Dice Question Streamline Icon: https://streamlinehq.com

Learning without strictly bounded interaction range

Determine whether a k-local Hamiltonian with potentially long-range interactions (e.g., beyond strictly bounded range) can be learned from access to e^{-iHt}, rather than assuming a hard cutoff on interaction range.

Information Square Streamline Icon: https://streamlinehq.com

Background

Much of the literature assumes Hamiltonians have bounded-range interactions (strict locality with a hard cutoff), but many physical models exhibit algebraic decay (e.g., power-law interactions).

The authors cite a central open question from prior work concerning what structural assumptions are truly necessary for Hamiltonian learning and contrast with classical settings where bounded local norms suffice.

References

However, many classes of Hamiltonians do not have this structure, and a central open question posed by prior work is to understand what kind of structure is necessary for Hamiltonian learning. Can we learn a local Hamiltonian without the assumption of strictly bounded range?

Structure learning of Hamiltonians from real-time evolution (2405.00082 - Bakshi et al., 30 Apr 2024) in Section 1 (Introduction), Question 2 (label: question:learning-without-locality)