ICACT20230157 Question.1
Questioner: yjk1425@163.com 2023-02-20 오후 7:07:52 |
ICACT20230157 Answer.1
Answer by Auhor myy98@mail.sdu.edu.cn 2023-02-20 오후 7:07:52
Chrome  Click!! |
| Has your network considered ablation experiments that consider only spatial attention or channel attention mechanisms? |
Thank you for your question. Some experiments have proved that the performance of CBAM is better than that of spatial attention module and channel attention module. Therefore, this paper does not only consider the spatial attention mechanism and the channel attention mechanism, but also considers the combination of both, and carries out adaptive feature optimization in the two dimensions of channel and space. |
ICACT20230157 Question.2
Questioner: yjk1425@163.com 2023-02-20 오후 7:08:05 |
ICACT20230157 Answer.2
Answer by Auhor myy98@mail.sdu.edu.cn 2023-02-20 오후 7:08:05
Chrome Click!! |
| This paper is a study in the field of RIS. Is RIS very helpful for long distance communication? |
Thank you for your question. RIS is helpful for long-distance communication. Through adjusting the reflecting phase, RIS generates a dynamic wireless channel that adapts to user mobility. Take the cell edge user as an example. When the cell edge user receives weak signals from the base station and there is interference from other base stations, RIS can be introduced to improve the signal power and suppress interference. |
ICACT20230157 Question.3
Questioner: hbw017@mail.sdu.edu.cn 2023-02-20 오후 7:08:18 |
ICACT20230157 Answer.3
Answer by Auhor myy98@mail.sdu.edu.cn 2023-02-20 오후 7:08:18
Chrome Click!! |
| You proposed the EDRSN model. What are its advantages over CNN model, and whether CNN model can also achieve noise reduction？ |
Thank you for your question. EDRSN introduces CBAM module on the basis of ResNet, optimizes features from two dimensions of space and channel, strengthens the recognition and extraction of signal features, and ignores secondary noise interference, so as to achieve better noise reduction effect. CNN can also achieve noise reduction, but its performance may be poor. |
ICACT20230157 Question.4
Questioner: lyh12208@163.com 2023-02-20 오후 7:08:30 |
ICACT20230157 Answer.4
Answer by Auhor myy98@mail.sdu.edu.cn 2023-02-20 오후 7:08:30
Chrome Click!! |
| Your article uses the residual network unit. I want to know what role this structure plays in this article? What is the purpose of your use? |
Thank you for your question. For the residual network, the forward process is linear, and the subsequent input is equal to the result of the input plus each residual element. The main purpose of using residual network is to solve the problem of gradient disappearance by reverse update. |
ICACT20230157 Question.5
Questioner: zhaomingyao1997@163.com 2023-02-22 오후 8:32:19 |
ICACT20230157 Answer.5
Answer by Auhor myy98@mail.sdu.edu.cn 2023-02-22 오후 8:32:19
Chrome Click!! |
| Your article uses the residual network unit. I want to know what role this structure plays in this article? What is the purpose of your use? |
Thank you for your question. For the residual network, the forward process is linear, and the subsequent input is equal to the result of the input plus each residual element. The main purpose of using residual network is to solve the problem of gradient disappearance by reverse update. |
ICACT20230157 Question.6
Questioner: zhaomingyao1997@163.com 2023-02-22 오후 8:37:04 |
ICACT20230157 Answer.6
Answer by Auhor myy98@mail.sdu.edu.cn 2023-02-22 오후 8:37:04
Chrome Click!! |
| Sorry, I entered the wrong question last time. My question is:Is the traditional model similar to yours put into use? |
Different from the traditional model, this paper uses the data-driven deep learning method and introduces the CBAM module on the basis of the deep residual shrinkage network to further improve the accuracy of channel estimation. |
ICACT20230157 Question.7
Questioner: sanjoy.basak@rma.ac.be 2023-02-23 오후 7:20:25 |
ICACT20230157 Answer.7
Answer by Auhor myy98@mail.sdu.edu.cn 2023-02-23 오후 7:20:25
Chrome Click!! |
| Thank you very much for the nice presentation!
I have two questions: (i) What is the impact of channel uncertainty on the model’s performance?
(ii) If you perform longer sensing with memory and use an LSTM model, does it improve the performance?
|
Thank you for your question. First, because the model is trained in a specific channel environment, the accuracy of channel estimation will decline when the channel becomes poor, but it is still better than traditional channel estimation methods. The LSTM method has not been compared, so it is impossible to determine whether its performance will improve. |
ICACT20230157 Question.8
Questioner: yfhou@okayama-u.ac.jp 2023-03-03 오후 9:55:54 |
ICACT20230157 Answer.8
Answer by Auhor myy98@mail.sdu.edu.cn 2023-03-03 오후 9:55:54
Chrome Click!! |
| Good work! Did you consider the special case with moving users? |
Good question! This paper assumes a fixed user location and future work is to consider mobile user scenarios. |
