Advanced techniques with Qunova HI-VQE Chemistry

Use the following techniques to get the most out of Qunova HI-VQE Chemistry. For every parameter and its default value, see the API reference.

---

Control the subspace size

max_states sets the maximum subspace dimension for the diagonalization, and max_expansion_states sets the maximum number of classically generated configurations added in each iteration. Larger values can improve accuracy but increase the classical diagonalization cost.

---

Choose an ansatz

Set ansatz in hivqe_options to "epa" (excitation-preserving, the default), "hea" (hardware-efficient), or "lucj" (local unitary cluster Jastrow). Use ansatz_entanglement and ansatz_reps to adjust the entanglement scheme and the number of layer repetitions.

---

Set shots and convergence

• shots sets the number of shots per iteration.
• max_iter caps the number of iterations. The algorithm can stop early when it converges.
• convergence_count, convergence_abstol, and reset_convergence_count control when the energy is considered converged.

---

Screen configurations

Use overlap_screening_tolerance to remove configurations predicted to contribute little before diagonalization, and amplitude_screening_tolerance to remove low-amplitude configurations after diagonalization. Set configuration_recovery to False to discard invalid sampled configurations instead of correcting them.

---

Restart from a previous result

Pass the "states" from a previous result as initial_basis_states to restart the algorithm from those configurations.