A naukri.com initiative
Arxiv
2d
175
Image Credit: Arxiv
Guided Graph Compression for Quantum Graph Neural Networks
- Graph Neural Networks (GNNs) face challenges with large graphs due to high memory requirements and inefficient sparse matrix operations on GPUs.
- Quantum Computing (QC) is seen as a solution to address GNN challenges and has inspired new algorithmic approaches like Quantum Graph Neural Networks (QGNNs).
- Current quantum hardware limitations restrict the effective encoding of data dimensions in QGNNs, leading to the manual simplification of datasets or the use of artificial graphs.
- The Guided Graph Compression (GGC) framework is introduced to tackle these limitations by employing a graph autoencoder to reduce the number of nodes and the dimensionality of node features.
- GGC compresses graphs to enhance downstream classification tasks, compatible with both quantum and classical classifiers.
- This framework is evaluated on the Jet Tagging task, a crucial classification problem in high energy physics for distinguishing particle jets initiated by quarks and gluons.
- GGC is compared favorably against using the autoencoder as a standalone step and a baseline classical GNN classifier, as proven by numerical results.
- The performance of GGC surpasses the alternatives and enables the testing of novel QGNN approaches on practical datasets.
Read Full Article
10 Likes
For uninterrupted reading, download the app