A Light-Weighted Machine Learning Approach to Channel Estimation for New-Radio Systems
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsIn general, this paper is well-written and easy to follow. This topic is interesting and the derivation is correct. Comments:
1. Why the computional complexity of LMMSE method is high?matrix inversion of high-dimensional matrix? It should be clarified since LMMSE is wildly adopted.
2. Are the pilots mutual-orthogonal？ Is the proposed method suitable the Non-orthogonal Random Pilots，such as R1?
[R1] ``Cell-Free IoT with non-orthogonal random pilots: Key research and future directions," IEEE Network , Early Access, DOI: 10.1109/MNET.133. 2200566.
Comments on the Quality of English LanguageNo comments. The presentation is clear and easy to follow.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThe authors have reported a lightweighted ML based channel estimation for NR systems. This topic is both interesting and promising. While I commend the authors for their efforts, I believe that the manuscript could benefit from additional investigations and modifications to enhance its overall quality.
1. In the title, I recommend using 'New Radio' instead of its abbreviation for the general audience to easily search and find related work.
2. In Table 1, there is an unrecognized character for the number of cited references (Machine learning using estimated channels).
3.For the comparison of this work with existing ML-based channel estimations in Table 1, the authors should not only list the content but also explain and summarize their pros and cons. This approach is essential to claim the superiority of this work when compared to previously reported research.
4. There is a typo in Line 172.
5. In Section 6, 'Simulation results,' the authors only investigate the conditions where the length of the CIR L is 1 and 6 separately. However, it appears that the number of L has a significant impact on the proposed machine learning model's performance. Specifically, when L=6, the performance decreases dramatically compared to when L=1. The authors should explain the reason behind this phenomenon. Therefore, the authors should investigate and discuss how the length of the CIR impacts the proposed method's performance in order to provide a more comprehensive analysis.
Given the points raised, it is advisable to conduct a thorough revision of the manuscript before considering it for publication in Electronics.
Comments for author File: Comments.pdf
Minor editing of English language required
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsIn this paper, the authors provided a lightweighted Machine Learning (ML) Approach to channel estimation for NR systems. Specifically, based on the equivalence between Channel Impulse Response (CIR) in the time domain and its corresponding Channel Frequency Response (CFR) in the frequency domain, the lightweighted ML model for the channel estimation is shown to be established in comparison to the existing ML-based channel estimator. Furthermore, for practical use, the quantized weights for the lightweighted ML-based estimator are shown to be feasible without significant performance degradation in the sense of Mean Squared Error (MSE), which shows the effectiveness of the proposed approach in the perspective of memory overhead. However, there are following comments, i.e.,
1、Incorrect title of chapter III. It should be "PELIMINARY ON CHANNEL ESTIMATION"
2、There is only one hidden layer inside the proposed algorithm, is one layer the optimal choice?
3、Spelling error at the beginning of the third paragraph of chapter IV.
4、The activation function of the proposed method is "tanh", Why was this function chosen, are there other activation functions available and what is the difference between them?
5、Chapter VI is incorrectly labeled.
6、The proposed neural network algorithm is well researched, what is the innovation of this paper?
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
Comments and Suggestions for AuthorsBased on the results obtained, the proposal:
“A Lightweighted ML Approach to Channel Estimation for NR Systems” was successful. Furthermore, there are other research fronts that can go in the same direction.
Some suggestions and comments aimed at improving the final version:
A question mark appears in Table 1. Remove and insert the relevant references;
Quote Figures 1, 2, 3, 4, 8, 9 and 10 before their appearance;
Center Fig. 1, Fig. 2, ...;
Doesn't Fig. 2 show KTC = 4 (there would be three subcarriers, KTC = 3, not four)?
Give a brief introduction after each Section: this applies to Section 1, Section 2, Section 3, Section 4, Section 5, Section 6;
Line 97: there is an error in the title of Section 3 (PRELIMINARY);
Uniformly place the title of Section 3 in capital letters;
If the proposed ML is on the order of Q*N, and Q is less than 2*N, does it seem like it is on the order of N^2 as well? It is close to the existing ML;
Why do wk-1 and wk come together (Fig. 6). A quick explanation would be in order;
Make uniform numbering: would it be Section 6.1 and not 6.0.1.?
Quote the acronym PAPR (line 229)?
Table 4: say something, or provide a reference, about “Optimizer Adam”?
Lines 264 and 272 (put the word "scenarios" in the plural);
Reference 14: the year of publication is missing;
References 7, 12, 19, 23, 24, 25 and 26 do not have the year in bold.
Comments on the Quality of English LanguageThe text is clear enough, but there could be improvements in terms of style (does not compromise technical aspects).
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsNo comments
Reviewer 2 Report
Comments and Suggestions for AuthorsThe Authors have addressed my concerns with the original manuscript. The revised manuscript is ready for publication.
