Multi-Dimensional Information Alignment in Different Modalities for Generalized Zero-Shot and Few-Shot Learning

Round 1

Reviewer 1 Report

The intended aim of the paper is interesting, but there are some fatal aspects that must be improved:

1. The method is not explained very clearly. Eq. 1, please describe it in an analytical way instead of descriptions. Then, the following deduced contents will be easy to read.

2. Picture 2 must be explained in detail since it is fundamental to this paper.

3. Please double-check the eq. 4. and 5.

4. The expression must be improved to be readable.

 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

1) What distinguishes continual learning from generalised zero-shot learning?

2) Cite the authors' claim that "high-dimensional data typically contain more discriminative information than low-dimensional data in the same class."

3) How did the author determine that it is preferable to project different dimensional data features to different dimensional distributions in the same latent space?

4) As shown in Table 2. Due to format issues, the results of AWA2 are cut off.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

I don't think my concerns have been clarified in the draft.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 3

Reviewer 1 Report

The draft is not improved to the previous two versions.