Machine-Learning-Based Framework for Coding Digital Receiving Array with Few RF Channels
Round 1
Reviewer 1 Report
see the attached file
Comments for author File: Comments.pdf
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
This paper proposes a novel framework for low-cost coding digital receiving array based on machine learning and implements on the FPGA. The results shows good performance. The content of the paper is very sufficient, and the suggestions and opinions are as follows:
1. The description of Table 1 is not clear, the enumerated architectures are not correspond to the characteristics;
2. The paper describes the existing coding network. Is this necessary?
3. The precoding/beamforming based on machine learning is very common. The earliest work, such as 'Beamforming design for large-scale antenna arrays using deep learning[J]. IEEE Wireless Communications Letters, 2019, 9(1): 103-107' and 'Fast beamforming design via deep learning[J]. IEEE Transactions on Vehicular Technology, 2019, 69(1): 1065-1069.' , please introduce the difference of machine learning framework between the above literatures and this paper.
4. In communication scene simulation, channels usually use cluster channel models. Therefore, each baseband equivalent channel in the actual scene usually contains multi-path. How will the multi-path effect affect the performance?
5. The contribution and novelty of the paper need to be re-described.
6. Why is the maximum SNR of sidelobe of the 4-in-1 higher than that of 6-in-1? To my knowledge, the more RFChains, the lower the side lobe. Moreover, why the max snr of 48-DRA is lower than 6-in-1 overloapping in FIg. 10?
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Some minor comments as follows:
1. The term "low-cost" is confusing to the reviewer, as this can refer to energy-efficient, time-saving or hardware-light. Please clarify and potentially change the title.
2. The hardware noise/imperfection should be elaborated, also the reason why only the additive RF noise is considered, e.g., how about the multiplicative noise.
3. The reason why using the non-linear decoding methods is not clear.
4. It would be practically important to compare the maximum likelihood estimation to other counterparts, e.g., MAP or max-log likelihood.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
The revised manuscript is OK for me.