An Optimal Adaptive Constellation Design Utilizing an Autoencoder-Based Geometric Shaping Model Framework
Round 1
Reviewer 1 Report
Overall, this manuscript is well written, presented and organised. After establishing the model, the authors not only did simulations but also experiments to verify their designed model, which is very good. Nevertheless, here are two suggestions:
1.In practice of underwater, different environments and characteristics may be encountered, so it would be better to verify the applicability and robustness of the model.
2.Compared to Norm-8QAM, what is the time cost of them? Have you compared the efficiency?
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The paper reports about the design of a optimal QAM constellation through an autoencoder. The proposed solution adapts to the actual features of the real-channel. The paper includes simulation and, in particular, experiments based on a real set-up, that add value to the paper. The paper is, in general, well organised and fits the interests of the journal readers. Below my comments:
- Sometimes the paper is repetitive. For example, the 2-step the system is based on are repeated several times. I suggest to re-read the paper to clean up the repetitions.
- Only paper from far-east authors are referenced. I suggest to acknowledge the work of researchers from different parts of the word by adding other papers, to give to this work a more wide audience.
- Figure 1 and 2 are referenced in the text, but not explained in details. I suggest to spend more words on their details and explain the meaning of all of their parts.
- The constellations in Fig.3-4 are a bit to small. I suggest to move them in a separate, larger panel placed, for example, at the side of the present figures.
- A key point in scientific application is that experiments should be reproducible by readers. I encourage the authors to add all the details needed for this purpose, that are now missing. For example which carrier frequency/frequencies have been used to achieve 2Gb/s? How distributed?
- Can authors give more details about the employed LED? Bandwidth? Power?
- In experiments authors employed an equalizer. Which its equalization curve was like?
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
This paper reports an autoencoder-based geometric shaping model (AEGSM) framework to jointly optimize quadrature amplitude modulation (QAM) signals symbol-wise and bit-wise for underwater visible light communication. Unlike traditional GS methods which only give theoretically optimal shaping solutions, our framework can always obtain the globally optimal shaping scheme for a specific channel or different application scenarios. I have the following comments/concerns about the submitted manuscript that could improve its overall quality and readability.
1. The authors need to provide the main contribution and novelty of their proposed technique over the already reported in the literature. There are a lot of techniques reported in the literature based on the autoencoder-based geometric shaping model.
2. Figure. 2 should be explained in more detail. It is better to represent the input in this figure as Sin
3. The authors should mention how was the equation (7) obtained. There is no reference provided for this equation.
4. In Figures 3 and 4, the authors should explain regions I, II, and III in more detail?
5. Why signal-to-noise ratio of Gaussian white noise is kept at 15 dB in Figure 5? The reference should be provided in the text. Also, what does Vpp (100 mV) means?
6. What is the reason for setting up the epoch to 100 in Figure. 7 ?
7. More references to the most recent work published in 2020, 2021, and 2022 should be included.
8. There are a few typos in the manuscript that need to be fixed.
Few typos in the manuscript that need to be fixed.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Authors made correction to the paper, however I'm not completely satisfied by the modifications.
- Figure description: authors added a description of what they did, not of what is reported in figures, For example Figures include a lot of blocks not described / referenced. The description should include a 1-by-1 reference to the figure content. I suggest to add, during textual description, clear references to the corresponding figure sections. For example, "collecting symbol data" panel includes a chain with several blocks. Where are they described/referred?
- I suggest to add a table with all of the parameters of the comunications like, carrier frequency, baandwidth, samplig rates, etc. I whould add as well the symbol time, which is not completely clear to me. With a 407MHz carrier, if a symbol per carrier period is transmitted, the rate is 407*3bit= 1.2Gb/s. To reach 2Gb/s authors tranmitted more symbols per cycle. Is this correct? How many?
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
