Vision GNN: An Image is Worth Graph of Nodes
Published:
The paper proposes to represent the image as a graph structure and introduce a new Vision GNN (ViG) architecture to extract graph-level feature for visual tasks.
Paper Reviews
Here I have posted the reviews of some selective papers that I have read and found interesting. The reviews follow the following format:- Summary
- Major Contributions
- Specific Comments- These include the things that I liked about the paper and limitations if any.
- Detailed Analysis of the solution approach
- Detailed Analysis of the experiments
- Comments on Possible Extensions/Improvements/Next Steps
- Conlusion
Note: The reviews are not to defame any work and just for my own journaling and study. If you find anything offensive you can mail me and I will remove it.
Beyond Real-world Benchmark Datasets: An Empirical Study of Node Classification with GNNs
Published:
The paper discusses the results of conducting extensive experiments with a synthetic graph generator that can generate graphs having controlled characteristics for fine-grained analysis for node classification tasks.
Feature Overcorrelation in Deep Graph Neural Networks: A New Perspective
Published:
The paper is a research paper that proposes a new perspective to look at the performance degradation of deep graph neural networks (GNNs), which is feature overcorrelation.
FedGNN: Federated Graph Neural Network for Privacy-Preserving Recommendation
Published:
The paper proposes a federated framework for privacy-preserving GNN-based recommendation which can train GNNs in a decentralized manner on user data.
DistDGL: Distributed Graph Neural Network Training for Billion-Scale Graphs
Published:
The paper discusses design and development of a distributed graph neural network training framework based on existing Deep Graph Library.