Predicting Critical Nodes in Temporal Networks by Dynamic Graph Convolutional Networks

Round 1

Reviewer 1 Report (New Reviewer)

Please explain why "but cannot predict the importance of nodes in the future." in Section 2.

The last paragraph lists related works without providing clear insight about the position of the paper.

Section 3 directly starts with a subsection without introduction paragraph.

 

 

Line 170 : explain why to add dummy nodes, please. Is it to have same shape matrices?

 

Line 184 : cite (again?) the previous researches.

 

Line 205 was not clear the first time.

 

At line 202 you specified s snapshots then the iterator becomes s for L snapshots.

 

At line 204 you specified a big-o complexity of O(n) which becomes O(N) at line 205. 

 

In the subsection 5.1 it will be better to use a "description" list instead of numbering distinct datasets in the text.

 

Line 226 why to specify only some features of the GPU? Is it important for the rest of the paper? Maybe the benchmark?

If it is important, explain why. Otherwise you can add an annex with hardware specification.

 

I cannot find definition of the mu parameterin equation 8 (≃ line 232).

 

Figure 5: increase the size. Text and labels on figures must be the same as the text.

Use subfigure for each dataset. Explain axis.

 

Line 263: apriori we do not know the real ranking. Maybe I am wrong but clarify please. (also at line 285)

I think it is better to claim that your method provides a better prediction according to the Kendall score;

 

Line 266: I also do not agree. The proposed combination cannot solve the real cases you have tested.

Maybe there are other combinations and/or other real cases.

 

After line 270:  people usually care more about top k nodes. Can you explain or use references, please?

 

You have time tables that you do not use. Please write comments. When s becomes bigger does the time is linear for each dataset?

Why is "Contact" the fastest?

 

When you compare the time of the 5 methods you have to explain.

What can be considered as reasonable time?

For instance: except TDC, DGCN requires more time than other methods on all dataset (except TK on UCI).

Why?

DGCN requires much more time on Email than other methods (except TDC).

But the training on Email was faster than DNC and UCI datasets.

Why?

 

At line 287, you specify that the method can be used on large-scale networks. But what about the ranking time? It requires more time than other methods (except TDC).

At line 291: do not use "what's".

 

 

Use "text~\cite{ref}" instead of "text\cite{ref}"

Write a space after comma (e.g. research field,it is a hot topic) and after a colon (e.g.  "topics in this area:(1)")

It is better to use a comma before "and" in a list (e.g. replace "such as degree centrality, betweenness centrality" and k-shell" by "such as degree centrality, betweenness centrality and, k-shell").

I am not a native speaker but some sentence seems not clear enough for me.

There are still typo (e.g "nodes in the future. proposes Dynamic Influence" the sentence is closed with a dot, then the word "proposes" is written between two sentences).

Use full-words for references (e.g. Figure instead of Fig.)

 

Page 10/13 has two tables with a wrong title. Please use space before parenthesis and write "seconds" instead of "senconds".

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report (New Reviewer)

Cite papers for the examples given in lines 24-29

Very good point at lines 40-44

line 112 - why do you only consider discrete time? you discuss continuous time in the conclusions as further work, yet you discard it here. You need to provide some sort of justification.

line 116 - define V and E properly

line 149 - need to clarify the link between L and δ, maybe re-iterate line 114 or reference it ?

line 153 - clarify that _your_ approach changes the "coarse-grained" approach

line 158 - "in this paper" ? clarify please

good example at lines 166-167

lines 176-179 - nice

Paragraph 4.3 - nice

line 212 - justify why did you choose these specific networks and not others? The table that presents their characteristics serves what point? Why are those values important?

lines 257-259 - good point, maybe needs some highlighting

line 260 - add something like "so far, the optimal value..." or  "the optimal value needs further research" , otherwise it's far less elegant than the rest of the method.

lines 264 - 266 - good point

line 268 - needs a better justification than "in real life"

line 273 - outperforms is clear, would you consider defining a magnitude of outperforming?

Is the code available in a repository?

Some reformulations are required:

lines 17-20

line 31-32

lines 34-36 (needs verb)

line 75 - needs a but or a yet, or some link between the sentences

line 86 - needs the subject

line 108

line 136 - misspell

line 140 - needs space before (2)

line 214 - missing word, space needed

line 244-245  

line 245 - space missing

line 249 - grammar

line 250 - missing connection words

line 255 - space missing

line 268 - punctuation

line 271 - broken from the rest of the paragraph, hard to read, as it's on the next page

line 275 - grammar

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report (New Reviewer)

The paper must be improved and revised before resubmission.

(1).For readers’ point of view, It would be beneficial if Authors include some more text about CNN and LSTM in the introduction section.

(2). The key novelty of proposed work should also be listed with key points in the introduction section.

(3). The related work section need to be improved, as it is written very briefly. Instead of writing 1-2 lines about each reference, more text needed for it.

(4). Although the algorithmic view of proposed method is missing, but proposed method is explained well graphically also.

 (5). Failure cases analysis and general observations of proposed method are completely missing in the study.

Comments for author File: Comments.docx

The paper should be checked for typo mistakes and grammatical mistakes before resubmission.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report (New Reviewer)

The authors have addressed my concerns.

No.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

in the method section, need to clarify the stages of research. and also shown instruments to measure predictions.

Results claiming that DGCN is the best should also be demonstrated with measurable results.

plagiarism too hight, can you cek again.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Several points for review:

(1) In the abstract, authors should add more description in terms of the proposed model’s novel design and interaction among different components in the model.

(2) Instead of some proposed concepts, such as DGCN, and LSTM. Please summarize the paper's contributions carefully.

(3) In the GNN branch in the related work, the authors should give good comparisons and analyses for similar or diverse methods in this task.

(4) How to explain figure 1 effectively by combining the unified symbol system and immediate logic and data flow? I feel like it is an important point. Besides, how to represent the dynamic features of GCN, the novelty of the model is limited.

(5) The problem definition has illustrated in Sec.4.1, why the authors describe the task in figure 3 from the perspective of schematics again.

 

(6) In analyzing the experimental results, the author should give a detailed and reasonable explanation for the performance of the proposed model that is better than the baselines of N2V-LSTM and S2V-LSTM.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Please see attachment for details.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

 Please add more novelty in the model design and formal method description.

Author Response

Please see the attachment.

Author Response File: Author Response.docx