Semi-Equivariant Conditional Normalizing Flows
This work addresses a domain-specific problem in molecular generation, presenting an incremental improvement by incorporating semi-equivariance into existing flow methods.
The paper tackled the problem of learning conditional distributions between 3D graphs using continuous normalizing flows, by deriving a semi-equivariance condition to ensure invariance to rigid motions, and demonstrated its effectiveness in receptor-aware ligand generation.
We study the problem of learning conditional distributions of the form $p(G | \hat G)$, where $G$ and $\hat G$ are two 3D graphs, using continuous normalizing flows. We derive a semi-equivariance condition on the flow which ensures that conditional invariance to rigid motions holds. We demonstrate the effectiveness of the technique in the molecular setting of receptor-aware ligand generation.