LSH Models in Federated Recommendation

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

1. In the introductory part of the work, detailed information regarding the research objectives, the assumptions from which they start and are to be verified and validated, the actions taken by the authors to fulfill the objectives are not explicitly stated.

2. In the current form of the work, the Related works section is missing, which makes it impossible to determine the current state of research in the field addressed by the authors and where exactly their work is placed.

3. I recommend the authors provide more details about what they had in mind when they stated that the MF model will reduce the recommendation effect in exchange for data privacy in federated learning and why it is difficult to ensure accurate recommendation, protecting at the same time data privacy.

4. In my opinion, it is not clearly specified in the paper, what are the results obtained by the authors that give them the right scientific context to state that: on the one hand, the proposed model has the effect of improving the federalized learning process, improving efficiency and accuracy of recommendations and, on the other hand, improved data privacy and protected customer data. I recommend the authors to present, in detail, the comparative experiments performed to validate those federated learning recommendation models proposed in the paper and which also allow the evaluation of the recommendation effect and the protection of users' privacy.

5. The list of bibliographic references is irrelevant and out of date. I recommend restoring it.

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files

Comments 1. In the introductory part of the work, detailed information regarding the research objectives, the assumptions from which they start and are to be verified and validated, the actions taken by the authors to fulfill the objectives are not explicitly stated.

Response 1:We have now modified in the introduction section to illustrate detailed information about the study objectives, the assumptions that they begin and will be verified and verified, the actions taken by the authors to achieve the objectives.

2. In the current form of the work, the Related works section is missing, which makes it impossible to determine the current state of research in the field addressed by the authors and where exactly their work is placed.

Response 2:At the same time, we have improved the relevant works of the article, which helps readers to determine the research status of the author's field and the exact location of their work, so that they can better understand the author's ideas.

3. I recommend the authors provide more details about what they had in mind when they stated that the MF model will reduce the recommendation effect in exchange for data privacy in federated learning and why it is difficult to ensure accurate recommendation, protecting at the same time data privacy.

Response 3:After our modification again, we illustrate our idea in federated learning when pointing out that the MF model will reduce the recommendation effect in exchange for data privacy, and why it is difficult to ensure accurate recommendation while protecting data privacy, that is, the MF model is a more detailed introduction, supplementing the missing parts.

4. In my opinion, it is not clearly specified in the paper, what are the results obtained by the authors that give them the right scientific context to state that: on the one hand, the proposed model has the effect of improving the federalized learning process, improving efficiency and accuracy of recommendations and, on the other hand, improved data privacy and protected customer data. I recommend the authors to present, in detail, the comparative experiments performed to validate those federated learning recommendation models proposed in the paper and which also allow the evaluation of the recommendation effect and the protection of users' privacy.

Response 4:We have made a summary in the second half of the article, including a detailed summary of the experiment, the results of the experiment and the evaluation of the recommendation effect through the experimental results and the protection of user privacy.

5. The list of bibliographic references is irrelevant and out of date. I recommend restoring it.

Response 5:We have revised the bibliography list to address this problem

Reviewer 2 Report

Comments and Suggestions for Authors

This manuscript deals with the use of neural collaborative filtering techniques in federated learning systems.

1) Unfortunately, the manuscript is very difficult to read and to understand. The authors fail in explaining clearly the problem, the background on the main techniques, and the model which is used in the paper. Ultimately, it is not possible to follow what problem is being addressed, hot the NCF techniques have been used or what novelty the paper shows.

Comments on the Quality of English Language

The English language used in the paper is poor, There are frequent errors in syntax, grammar, and meaning, up to the point that often the sentences have low meaning or are almost unreadable. A strong suggestion that the text is revised thoroughly is advised.

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files

Comments 1:Unfortunately, the manuscript is very difficult to read and to understand. The authors fail in explaining clearly the problem, the background on the main techniques, and the model which is used in the paper. Ultimately, it is not possible to follow what problem is being addressed, hot the NCF techniques have been used or what novelty the paper shows.

Response 1:In order to solve this problem, we have made a new arrangement of the content of the article, and solved the grammar problems involved, so that the article is easy to read and understand. And supplement the explanation problem, the background of the main techniques, the model used in the paper, and which NCF techniques are used. Comments on the Quality of English Language

The English language used in the paper is poor, There are frequent errors in syntax, grammar, and meaning, up to the point that often the sentences have low meaning or are almost unreadable. A strong suggestion that the text is revised thoroughly is advised.

Response:we made a new arrangement of the content of the article and solved the grammatical problem

Reviewer 3 Report

Comments and Suggestions for Authors

In this manuscript, the authors integrate Neural Collaborative Filtering (NCF) with federated learning and propose three types of federated learning improvements – Generalized Matrix Factorization, Multilayer Perceptron and Neural Matrix Factorization for better predictions. Additionally, they utilize Laplacian noise in differential privacy to safeguard the uploaded parameters during federated learning training. The topic is very interesting and worth investigating.

 

My remarks are as follows:

The English language quality in the manuscript hinders comprehension and replication of the proposed improvements. A native English editor should edit the entire text.

The manuscript is not well organised, for example, “Related Work” section is missing and “Introduction” section should conclude with a brief overview of the manuscript’s structure.

The proposed methodology requires clearer presentation, potentially with pseudocode or flowchart to explain the authors’ ideas. Figure 2 needs editing – model M’ and model M’’ should be included. Details regarding the development environment and programming languages utilized should be mentioned. Link to the programming code could be added.

In the “Results” section, a comparison with results obtained in previous similar studies should be added.

In “5. Conclusions” section, future research directions are missing.

 

Some technical remarks:

Abstract: “Due to the communication cost problem of federal learning and the privacy protection problem by neural collaborative filtering recommendation. Model Weight Setting Based on Partici…” – This fragment should be edited.

p. 2: Section “Related Work” contains random text and requires rewriting.

p. 2: “GAN” is undefined.

p. 4: “3. NCF FEDERAL RECOMMENDATION BASED ON DP esults” -> “3. NCF Federal Recommendation Based on DP”

p. 9, teble 3:  “This is a table. Tables should be placed in the main text near to the first time” – Edit.

p. 9: “4.4. Pre-training on NeuMF model ample Number of NCF Federal Recommendation” – Edit.

p. 10, Table 10 caption: “Ncf” -> “NCF”.

p. 12, l. 401 – 415: Epsilon is missing.

The references should be edited according to the Journal’s instructions for authors.

Comments on the Quality of English Language

The English language quality is poor.

Author Response

 

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files

Comments:The English language quality in the manuscript hinders comprehension and replication of the proposed improvements. A native English editor should edit the entire text.

The manuscript is not well organised, for example, “Related Work” section is missing and “Introduction” section should conclude with a brief overview of the manuscript’s structure.

The proposed methodology requires clearer presentation, potentially with pseudocode or flowchart to explain the authors’ ideas. Figure 2 needs editing – model M’ and model M’’ should be included. Details regarding the development environment and programming languages utilized should be mentioned. Link to the programming code could be added.

In the “Results” section, a comparison with results obtained in previous similar studies should be added.

In “5. Conclusions” section, future research directions are missing.

Response:

Language Quality: A native English editor has been engaged to improve the overall clarity and coherence of the text, ensuring that it is comprehensible and replicable for readers.

Organization: The manuscript has been restructured to include a "Related Work" section, and the "Introduction" now concludes with a brief overview of the manuscript's structure, providing readers with a clearer roadmap of the content.

Methodology Presentation: The proposed methodology has been presented with clearer illustrations, including pseudocode or flowcharts to elucidate the authors' ideas. Figure 2 has been edited to include model M' and model M''. Additionally, details regarding the development environment and programming languages used have been incorporated, along with a link to the programming code for accessibility.

Results Comparison: The "Results" section now includes a comparison with results obtained in previous similar studies, providing readers with valuable insights into the performance of the proposed approach relative to existing methods.

Conclusions and Future Directions: In the "Conclusions" section, future research directions have been added to guide further exploration and development in the field.

In this manuscript, the authors integrate Neural Collaborative Filtering (NCF) with federated learning and propose three types of federated learning improvements – Generalized Matrix Factorization, Multilayer Perceptron and Neural Matrix Factorization for better predictions. Additionally, they utilize Laplacian noise in differential privacy to safeguard the uploaded parameters during federated learning training. The topic is very interesting and worth investigating.

 

My remarks are as follows:

The English language quality in the manuscript hinders comprehension and replication of the proposed improvements. A native English editor should edit the entire text.

The manuscript is not well organised, for example, “Related Work” section is missing and “Introduction” section should conclude with a brief overview of the manuscript’s structure.

The proposed methodology requires clearer presentation, potentially with pseudocode or flowchart to explain the authors’ ideas. Figure 2 needs editing – model M’ and model M’’ should be included. Details regarding the development environment and programming languages utilized should be mentioned. Link to the programming code could be added.

In the “Results” section, a comparison with results obtained in previous similar studies should be added.

In “5. Conclusions” section, future research directions are missing.

 

Some technical remarks:

Abstract: “Due to the communication cost problem of federal learning and the privacy protection problem by neural collaborative filtering recommendation. Model Weight Setting Based on Partici…” – This fragment should be edited.

1. 2: Section “Related Work” contains random text and requires rewriting.
2. 2: “GAN” is undefined.
3. 4: “3. NCF FEDERAL RECOMMENDATION BASED ON DP esults” -> “3. NCF Federal Recommendation Based on DP”
4. 9, teble 3:“This is a table. Tables should be placed in the main text near to the first time” – Edit.
5. 9: “4.4. Pre-training on NeuMF model ample Number of NCF Federal Recommendation” – Edit.
6. 10, Table 10 caption: “Ncf” -> “NCF”.
7. 12, l. 401 – 415: Epsilon is missing.

The references should be edited according to the Journal’s instructions for authors.

Response：

The mentioned issues have been addressed as follows:

Abstract: The fragment has been edited for clarity and coherence.

p. 2: Section "Related Work" has been rewritten to provide relevant and coherent content.

p. 2: The term "GAN" has been defined upon first mention.

p. 4: Section "3. NCF FEDERAL RECOMMENDATION BASED ON DP esults" has been corrected to "3. NCF Federal Recommendation Based on DP" for clarity.

p. 9, Table 3: The table has been appropriately placed in the main text near its first mention.

p. 9: Section "4.4. Pre-training on NeuMF model ample Number of NCF Federal Recommendation" has been edited for clarity.

p. 10, Table 10 caption: "Ncf" has been corrected to "NCF".

p. 12, l. 401 – 415: The epsilon has been appropriately included.

References: The references have been edited according to the instructions for authors provided by the journal.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I think that the bibliography could still be worked on in order to update it, but the paper can be published in its current form.

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files

Comments 1:I think that the bibliography could still be worked on in order to update it, but the paper can be published in its current form. Response 1:Thank you for your suggestion. We will consider updating the bibliography to ensure it reflects the latest research findings and resources. And  we plan to submit it to a journal in the near future. Thank you once again for your feedback!

Reviewer 2 Report

Comments and Suggestions for Authors

I appreciate the improvements and changes introduced by the authors in this manuscript, which combines neural collaborative filtering with federated learning. The changes have made the work more comprehensible and sound, and also allowed to better understand the experimental results. There are still some issues in this version that need further refinement:

1) Carefully review the English usage, e.g., federal->federated, which appears across all the manuscript.

2) Put appropriate legends and labels in all the Figures (e.g., Figure 4, all the subfigures are void of a legend)

3) Place figures and tables at the top of the page.

4) Highlight the contributions of the paper in the Introduction, and renumber this and the following Sections (i.e., Introduction is section 0)

Comments on the Quality of English Language

The English usage has improved substantially from the original version to the present one, but some corrections are still necessary. Most striking among these is the repeated use of 'federal' instead of 'federated' across the text. Additionally, there are some inconsistencies in the mathematical typesetting and notations. The authors are advised to review these too.

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files

Comments 1:Carefully review the English usage, e.g., federal->federated, which appears across all the manuscript.

Response 1:Thank you for bringing this to our attention. We will carefully review the English usage throughout the manuscript, including replacing "federal" with "federated" where appropriate. Your feedback is greatly appreciated as we strive to ensure clarity and accuracy in our writing.

Comments 2:Put appropriate legends and labels in all the Figures (e.g., Figure 4, all the subfigures are void of a legend)

Response 2:We will ensure that appropriate legends and labels are added to all figures, including providing legends for each subfigure in Figure 4. Your feedback is valuable, and we appreciate your attention to detail as we work to enhance the clarity and comprehensibility of our manuscript.

Comments 3:Place figures and tables at the top of the page.

Response 3:We have already ensured that figures and tables are placed at the top of their respective pages in the manuscript. Thank you for your feedback, as our goal is to improve the organization and readability of the document.

Comments 4: Highlight the contributions of the paper in the Introduction, and renumber this and the following Sections (i.e., Introduction is section 0)

Response 4:Thank you for your suggestion. We will highlight the contributions of the paper more prominently in the Introduction section and renumber the sections accordingly, starting with the Introduction as section 0. Your feedback is valuable as we strive to enhance the clarity and structure of our manuscript.

 

Comments on the Quality of English Language:The English usage has improved substantially from the original version to the present one, but some corrections are still necessary. Most striking among these is the repeated use of 'federal' instead of 'federated' across the text. Additionally, there are some inconsistencies in the mathematical typesetting and notations. The authors are advised to review these too. Response: We appreciate your acknowledgment of the improvements made since the original version. We will address the repeated use of 'federal' instead of 'federated' throughout the text, as well as any inconsistencies in mathematical typesetting and notations. Your comments are valuable to us, and we will ensure to review and make necessary corrections to enhance the clarity and accuracy of our writing.

 

 

Reviewer 3 Report

Comments and Suggestions for Authors

Despite some improvements in the manuscript, it is still not ready for publication.

My remarks are as follows:

The entire manuscript requires final editing.

In “4. Conclusions” section, future research directions are missing.

 

Some technical remarks:

Figure 1: The upper part should be redrawn – the name of upper level layer and other labels can not be identified.

Eq. (2): ‘y bar’ is undefined. It should be changed.

Eq. (4): The superscript text is blurred. Change the font style.

l. 229: ‘em-bedding’ – Edit.

l. 294: hab formula is incorrect.

…

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files

Comments 1:The entire manuscript requires final editing.

Response 1:We will proceed with a final editing pass of the entire manuscript to address any remaining errors or inconsistencies. Your input is appreciated as we work towards ensuring the quality and polish of our document.

Comments 2:In “4. Conclusions” section, future research directions are missing.

Response 2：We will make sure to include future research directions in the "Conclusions" section of the manuscript. Your feedback is valuable as we strive to provide comprehensive insights and guide potential avenues for further study.

Some technical remarks:

Figure 1: The upper part should be redrawn – the name of upper level layer and other labels can not be identified.

Eq. (2): ‘y bar’ is undefined. It should be changed.

Eq. (4): The superscript text is blurred. Change the font style.

1. 229: ‘em-bedding’ – Edit.
2. 294: hab formula is incorrect.

Response:

Thank you for your technical remarks. We will address the following issues:

Figure 1: We will redraw the upper part of the figure to ensure that the labels, including the name of the upper level layer, are clearly identifiable.

Eq. (2): We will change 'y bar' to a defined and appropriate symbol.

Eq. (4): We will modify the font style of the superscript text to improve clarity.

Line 229: We will edit the term 'em-bedding' accordingly.

Line 294: We will correct the formula for the 'hab' calculation.

Your detailed feedback is greatly appreciated, and we will make the necessary revisions to improve the accuracy and readability of our manuscript.