GCARe: Mitigating Subgroup Unfairness in Graph Condensation through Adversarial Regularization

Round 1

Reviewer 1 Report

What is being missed most in the paper, is a self-contained description of practical utility of the presented graph compression approach. While learning a condensed graph seems to be a regular, lenghty training process, how its results can be used to accelerate solving of similar problems?

In case of graph-level analysis, I guess a condensed graph is being created in some way first --- but how, in detail?

But in case of nodel-level tasks, e.g. classification, how is your procedure being applied? Continuous referring to [10] and [11] throughout the text does not help much in this regard.

Editorial remarks:

- Fig 1d, both fairness criteria are undefined at this stage of narrative;

- line 185, is "three", should be "four".

Nothing serious.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper reports a superior graph condensation method namely, Graph Condensation with Adversarial Regularisation (GCARe) to regularise the graph condensation process for more fair distillation of a node subgroups. Compared with current graph condensation methods, GCARe achieves superior performance in terms of accuracy and fairness metrics.

The paper is succinctly and clearly written and advances the knowledge in graph condensation in deep learning.

The literature review is well written and the gap this study fills is justified.

I have a small concern on the title – can the title be made more revealing of the content of manuscript than it is?

The authors should also consider strengthening the discussion of their results.

I do recommend this work for publication with minor revisions to include:

-spelling mistakes (line 137 – 138)

-Table 3 – acronyms: SGC, MLP, APPNP, Cheby need to be defined.

Comments for author File: Comments.pdf

The manuscript is well written and easy to understand.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

- The article is well written, and explained in detail.

- Please add a comparative table of your work and state-of-the-art methods that could clearly show the importance of your work. Currently, this is not sufficient to prove contributions. 

- Please improve list of references

No, it look fine to me. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf