Recognition of Noisy Digital Images Using the Asymmetric Coupling Semiconductor Chaotic Lasers Network

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

A theoretical analysis about asymmetric coupling semiconductor lasers network (ACSCLN) has been performed in this manuscript, where the influence of digital images and noises were verified. I will be able to recommend its publication after major revisions that are subjected to the editor’s decision.  

 

1-      I suggest removing the English mistakes along the manuscript, such as in the section Results and discussions, Page 4 (line 4 “trajectores”, line 7 “differen”, line 10 “dirven”, line 17 “has similar with”, line 24 “These idicates”, …).

2-      Figures 3 and 4. Based on these Figures, the authors have concluded that the noises had influence about the temporal trajectories of the semiconductor laser (SL8), although different digital images had the same answer when they were submitted at the same noise (ԑ). Has the literature ever reported different results? What is the novelty? I suggest including this information along the manuscript.

3-         Figure 5. The correlations of the output of the SL 8 induced by one blurred digital image and a noise must be clarified along the manuscript. There are images hard to understand only seeing through them.       

Comments on the Quality of English Language

I    I suggest removing the English mistakes along the manuscript, such as in the section Results and discussions, Page 4 (line 4 “trajectores”, line 7 “differen”, line 10 “dirven”, line 17 “has similar with”, line 24 “These idicates”, …).

 

Author Response

Please  uploaded file 

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

In this paper the authors modelled a chaotic laser network consisting of eight asymmetrically coupled semiconductor lasers for the recognition of the blurred digital images of 0-9 with a certain noise. It is shown that for a fixed injection strength all blurred digital images can be accurately recognized under low noise and some blurred digital images can be completely recognized under arbitrary noise. By varying the injection strength the recognition accuracies of some of the blurred images can be improved. This scheme can have applications as a photonic accelerator. The paper is good as it is and I do not have any further comments to make. I recommend it for publication.

Author Response

Please sees uploaded files 

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript of D. Zhong et al. is considering a theoretical description of a image recognizer based on chaotic laser networks. The paper investigates the possibility to utilize asymmetric coupling semiconductor chaotic lasers network for the recognition of the noisy images of digitals. In my opinion this idea is rather original and novel and thus it deserves to be published. However, the presentation of the research should be improved. Below are my more specific remarks:

1. The most important thing missing in the paper is valdiation of the method against different sets of input data. As far as I understand from it, a single set of 10 digitals has been used and thus it is not clear how the method will generalize with other sets. Moreover, it is not completely clear if the very same images with noise and without noise were used as the reference and test signals It seems that they were essentially the same (with difference in noise level only). If so, it also doubts the significance of the obtained results because in real application you don't usually have the original image without noise and have to test a noisy image against references of different origin. I believe the authors must clarify this point and maybe change the design of their research.

2. I find the term "blurred" to be misleading because usually blurring is some kind of diffusive transformation, for example, gaussian convolution. Here, however, added noise is random so I would use the term "noisy" instead.

3. Eqs. (5)-(7) seem to be essentialy the same as Eqs. (1)-(3) with primes only added to some quantities. If so, I don't think there is need in rewriting them.

4. On page 5, it is said that ?7,3(?), and ?7,7(?) have comparable correlation peaks due to the similarity of the digital image 7 with 3. First, I don't see from the figure that those peaks are comparable and second I don't feel that 3 and 7 have similar spelling.

5. The surfaces at Figure 5 are very hard to read. I would suggest to use a simple 2D color map as produced for example by Python matplotlib.pyplot.pcolormesh() function or similar.

6. On page 7 it is said that "recognition accuracies can be improved by varying the injection strength." However, the dependence of obtained results on Kinj is rather unclear. What should be done to improve the accuracy? It seems that there is some optimal Kinj and it is different for different noise levels. Could the authors somehow comment on this in the paper?

7. There are lot of misprints in the text. Just as an example, the first paragraph of the Part 3 contains "Rung" instead of "Runge", "trajectores" instead of "trajectories", "differen" instead of "different", "dirven" instead of "driven", "indcued" instead of "induced". The whole manuscript should be checked before further submission. Please, also pay attention that in first line of Page 7 "Kinj=9.4" should be "Kinj=12.1"

8. I believe the paper misses an important citation of the paper by Yu Huang et al. [IEEE Photonics Journal, 13, 8700109 (2021)] which is devoted to a similar problem but based on a different optical system. Some comparison would be hepful.

Comments on the Quality of English Language

English of the manuscipt must be improved. There are lot of grammar mistakes and the paper is somewhat hard to read. Just few examples from Page 1:

1. "can play a role of a like PRC for information processing" should be "can play a role similar to PRC for information processing"

2. "As we all known" should be "as well known"

3. "when the connections between the neurons of the photonic reservoir are decided" should be "whereas the connections between the neurons of the photonic reservoir are decided"

4. "our proposed scheme has obviously different with a PRC." should be "the scheme proposed by us is obviously different from the PRC"

5. "traditional photonic neural network that difficultly change the connections in real time." is not understandable by me

6. "we move a critical step forward" should be "we take a critical step forward"

7. And so on.

Author Response

Please sees uploaded file 

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript has been improved, issues were removed and all questions were perfectly answered. So, I recommend its publication in present form.    

Comments on the Quality of English Language

No issue detected.

Reviewer 3 Report

Comments and Suggestions for Authors

I am satisfied with the author's reply and with the improvements they made in the manuscript. However I should say that the paper is still hard to read as the English must be improved.

Comments on the Quality of English Language

There are lot of hard-to-understand sentences. Just an example, lines 105-107 read:

Most notably, under higher pixels, one sample from arbitrary digital image or its corresponding noisy digital image also can be converted into the matrix with the rows of more than 112 and the 7-column using the above-mentioned method.

What is "higher pixels"? How can an image have a sample (there couls be a sample from a set but not from an image"? "the matrix with the rows of more than 112" should be "the matrix with more than 112 rows". "matrix with ... the 7-column" should be "matrix with ... 7 columns" or "7-column matrix".

And there are a lot of similar problems.

Another problem is repetition of very-hard-to-follow phrases like "digital images 0-9 from different samples of the digital image sets 0-9 in the MNIST dataset" - there are two almost consecutive "digital images", as well as strange construction "images from samples" and "sets from dataset"