Fine-Grained Graph Generation through Latent Mixture Scheduling
TLDR
This paper introduces a CVAE with a latent mixture scheduler for fine-grained, controllable graph generation, outperforming existing methods.
Key contributions
- Introduces a novel CVAE for fine-grained, structure-aware graph generation.
- Refines the decoder's latent space by aligning graph and property representations.
- Employs a mixture scheduler to progressively integrate graph and control priors.
- Achieves high generation quality and controllability on five real-world datasets.
Why it matters
Existing graph generation methods lack precise control over structural properties, limiting their utility in critical applications like drug discovery. This paper's CVAE with a mixture scheduler enables fine-grained control, leading to more accurate and useful generated graphs. This advancement significantly improves the applicability of graph generation across various domains.
Original Abstract
Structure aware graph generation aims to generate graphs that satisfy given topological properties. It has applications in domains such as drug discovery, social network modeling, and knowledge graph construction. Unlike existing methods that only provide coarse control over graph properties, we introduce a novel conditional variational autoencoder for fine-grained structural control in graph generation. The approach refines the decoder's latent space by dynamically aligning graph- and property-driven representations to improve both graph fidelity and control satisfaction. Specifically, the approach implements a mixture scheduler that progressively integrates graph and control priors. Experiments on five real-world datasets show the efficacy of the proposed model compared to recent baselines, achieving high generation quality while maintaining high controllability.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.