Dimension Reduction Using New Bond Graph Algorithm and Deep Learning Pooling on EEG Signals for BCI
Round 1
Reviewer 1 Report
In this manuscript, the authors propose a new optimization algorithm - Bond Graph Algorithm (BGA). The algorithm is employed for feature extraction and feature selection in a series of experiments for EEG electrode number reduction and frequency interval reduction.
The manuscript is well-structured and very informative.
The proposed optimization approach is clearly presented.
The analysis of experimental results is thorough.
My remarks are as follows:
In the "Introduction" section, please remove the results description part (l. 98-132).
In the “Related work” section, please add an analytical summary of advantages and disadvantages of previous similar studies.
In section “3.2. Bond Graph Optimization and Algorithm”, my suggestion is to include a practical example - a classical optimization task, and model and solve it (step-by-step) by using the new proposed approach.
What are the time and space complexity of your optimization algorithm?
In the "Conclusion" section, future plans are missing.
Technical remarks:
p. 286: C, I, R, SE, SF, TF, and GY are undefined.
l. 342: “gest” -> “best”.
l. 449: “In this article, we will only mention the formulas used for our article. Below we briefly show the formulas used in the formulation. Which include [30,46].” – Please, edit this fragment.
l. 454: “If one formula based on algorithms is active, which means feature selected.” – Please, edit this sentence.
l. 509: “We have a 63% reduction” – Please, explain.
Figure 3, 8, 10, 12: The font size should be enlarged.
l. 626: “Fiat” -> "filter".
l. 637: “VGA” – “BGA”.
Author Response
Thank you for your good comments on improving our paper.
Reviewer#1, Concern # 1: In the "Introduction" section, please remove the results description part (l. 98-132)
Author response and action: Thanks for your comment. We removed these parts.
Reviewer#1, Concern # 2: In the "Related work" section, please add an analytical summary of advantages and disadvantages of previous similar studies.
Author response and action: We added one paragraph in Related work for an analytical summary of the advantages and disadvantages of previous similar studies.
Reviewer#1, Concern # 3: In section "3.2. Bond Graph Optimization and Algorithm", my suggestion is to include a practical example - a classical optimization task, and model and solve it (step-by-step) by using the new proposed approach.
Author response and action: Firstly, we improved and revised the introduction of the bond graph for better understanding. Secondly, we discussed the difference between PSO and BGA algorithms's structures.
Reviewer#1, Concern # 4: What are the time and space complexity of your optimization algorithm?
Author response and action: We explained the time and space complexity of the BGA algorithm and compared the execution time for one benchmark example.
Reviewer#1, Concern # 5: In the "Conclusion" section, future plans are missing.
Author response and action: We added a paragraph related to the future works.
Reviewer#1, Concern # 6: C, I, R, SE, SF, TF, and GY are undefined.
Author response and action: We defined all of these elements in the text.
Reviewer#1, Concern # 7: “gest” -> “best” (I. 342), “Fiat” -> "filter” (I. 626) ", and “VGA” – “BGA” (I. 637)”.
Author response and action: We correctd these mistakes
Reviewer#1, Concern # 8: "In this article, we will only mention the formulas used for our article. Below we briefly show the formulas used in the formulation. Which include [30,46]." – Please, edit this fragment.
Author response and action: we revised this part.
Reviewer#1, Concern # 6: If one formula based on algorithms is active, which means feature selected." – Please, edit this sentence.
Author response and action: we revised this part.
Reviewer#1, Concern # 6: "We have a 63% reduction" – Please, explain.
Author response and action: We elaborated these sentences in the text.
Reviewer#1, Concern # 6: Figure 3, 8, 10, 12: The font size should be enlarged
Author response and action: Figure 3 is redesigned. And the font sizes are corrected.
Author Response File: Author Response.DOC
Reviewer 2 Report
The paper focuses on dimension reduction in the context of studying brain signals.
The topic is interesting and worth investigating. The paper includes an adequate literature review. The results of the approach compare favorably to other approaches in the scientific literature. However, there are several issues that should be addressed.
The explanation of the concept of bond graph at the beginning of section 3.1 (first five-six paragraphs) is difficult to understand and the authors are kindly asked to revise it.
There seems to be a formatting issue at line 256 which ends with "Connection 1:".
Between lines 349 and 265, the numbering of the equations in not clear. The authors are kindly asked to check all the equations in their paper. Issues are also present at other equations, including 11, 12, 13, 14, etc.
In the case of equation 10, the meaning of trace should be better explained.
The quality of figure 3 is very low. Moreover, conventionally, a flow diagram should have the start block at the top and not somewhere in the middle. The authors are kindly asked to revise the figure.
At line 663 the paper mentions that the dataset also includes foot and tongue as classes. The authors are kindly asked to clarify the relevance of these classes for their study, focused on analysing brain information.
Author Response
Thank you for your good comments on improving our paper.
Reviewer#1, Concern # 1: The explanation of the concept of the bond graph at the beginning of section 3.1 (first five-six paragraphs) is difficult to understand and the authors are kindly asked to revise it.
Author response and action: Firstly, we improved and revised the introduction of the bond graph for better understanding. Secondly, we discussed the difference between PSO and BGA algorithms' structures.
Reviewer#1, Concern # 2: There seems to be a formatting issue at line 256 which ends with "Connection 1:".
Author response and action: We correctd the formats.
Reviewer#1, Concern # 3: Between lines 349 and 265, the numbering of the equations in not clear. The authors are kindly asked to check all the equations in their paper. Issues are also present at other equations, including 11, 12, 13, 14, etc.
Author response and action: We added the equation numbers for all of the formulas in the paper.
Reviewer#1, Concern # 4: In the case of equation 10, the meaning of trace should be better explained.
Author response and action: We added a sentence explaining the trace.
Reviewer#1, Concern # 5: The quality of figure 3 is very low. Moreover, conventionally, a flow diagram should have the start block at the top and not somewhere in the middle. The authors are kindly asked to revise the figure.
Author response and action: Figure 3 is redesigned accordingly. This new model addresses your comment.
Reviewer#1, Concern # 6: At line 663 the paper mentions that the dataset also includes foot and tongue as classes. The authors are kindly asked to clarify the relevance of these classes for their study, focused on analysing brain information.
Author response and action: One sentence is added for explaining this. Similar to the other works in the letereture we have used only two classes of the dataset that includes the left and right hands.
Author Response File: Author Response.DOC
Round 2
Reviewer 2 Report
I would like to thank the authors for the changes made.
