Speaker
Description
In this talk, we present a new approach to improve the accuracy of ground state approximations in Variational Quantum Eigensolver (VQE) algorithms. We employ subspace representations where orthogonality is enforced via "soft-coded" constraints within the cost function, rather than "hard-coded" at the circuit level.
Similar to other subspace-based methods like Subspace-Search VQE (SSVQE) and Multistate Contracted VQE (MCVQE), our method trains parameters to overlap with the low-energy sector before diagonalizing the Hamiltonian restricted to the subspace. However, by shifting the orthogonality constraints via penalty terms during the minimization, we find that significantly shallower quantum circuits can be used while maintaining high fidelity.
We validate this approach on two benchmark cases: a 3x3 transverse-field Ising model and random realizations of the Edwards-Anderson spin-glass model on a 4x4 lattice. We show that our soft-coded representation outperforms single-state (standard VQE) and multi-state (SSVQE/MCVQE) approaches, offering a favorable trade-off between circuit depth and accuracy.
Reference:
G. Clemente and M. Intini, arXiv:2602.05980 (2026)