CellS: A Cell-Inspired Efficient Software Framework for AI-Enabled Application on Resources-Constrained Mobile System

Round 1

Reviewer 1 Report

Overall, I believe that the paper has some merit, but there are still changes to be made prior to publication.

First, the introduction does not clearly state the purpose of the work.

The authors might try to see whether the same work has been made in the scientific literature. Please try to add such a discussion.

The experimental evaluation of the performance of the proposed framework CellS is the most important part of the research. The authors provide quantitative evaluation of two experiments.

Why were no statistical analyses performed at all in these comparisons? The current presentation of the results seems very unscientific. The plots are, in many cases, not so clear that trends are obvious without quantification and, given the lack of error bars, it is often completely impossible to interpret the results.

The authors could also provide a qualitative comparison of the proposed framework for the different programming methods such as programming efforts, lines of code, etc.

Please add a discussion related to the future directions of the work.

Author Response

Please see the attachment.

We thank the reviewer for providing valuable comments on improving the paper quality and presentation. We have tried our best to answer all the comments, and have revised the manuscript according to them. We hope that the revised manuscript is now of publishable quality. Responses to reviewer’s comments are given as follows.

1. The introduction does not clearly state the purpose of the work.

Reply: Thanks for the reviewer's comment. Our purpose is to solve the issues raised in the second page of Introduction.

• How to speed up the response time on resources constrained mobile system?
• How to establish an integration framework suitable for artificial intelligence appli-cations?
• How to refactor the software flow with a small cost?

 

2. The authors might try to see whether the same work has been made in the scientific literature. Please try to add such a discussion.

Reply: Thanks for the reviewer's comment. We added references [20-23], and compared and discussed them in Table 2 on page 29 of section 4.5.

 

3. The experimental evaluation of the performance of the proposed framework CellS is the most important part of the research. The authors provide quantitative evaluation of two experiments. Why were no statistical analyses performed at all in these comparisons? The current presentation of the results seems very unscientific. The plots are, in many cases, not so clear that trends are obvious without quantification and, given the lack of error bars, it is often completely impossible to interpret the results.

Reply: Thanks for the reviewer's comment. We have performed statistical analysis of all experimental results and presented the results using boxplot (Figure 13 on pages 21 and 22, Figure 14 on pages 23 and 24, Figure 16 on pages 26 and 27, Figure 17 on page 27 and Figure 18 on page 28).

 

4. The authors could also provide a qualitative comparison of the proposed framework for the different programming methods such as programming efforts, lines of code, etc.

Reply: Thanks for the reviewer's comment. We added Section 4.6 and Table 3 for discussion and comparison of programming methods.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper developed a framework called CellS with inspired by the cell theory. The CellS is a light-weight framework that does not have to depend on other systems; it is easy to integrate with other systems, and it can be used within the resources constrained mobile system. The results seem to be interesting and useful for researchers in the field. I have some comments below. The paper will be acceptable only if all questions are addressed satisfactorily.

1. Are the questions summarized on page 2 the key research questions the authors would like to address in this paper? They should be better summarized.
2. What are the similarities between dataflow programming and multithreading?
3. There are some grammatical errors in the paragraph starting from line 169. Please double check.
4. Figure 1 should be better explained. Perhaps some more caption details should be given.
5. I find Figure 4 difficult to read. There is too much information and the figure is not of good resolution.
6. I tried the code in Figure 9, which does not work. I got an error saying line 12 in override gets warning. Please sort it.
7. How do you evaluate the performance of the CellS framework in different multi-core CPU?
8. Refs 18 and 19 are not relevant.

Author Response

Please see the attachment.

 

We thank the reviewer for providing valuable comments on improving the paper quality and presentation. We have tried our best to answer all the comments, and have revised the manuscript according to them. We hope that the revised manuscript is now of publishable quality. Responses to reviewer’s comments are given as follows.

 

1. Are the questions summarized on page 2 the key research questions the authors would like to address in this paper? They should be better summarized.

Reply: Thanks for the reviewer's comment. Yes, we have added Section 4.6 and Table 3 on pages 29 and 30 for discussion and comparison. On the other hand, we add a summary of this part in section 5 on page 31.

 

2. What are the similarities between dataflow programming and multithreading?

Reply: Thanks for the reviewer's comment. The first section on page 3 explains that both dataflow programming and multithreading are parallel frameworks that can improve software performance.

 

3. There are some grammatical errors in the paragraph starting from line 169. Please double check.

Reply: Thanks for the reviewer's comment. We have corrected the grammatical errors.

 

4. Figure 1 should be better explained. Perhaps some more caption details should be given.

Reply: Thanks for the reviewer's comment. We have added further descriptions of isTrigger, selectPlans, and execute in Figure 1 on page 5.

 

5. I find Figure 4 difficult to read. There is too much information and the figure is not of good resolution.

Reply: Thanks for the reviewer's comment. We have split Figure 4 on page 9 into (a) and (b) and added comments to illustrate this figure.

 

6. I tried the code in Figure 9, which does not work. I got an error saying line 12 in override gets warning. Please sort it.

Reply: Thanks for the reviewer's comment. In Figure 9 on page 15 we add the library that the program needs to import, because CellS requires additional jar files, which may be the reason why it cannot run.

Author Response File: Author Response.pdf

Reviewer 3 Report

This article is very difficult to read.
Much of the text, especially in the abstract, introduction, and conclusion, is very repetitive, which makes the reading difficult.
The authors present and describe some experiments, but the overall relevance is unclear. The connection between the proposed software design method is too vague, which compromises the scientific quality of the work.
Therefore, I do not recommend the publication of this work in its current form.

Author Response

Please see the attachment.

We thank the reviewer for providing valuable comments on improving the paper quality and presentation. We have tried our best to answer all the comments, and have revised the manuscript according to them. We hope that the revised manuscript is now of publishable quality. Responses to reviewer’s comments are given as follows.

1.    Much of the text, especially in the abstract, introduction, and conclusion, is very repetitive, which makes the reading difficult. 
Reply: Thanks for the reviewer's comment. We have rewritten Abstract, Introduction, and Conclusion, and removed similar parts.

2.    The authors present and describe some experiments, but the overall relevance is unclear. The connection between the proposed software design method is too vague, which compromises the scientific quality of the work.
Reply: Thanks for the reviewer's comment. Based on the issues we want to solve on the second page of the Introduction, and through the statistical analysis in Section 4 and the discussion in Section 5, we prove that our framework can indeed solve the issues we encountered.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The revision has greatly improved the article and the author has addressed the concerns raised in the previous review. There are a few minor grammatical errors that another proofreading should be able to capture.

Author Response

We thank the reviewer for providing valuable comments on improving the paper quality and presentation. We have tried our best to answer all the comments, and have revised the manuscript according to them. We hope that the revised manuscript is now of publishable quality. Responses to reviewer’s comments are given as follows.

 

1. There are a few minor grammatical errors.

Reply: Thanks for the reviewer's comment. We have corrected the grammatical errors.

Reviewer 2 Report

My previous questions have been addressed well.

Author Response

We thank the reviewer for providing valuable comments on improving the paper quality and presentation.

Reviewer 3 Report

The authors have made significant changes that have improved the technical quality of this work.

However, there are still some small issues with the English style that should be improved. An example is the sentence in line 14-15 in the abstract that reads: "This paper proposes a(n) intelligent system software framework: CellS to improve smart software development on multi-core mobile processor systems." The use of semicolon here is not correct. It would be better as "This paper proposes an intelligent system software framework, CellS, to improve smart software development on multi-core mobile processor systems."
This kind of issues repeats throughout the text. For example, in line 134 there is a missing semicolon after the word including. The authors can follow their own example on line 79, which was done correctly.

I recommend the publication of this work, with minor language review to improve the overall quality of the presentation.

Author Response

We thank the reviewer for providing valuable comments on improving the paper quality and presentation. We have tried our best to answer all the comments, and have revised the manuscript according to them. We hope that the revised manuscript is now of publishable quality. Responses to reviewer’s comments are given as follows.

 

1. There are a few minor grammatical errors. However, there are still some small issues with the English style that should be improved. An example is the sentence in line 14-15 in the abstract that reads: "This paper proposes a(n) intelligent system software framework: CellS to improve smart software development on multi-core mobile processor systems." The use of semicolon here is not correct. It would be better as "This paper proposes an intelligent system software framework, CellS, to improve smart software development on multi-core mobile processor systems." This kind of issues repeats throughout the text. For example, in line 134 there is a missing semicolon after the word including. The authors can follow their own example on line 79, which was done correctly.

 

Reply: Thanks for the reviewer's comment. We have corrected the grammatical errors.