- Rating
- 1
- Confidence
- 1
- Informativeness
- 2
- Presentation
- 1
- Soundness
- 2
Read review
Paper summary
The manuscript investigated a study for learning constrained graph dynamics by introducing a decomposition into pressure, vorticity, and reaction roles. Therefore, several strategies are developed, such as divergence-free graph memory, and shared-capacity reaction. Various evaluations were conducted on synthetic benchmarks.
Critical strengths
- Attempting a challenge in learning constrained graph dynamics
- Undertaking a wide range of evaluation with good discussions
Critical weaknesses
-
Very difficult to gain the context of the work. In particular, many terms appear suddenly, such KKT in Abstract, GNN (Graph Neural Network???), GRU (Gated Recurrent Unit???); PVR-Compact, PVR-Strong, Compact-Strong, …
-
Unclear motivation. In line 044, several aspects of the graph systems were mentioned, such as nodal conservation, circulation, and capacity. It lacks transition into Line 049 “We propose”.
-
The language presentation makes the thesis read more like an experimental report.
Questions for authors
-
Clearly justify the motivation of the work and articulate the contributions. The current list of contributions read more like a list of system functions.
-
The related Work reads like this manuscript is attempting many aspects or things all together.
-
Provide clear formulation of the task. For example, it is challenging to understand Equations (1) and (2).