A Study on Quantum Graph Neural Networks Applied to Molecular Physics
- Piperno, Simone et al
- arXiv:2408.03427
| Graphical representation of a single dataset entry. The latter describes the arrangement of atoms within an $H_2O$ molecule, showcasing the coordinates and forces acting on each atom. |
| The QGNN schema. |
| Encoding layer. As this figure visualizes the first layer of the architecture, the qubits are initialized in $\ket{0}$. |
| Edge layer of the $n$-th complete layer. |
| Embedding layer of the $n$-th complete layer. |
| Unrolled representation of a QGNN circuit with 2 layers. |
| Pooling for energy prediction. |
| Train and validation losses for our 2-layer QGNN architecture. |
| Components of the train loss for our 2-layer QGNN architecture. |
| Expressibility of the QGNN by the number of layers. |