct scores of your approach. Specifically, the results in Table 2 are not updated so I can not make fair comparison with your approach and other baselines you employed.

Yes, please wait 1 or 2 weeks.