Pith. sign in

REVIEW 5 cited by

Graph Condensation via Receptive Field Distribution Matching

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2206.13697 v1 pith:SSMKSZPT submitted 2022-06-28 cs.LG cs.AI

Mengyang Liu , Shanchuan Li , Xinshi Chen , Le Song This is my paper
classification cs.LGcs.AI
keywords graphdistributionreceptivematchingfieldfieldsgcdmgnns
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Graph neural networks (GNNs) enable the analysis of graphs using deep learning, with promising results in capturing structured information in graphs. This paper focuses on creating a small graph to represent the original graph, so that GNNs trained on the size-reduced graph can make accurate predictions. We view the original graph as a distribution of receptive fields and aim to synthesize a small graph whose receptive fields share a similar distribution. Thus, we propose Graph Condesation via Receptive Field Distribution Matching (GCDM), which is accomplished by optimizing the synthetic graph through the use of a distribution matching loss quantified by maximum mean discrepancy (MMD). Additionally, we demonstrate that the synthetic graph generated by GCDM is highly generalizable to a variety of models in evaluation phase and that the condensing speed is significantly improved using this framework.

Discussion (0). Sign in to comment.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Semi-Supervised Text-Attributed Graph Distillation

    cs.AI 2026-05 reject novelty 6.0 of 10

    STAD distills large text-attributed graphs into tiny human-readable graphs that match or beat full-graph semi-supervised node classification.

  2. Dynamic Graph Condensation

    cs.LG 2025-06 conditional novelty 6.0 of 10

    DyGC, the first framework for dynamic graph condensation, synthesizes a small temporal graph that preserves enough spatiotemporal structure to train dynamic GNNs with up to 1846 times speedup and around 96 percent fidelity.

  3. Simple yet Effective Graph Distillation via Clustering

    cs.LG 2025-05 conditional novelty 6.0 of 10

    ClustGDD distills large graphs by clustering node embeddings and refining synthetic attributes, achieving state-of-the-art node classification accuracy at orders of magnitude lower time cost.

  4. Dataset Distillation via Vision-Language Category Prototype

    cs.CV 2025-06 conditional novelty 5.0 of 10

    A dataset distillation method that combines K-means image prototypes with LLM-generated text prototypes to synthesize small, high-accuracy training sets.

  5. Heterogeneous Graph Condensation via Role-Aware Clustering

    cs.LG 2026-07 conditional novelty 4.5 of 10

    Role-aware clustering (class-partitioned targets, type-wise non-targets) plus cluster-level reconstruction yields compact heterogeneous graphs that train HGNNs near full-graph accuracy at far lower condensation cost.

Pith tools