Search papers, labs, and topics across Lattice.
This paper reformulates the operator selection process in adaptive variational quantum algorithms as a graph-based decision problem, employing a graph neural network (GNN) to predict the next entangling operator from the interaction graph and state-dependent observables. By training the GNN on data from exact simulations of disordered long-range spin chains, the authors show that their approach significantly reduces the classical cost associated with full-pool gradient evaluations while maintaining energy accuracy comparable to standard ADAPT-VQE. The GNN also demonstrates transferability to molecular benchmarks, effectively generating a shortlist of candidates that allows for near-optimal performance with minimal computational overhead.
A graph neural network can drastically reduce the computational burden of operator selection in variational quantum algorithms while achieving near-optimal performance.
Adaptive variational quantum algorithms like ADAPT-VQE construct tailored ans盲tze by iteratively selecting operators from a pool using gradient-based criteria. While this avoids oversized parameter spaces, repeatedly scanning the full pool incurs a classical cost that scales linearly with pool size-a major bottleneck for systems with long-range interactions or large operator sets. Here, we reformulate adaptive operator selection as a graph-based decision problem and introduce a graph neural network (GNN) policy that predicts the next entangling operator directly from the interaction graph and state-dependent observables. Training data are generated from exact simulations of disordered long-range spin chains, using gradient magnitudes as supervision signals. The learned policy accurately reproduces the dominant structure of the greedy gradient-based selection rule, significantly outperforming heuristics based solely on interaction strength. Integrated into a variational quantum eigensolver (VQE) workflow, this GNN-VQE approach achieves energy errors close to standard ADAPT-VQE while drastically reducing full-pool gradient evaluations. To test transferability beyond spin models, we evaluate the policy on small active-space molecular benchmarks (LiH and BeH_$2$). We find the GNN is highly effective as a shortlist generator: exact rescoring over just a few GNN-proposed candidates recovers near-oracle rollout behavior while searching only a small fraction of the pool. These results demonstrate that adaptive circuit construction contains learnable structure that can be exploited to accelerate variational quantum algorithms.
