Equivariant Denoisers Cannot Copy Graphs: Align Your Graph Diffusion Models
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866918048810663936 |
|---|---|
| author | Laabid, Najwa Rissanen, Severi Heinonen, Markus Solin, Arno Garg, Vikas |
| author_facet | Laabid, Najwa Rissanen, Severi Heinonen, Markus Solin, Arno Garg, Vikas |
| contents | Graph diffusion models, dominant in graph generative modeling, remain underexplored for graph-to-graph translation tasks like chemical reaction prediction. We demonstrate that standard permutation equivariant denoisers face fundamental limitations in these tasks due to their inability to break symmetries in noisy inputs. To address this, we propose aligning input and target graphs to break input symmetries while preserving permutation equivariance in non-matching graph portions. Using retrosynthesis (i.e., the task of predicting precursors for synthesis of a given target molecule) as our application domain, we show how alignment dramatically improves discrete diffusion model performance from 5% to a SOTA-matching 54.7% top-1 accuracy. Code is available at https://github.com/Aalto-QuML/DiffAlign. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2405_17656 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Equivariant Denoisers Cannot Copy Graphs: Align Your Graph Diffusion Models Laabid, Najwa Rissanen, Severi Heinonen, Markus Solin, Arno Garg, Vikas Machine Learning Quantitative Methods Graph diffusion models, dominant in graph generative modeling, remain underexplored for graph-to-graph translation tasks like chemical reaction prediction. We demonstrate that standard permutation equivariant denoisers face fundamental limitations in these tasks due to their inability to break symmetries in noisy inputs. To address this, we propose aligning input and target graphs to break input symmetries while preserving permutation equivariance in non-matching graph portions. Using retrosynthesis (i.e., the task of predicting precursors for synthesis of a given target molecule) as our application domain, we show how alignment dramatically improves discrete diffusion model performance from 5% to a SOTA-matching 54.7% top-1 accuracy. Code is available at https://github.com/Aalto-QuML/DiffAlign. |
| title | Equivariant Denoisers Cannot Copy Graphs: Align Your Graph Diffusion Models |
| topic | Machine Learning Quantitative Methods |
| url | https://arxiv.org/abs/2405.17656 |