Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper introduces a novel and efficient adaptive feature representation learning method called AMF-DP. Overall, the paper is well motivated, and the problem is easy to understand. However,  multiple concern points need to be addressed:

1. My biggest concern is that the description of how DP noise is applied, with a specific privacy parameter ε, and of the total privacy guarantees accumulated (calculated with a moment accountant or so) from a theoretical perspective, is missing. Since DP is primarily a theoretical privacy framework, the paper would benefit from providing more detail on these important points.
2. The related work section could be supplemented with a technical section that outlines how the proposed method differs from existing methods and why they cannot be directly implemented to solve the problem.
3. It isn't easy to understand the overall workflow of the proposed method at first glance, as Figure 1 isn't very informative. Please consider adding the algorithm pseudocode of the proposed method or improving the overall figure.
4. Please provide more details on the selection of hyperparameters, such as α and β. Is there any theoretical or empirical analysis conducted?  The experimental section also does not specifically address these values or any other specific hyperparameter used for training and DP noise generation when introducing the observed results.
5. While the experiments are conducted to evaluate accuracy and privacy leakage, it would be beneficial to supplement the results with varying privacy parameters, such as the DP parameter. Also, a clear statement of exactly what is sensitive in the experimental results would improve the description.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

The paper is very thoughtful and timely. It proposed an innovative approach,  AMF-DP. However, it suffers from a lack of readability. Some sections are difficult to follow due to their lengthy sentences.  I strongly recommend a thorough English editing if possible. Additionally, I think the authors should consider discussing some potential limitations of AMF-DP in the revision.

Comments on the Quality of English Language

A thorough language edit is needed. 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

In the section of the results and its discussions, the language could be simplified or flowcharts or didactic diagrams could be added that can facilitate the understanding of the modules (ALEL, ADS, etc.).

It is worth further justifying the selection and sensitivity of hyperparameters (α, β, λ₁, λ₂) to discuss how different configurations affect the trade-off between privacy, fairness, and performance.

It is worth commenting on how the model could be applied or tested in real or simulated scenarios. What are the limitations?

In the conclusions of the article, a few words could be dedicated to help understand the limitations of the model (cost, dependency, etc.)

What are the practical and concrete contributions of the AMF-DP?

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 4 Report

Comments and Suggestions for Authors

Please find attached

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Thank you, the comments previously mentioned were addressed well. The revised manuscript is significantly improved. 

However, there are minor corrections that should be done, such as re-checking the grammar, and improving the typo. Specifically, I would note typos on Fig.1 ("Classifier"), and Fig.2 ("Softmax") which catches the eye due to being underlined in red. Please ensure correct display of the figures in the final version.

Author Response

Comments 1:
However, there are minor corrections that should be done, such as re-checking the grammar, and improving the typo. Specifically, I would note typos on Fig.1 ("Classifier"), and Fig.2 ("Softmax") which catches the eye due to being underlined in red. Please ensure correct display of the figures in the final version.

Response 1:
Thank you very much for this helpful comment. We have carefully re-checked the whole manuscript for grammar and typos and made several minor corrections, which are marked in red in the revised version.

Specifically, we have corrected the typo in Fig.1 (“Classifier”) and regenerated Fig.2 to remove the red underline under “Softmax” that came from the spell-check marks in the original drawing tool. We have also double-checked all other figures and captions to ensure that the words are correctly spelled and the figures are clearly displayed.

These changes can be found in the revised manuscript on page 5, Fig.1 and page 9, Fig.2.

Reviewer 4 Report

Comments and Suggestions for Authors

The current version comprises all suggestions and improvements given by me. My only current concern is still the size of figures, which is tiny (particularly the labels there included), which could be very difficult to see for several audiences. I suggest increse the size, for instance using the wider format with adjustwidth command in LaTeX, or inclusively changing the current arrangement, letting increase the size of some labels. The authors should be sure that labels will be readable for the most potential readers, otherwise they will become futile.

Author Response

Comments 1:
The current version comprises all suggestions and improvements given by me. My only current concern is still the size of figures, which is tiny (particularly the labels there included), which could be very difficult to see for several audiences. I suggest increase the size, for instance using the wider format with adjustwidth command in LaTeX, or inclusively changing the current arrangement, letting increase the size of some labels. The authors should be sure that labels will be readable for the most potential readers, otherwise they will become futile.

Response 1:
Thank you very much for this constructive suggestion. We completely agree with the reviewer that the labels in some figures should be enlarged to improve readability. In the revised manuscript, we have increased the font size of the labels and legends in Fig. 4 (page 15) and Fig. 8 (page 18), and slightly adjusted their layout to make them clearer for readers.

We have also double-checked the other figures to ensure that all labels are now easily readable in the final version.