Sparse-QITE: Discovering Compact Generators for Quantum Imaginary-Time Evolution via Regularization Methods

arXiv:2610.20791 · quant-ph · Submitted 2026-10-14 · Read on arXiv

Listen

Transcript

Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.

Kai: Today's paper: "Sparse-QITE: Discovering Compact Generators for Quantum Imaginary-Time Evolution via Regularization Methods".

Mira: Sparse-QITE introduces a regularized formulation for quantum imaginary-time evolution that promotes compact local generators by selecting only a subset of the Pauli strings available within the QITE domain,

Kai: First, who's behind it and why it matters.

Paper summary: Kai: So, this paper is about Sparse-QITE: Discovering Compact Generators for Quantum Imaginary-Time Evolution via Regularization Methods. It basically addresses a big problem in quantum simulation where the cost of preparing ground states using imaginary time evolution grows really fast with how local the approximation is.

Mira: Right, so standard QITE has a cost that scales as four to the power of D, where D is the locality of your unitary approximation, meaning you've got four to the power of D local Pauli basis strings to deal with. This paper introduces Sparse-QITE as a regularized formulation that aims to promote these compact generators.

Lev: I’m interested in how it reduces the number of quantum operations and yields circuits with fewer gates and lower depth, because that’s what matters when we think about running this on real hardware, you know?

Kai: Exactly. The core thesis is that they derive rigorous bounds on the sparsification error and then benchmark this method using one-dimensional transverse-field Ising and antiferromagnetic Heisenberg models. They show that Sparse-QITE exhibits a controllable sparsity to accuracy trade-off while still retaining strong compression even in the high-accuracy regime.

Mira: And looking at their scalability experiments with respect to D, they found that the active support grows much more slowly than in standard QITE and it approximately saturates over the domain sizes they studied. This suggests that you don't need to increase your implemented generator terms just because you are looking at a larger domain, which is pretty significant.

Lev: That slow growth in active support is promising for error correction overheads because it means we don't have to deal with an exponentially growing number of Pauli strings during the evolution steps. But what about the measurement cost? The paper mentions that the candidate Pauli set on a domain of D spins can contain up to four to the power of D Pauli strings, which creates two main bottlenecks: measurement cost and circuit complexity.

Kai: That’s where they introduce two extensions, Adaptive SparseQITE and Active-Discovery Sparse-QITE. Adaptive SparseQITE adjusts the regularization strength at each update according to a local approximation tolerance, which provides additional compression when you match the trajectory accuracy.

Mira: And Active-Discovery Sparse-QITE addresses that remaining measurement cost by reusing discovered support from an initial discovery stage, which reduces subsequent updates to just what’s necessary. This is designed to reduce the number of Pauli expectation values that need to be measured during the rest of the evolution steps.

Lev: From a hardware view, reducing those required Pauli measurements directly cuts down on measurement time and associated errors, which is huge when you're looking at noisy superconducting qubits or whatever physical platform we're using.

Conclusion: Kai: So, to wrap up this paper on Sparse-QITE: Discovering Compact Generators for Quantum Imaginary-Time Evolution via Regularization Methods. The main point is that accurate QITE trajectories can have representations substantially more compact than the full local operator basis.

Mira: They show this by using regularization methods to select only a subset of the Pauli strings available within the QITE domain, and this directly reduces circuit resources and yields circuits with fewer gates and lower depth.

Lev: I think what they’re really saying is that this compression persists after you actually build the circuit and do the transpilation, which translates into substantial reductions in one-qubit gate counts and two-qubit gate counts.

Kai: They also introduced those variants, Adaptive SparseQITE and Active-Discovery SparseQITE, showing that sparsity-promoting regularization is a route toward more resource-efficient QITE. It’s about finding a way to make the simulation cheaper than the standard approach for certain problems.

Mira: Broadly speaking, these results suggest that sparsity-promoting regularization might also benefit variational quantum algorithms too, which is an interesting implication beyond just this specific imaginary-time evolution method.

Lev: The main limitation they flag is that they still require the expectation values associated with the full candidate pool for the basic formulation, and while Active-Discovery Sparse-QITE helps, it depends on that initial discovery stage. They also leave open questions about how compressible QITE generators are depending on locality or correlation length.

Kai: So, Sparse-QITE reduces circuit resources rather than just the formal number of generator coefficients. It's a way to get better performance when you don't know the domain required for convergence in advance.

Mira: That’s the gist of it, focusing on making these quantum simulation methods more resource efficient by using smart regularization techniques.

Rafael Gáomez-Lurbe, Luca Ion, Alberto Acevedo, Andreu Anglés-Castillo

Departament de Fósica Te`orica and IFIC, Universitat de Val`encia-CSIC · Departamento de Matem´atica, Fósica y Ciencias Tecnol´ogicas, Universidad CEU Cardenal Herrera · Instituto Universitario de Investigación de Matem´atica Pura y Aplicada, Universitat Politecnica de Val`encia

quant-ph

Submitted: 2026-10-14

Updated: 2026-10-14

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 90/100

The gist: Sparse-QITE introduces a regularized formulation for quantum imaginary-time evolution that promotes compact local generators by selecting only a subset of the Pauli strings available within the QITE

Key concepts

Quantum Imaginary-Time Evolution (QITE): QITE is a method used to prepare the ground state of a quantum system by evolving it in imaginary time. It involves applying a unitary operator repeatedly, which is computationally intensive because the required local operators scale exponentially with the system's locality.
Sparsity-Promoting Regularization: This technique uses an elastic-net formulation, specifically an l1 penalty, to encourage only essential Pauli strings to be included in the evolution. This regularization promotes compact generators by penalizing unnecessary coefficients, effectively selecting a smaller subset of the full local basis needed for accurate simulation.
Active Support: The active support at a specific time step $n$ is the set of Pauli strings that are non-zero after applying the evolution approximation. By tracking and minimizing the size of this active support, Sparse-QITE ensures that only the most relevant local interactions are retained for each step in imaginary time.

Terminology

Generator — 1× in this paper · explained in 2 episodes
Regularization — 1× in this paper · explained in 3 episodes

Summary

Sparse-QITE introduces a regularized formulation for quantum imaginary-time evolution that promotes compact local generators by selecting only a subset of the Pauli strings available within the QITE domain, which reduces circuit resources and yields circuits with fewer gates and lower depth<ref:2610.20791#pg2>.

How it works

Sparse-QITE addresses the rapidly growing cost of standard Quantum Imaginary-Time Evolution (QITE), where the local Pauli basis scales as 4D, by using sparsity-promoting regularization methods<ref:2610.20791#pg4>. It introduces an elastic-net formulation by minimizing an objective function that includes both the QITE dynamics approximation and a penalty term to promote sparsity, specifically using an l1 penalty which promotes vanishing coefficients and an l2 term to stabilize the determination<ref:2610.20791#pg5>. The active support for time step n is defined as Cn = nI ∈ ∪kPn,k,: a(n,k)λ,I ≠ 0 o, and its cardinality Cn gives the number of Pauli strings retained at imaginary-time step n<ref:2610.20791#pg6>.

Sparse-QITE Variants

The paper introduces two extensions to the basic Sparse-QITE formulation<ref:2610.20791#pg6>:

Adaptive SparseQITE selects the regularization strength dynamically during the evolution according to a local approximation tolerance, providing additional compression at matched trajectory accuracy<ref:2610.20791#pg6>. The choice of λe is determined by λe = sup λ ∈ [0, λmax]: R(λ) ≤ εobj, where R(λ) determines the local infidelity between SparseQITE and QITE states<ref:2610.20791#pg6>.
Active-Discovery Sparse-QITE reduces the number of Pauli expectation values that need to be evaluated by restricting subsequent updates to support identified during an initial discovery stage, which is based on a heuristic freezing criterion using the Jaccard overlap between cumulative supports<ref:2610.20791#pg9>. This allows for a direct reduction in the number of Pauli expectation values that must be measured during the remaining evolution steps<ref:2610.20791#pg9>.

Error Bounds and Stability

The paper derives rigorous bounds quantifying the trajectory error between Sparse-QITE and ordinary QITE<ref:2610.20791#pg6>. For the basic formulation, Proposition 1 provides a bound on the difference between the regularized and ordinary QITE coefficients<ref:2610.20791#pg9>. Corollary 3 establishes a trajectory-level estimate, showing that the discrepancy is bounded by δR ≤ eΛR ∆τ X R r=1 qe(r)λ,ρ + ΓR(∆τ)2<ref:2610.20791#pg29>. The stability of the ridge-QITE update is proven in Lemma 4, which shows that the local first-order remainder is bounded by Γ(∆τ)2, where Γ:= Γsp + Γridge<ref:2610.20791#pg30>.

Numerical Results and Scalability

Numerical benchmarks on the one-dimensional transverse-field Ising model (TFIM) and the isotropic Heisenberg model demonstrate substantial generator compression, including regimes with ground-state accuracy comparable to standard QITE<ref:2610.20791#pg10>. The results show that the active support grows much more slowly than in standard QITE and approximately saturates over the domain sizes studied<ref:2610.20791#pg4>. Furthermore, Sparse-QITE can eliminate many terms that are not required to attain the accuracy reached by standard QITE, yielding substantially more compact generators even in the high-accuracy regime<ref:2610.20791#pg6>. The results also show that the compression persists after circuit construction and transpilation, producing substantial reductions in gate counts and circuit depth<ref:2610.20791#pg6>.

Resource Cost Analysis

The analysis of compiled circuit resources shows that the reduction in active Pauli terms translates into actual circuit-resource savings, with the largest savings occurring at larger domain sizes<ref:2610.20791#pg32>. The compiled circuits follow the same qualitative trend as the number of active generator terms, translating into substantial reductions in both one- and two-qubit gate counts and circuit depth<ref:2610.20791#pg33>.

Conclusion

Sparse-QITE demonstrates that accurate QITE trajectories can admit representations substantially more compact than the full local operator basis, suggesting it may be especially useful when the domain required for convergence is not known in advance. The paper concludes that Sparse-QITE reduces circuit resources rather than only the formal number of generator coefficients<ref:2610.20791#pg6>. Several open questions remain regarding the compressibility of QITE generators depending on locality, correlation length, symmetry, entanglement, and proximity to criticality. The active-discovery strategy can also be generalized to allow operators to leave and re-enter the candidate pool. Testing Sparse-QITE under realistic noise will determine whether its lower gate counts and depths translate into improved hardware performance. The same regularization principle may apply to other quantum algorithms that determine generators or circuit parameters through optimization over large operator sets<ref:2610.20791#pg6>.

The gist Sparse-QITE introduces a regularized formulation for quantum imaginary-time evolution that promotes compact local generators by selecting only a subset of the Pauli strings available within the QITE domain, which reduces circuit resources and yields circuits with fewer gates and lower depth<ref:2610.20791#pg2>.

Improvements for AI systems

Sparse-QITE for Quantum Imaginary-Time Evolution (QITE) can generate circuits with fewer gates and lower depth by selecting a compact set of generators from the available operator pool, reducing circuit complexity significantly compared to standard QITE, especially at larger domain sizes D.
The Adaptive SparseQITE variant allows the regularization parameter lambda to vary dynamically, choosing it based on a local approximation tolerance, which provides additional compression at matched trajectory accuracy and controls the leading local state error directly.
The Active-Discovery Sparse-QITE variant reduces measurement requirements by reusing operators discovered during an initial stage, leading to a reduction in the number of required Pauli expectation values, converting learned sparsity into a direct reduction in the number of Pauli expectation values that need to be evaluated and consequently into a reduction in the number of shots required on the quantum computer.
The method provides a controllable sparsity–accuracy trade-off, allowing researchers to select the largest tested regularization strength lambda for which the final-state fidelity satisfies a prescribed threshold, enabling resource-efficient exploration of larger domains without inheriting the full growth of the standard QITE generator.

Abstract

Quantum imaginary-time evolution (QITE) provides a systematic route to ground-state preparation, but its cost grows rapidly with the locality D of the unitary approximation, for which the local Pauli basis scales as 4 D. We introduce Sparse-QITE, a regularized formulation that promotes compact generators, reducing the number of quantum operations and yielding circuits with fewer gates and lower depth. We derive rigorous bounds on the sparsification error and benchmark the method in one-dimensional transverse-field Ising and antiferromagnetic Heisenberg models. Sparse-QITE exhibits a controllable sparsity--accuracy trade-off while retaining strong compression even in the high-accuracy regime. In scalability experiments with respect to D, the active support grows much more slowly than in standard QITE and, in several regimes, approximately saturates over the domain sizes studied. After circuit construction and transpilation, some of the largest-domain cases require only about 1% of the standard-QITE gate count while maintaining high accuracy. We further introduce adaptive and active-discovery variants: the former selects the regularization strength dynamically during the evolution, while the latter reduces the number of Pauli expectation values that need to be evaluated and, consequently, the measurement requirements. These results establish sparsity-promoting regularization as a route toward more resource-efficient QITE. More broadly, these results suggest that sparsity-promoting regularization may also benefit variational quantum algorithms.