Machine-Learning-Based Optimization for Multiple-IRS-Aided Communication System
Round 1
Reviewer 1 Report
The authors propose a machine learning algorithm for IRSs wireless system
- The English writting should be revised
- The authors should clarify the novelty and motivation of their work
- A comparison should be included and result discussion should be enhanced
- The discussion of the proposed GRNN should be enhanced
- The contributions should be clarified.
Author Response
We would like to thank the reviewer for his/her valuable comments, which have helped us to further improve the quality of the paper. We have addressed all the comments and suggestions and revised our manuscript as attached.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
å¯¹æ–‡ç« æ•´ä½“è¯„ä»·
This paper substantially investigates machine learning-based optimization for multiple IRS-aided communication systems. A low-complexity optimization algorithm based on the new GRNN is proposed. In addition, the DRL is utilized for the dynamic UEs’ scenario. Specifically, a DDPG-based algorithm is designed to make up for the GRNN algorithm restrictions.
The considered topic is interesting and useful. The theoretical derivation is rigorous, and the results appear correct and believable.
In general, the manuscript is well-written (although I gave a good number of corrections and suggestions) and provides an interesting conclusion.
Nevertheless, the reviewer holds some concerns about this work which you can find below. I suggest that the authors revise and improve the manuscript accordingly.
1. The co-channel interference is mentioned in the introduction part. In addition, channel estimation is emphasized in this part. All these issues are strongly related to the channel model of the constructed system model, which further influences the prediction result. I would like to provide more reviews about the channel model of similar work in the introduction part, which will increase the cohesiveness of this work. The following reference could help the readers:
Outage Probability and Average BER of UAV-assisted Dual-hop FSO Communication with Amplify-and-Forward Relaying
Statistical Sparse Channel Modeling for Measured and Simulated Wireless Temporal Channels
2. More explanation needs to be included regarding imperfect channel state information conditions for better clarity.
3. The abstract should also be revised. The author should highlight the results, not simply list what has been done.
4. Line 1: Intelligent reflecting surfaces --> intelligent reflecting surfaces
5. The symbols used in the system modeling section are not well defined. It is advised to define all the symbols clearly to make the paper easy to understand.
Author Response
We would like to thank the reviewer for his/her valuable comments, which have helped us to further improve the quality of the paper. We have addressed all the comments and suggestions and revised our manuscript as attached.
Author Response File: Author Response.pdf
Reviewer 3 Report
The authors discuss the area of energy efficiency of intelligent reflective surfaces (IRS) as a promising technology for future data networks. In this paper, they focus a single cellular network where multiple IRSs are deployed to assist in downlink transmission from a base station (BS) to multiple user equipment (UE).
The authors focus on minimizing the overall transmit power of the BS by jointly optimizing the transmit beamforming matrix of the BS and the reflection matrices of all IRSs. The alternating approach is widely used to find convergent solutions, but its applicability is limited by its high complexity, which is more complex in dynamic environments. Therefore, machine learning (ML) is used to find the optimal solution with lower complexity. For the static scenario, UEs propose a low complexity optimization algorithm based on a novel generalized neural network (GRNN). For the dynamic scenario, UEs propose to use deep reinforcement learning (DRL).
A deep deterministic gradient-based algorithm (DDPG) has been proposed, which forms a constraint on the GRNN algorithm.
Simulation results confirm that the proposed algorithms can achieve better energy savings and convergence with remarkable reduction in computation time compared to alternating optimization-based approaches.
The text of the paper is written in a systematic and scientifically based manner, is complemented by mathematical apparatus and contains an element of novas. Formally, I have no major comments.
No plagiarism has been found.
Author Response
We would like to thank the reviewer for his/her valuable comment on our manuscript, which has motivated us in future work.
Author Response File: Author Response.pdf
Reviewer 4 Report
The main content of the research presented in the article is the investigation of the joint optimization of the active and passive beamforming vectors for a single cellular network assisted by multiple Intelligent reflecting surfaces, intending to minimize the transmit power of the base station.
The topic is not unique, but it is worthy of researching.
The main proposal is a low-complexity optimization algorithm based on the new generalized neural network for the static user equipments scenario.
The deduced conclusions based on the research methods are that the simulation results confirm that the proposed algorithms can achieve better power-saving performance and convergence with a noteworthy reduction in the computation time compared to the alternating optimization-based aproaches.
The conclusions are tenable. However, the article is not sufficiently clear what progress has been made compared with the current research results.
The abstract is informative. It reflects the body of the paper.
The introduction provides sufficient background information for readers in the immediate field to understand the problem.
The text is well arranged and the logic is clear. There are virtually no grammatical errors in this article. The related concepts are introduced clearly. The readability is sufficient.
The novelty of the proposals and approaches applied in the study is not clear enough in the article.
The derivation of formulas and equations is sufficiently rigorous.
The theoretical analysis is sufficient for the purposes of the article.
All figures and tables are clear enough to summarize the results for presentation to the readers. However, not all figures and tables are well referred to in the text. Some of the figures (Figures 7, 8, 9, 10 and 11) and Table 1 are referred to after their appearance in the text whereas they should be referred to before their appearance in the text.
The reference section is informative. However, not all references are accurate. Authors must review and correct the formatting of references in the References section to make it more homogeneous and in accordance with the journal's rules.
Author Response
We would like to thank the reviewer for his/her valuable comments, which have helped us to further improve the quality of the paper. We have addressed all the comments and suggestions and revised our manuscript as attached.
Author Response File: Author Response.pdf
Reviewer 5 Report
The paper Machine Learning-based Optimization for Multiple IRSs-aided Communication System is focused on Intelligent reflecting surfaces (IRSs) and their application in modern future cellular networks. The main goal inside the article is to minimize the total transmit power of a base station (BS) by jointly optimizing the transmit beamforming matrix of the BS and the reflection matrices of all IRSs. In order to achieve the optimized solution, two algorithms are introduced in this paper. For the static user equipment (UE) scenario, a low-complexity optimization algorithm based on the new generalized neural network (GRNN) is proposed, while the deep reinforcement learning (DRL) is utilized for the dynamic UE situation. The paper contains the definition of both algorithms as well as the simulations and experimentations for several different scenarios. The comparison with other existing methods is also provided.
First, the language level throughout the entire paper is very good, I have noticed only several minor errors and mistakes. The language is technically sound.
The paper is well organized, containing sufficient theoretical background as well as the modeling of network, algorithm description and results. The presentation of results is very good, clear and sufficient. The presented results are interesting.
I recommend to add some additional simulation with more UEs and multiple BSs.
Author Response
We would like to thank the reviewer for his/her valuable comments, which have helped us to further improve the presentation of the paper. We have addressed all the comments and suggestions and revised our manuscript as attached.
Author Response File: Author Response.pdf
