ConQUESTA: Topology-Aware Post-Training Structural Pruning of Quantum Neural Networks

Saim Rehman; Muhammad Shafique

Home/Archive/arXiv

arXiv · 2610.23415

ConQUESTA: Topology-Aware Post-Training Structural Pruning of Quantum Neural Networks

Saim Rehman·Muhammad Shafique

View original source ↗Read source document ↗Export metadata ↓

Abstract

Deploying quantum neural networks (QNNs) on sparse hardware requires preserving task accuracy while reducing compiler-induced two-qubit gates, SWAPs, and depth. We present ConQUESTA (\underline{C}onstrained \underline{Qu}antum \underline{E}ntangler \underline{S}election with \underline{T}opology \underline{A}wareness), a post training pruning framework that combines task saliency with compiler cost, screens candidate entanglers, and searches complete pruning masks under an explicit accuracy constraint. Unlike one-shot ranking, each mask is transpiled as a whole across multiple coupling topologies. Across 4-, 6-, and 8-qubit QNNs and eight model seeds at 55% pruning, ConQUESTA achieves mean resource reductions of 61.6% on Breast Cancer, 72.0% on Digits01, and 82.2% on Iris, exceeding the strongest matched simple baseline by 3.5, 1.7, and 8.9 percentage points, respectively. Fixed-mask recovery raises feasibility from 16/24 to 22/24 runs. In a fixed-k=4 scaling study on a scaling probe of 4 to 12-qubits, the raw mask space grew from 495 to 12.1M while the evaluated masks increase only from 189 to 305. Resource-weight changes alter depth/routing priorities with at most a 0.44-point mean validation shift.

Explore related subjects

quant-ph cs.AR cs.SY eess.SY

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Saim Rehman, Muhammad Shafique. 2026-10-16. ConQUESTA: Topology-Aware Post-Training Structural Pruning of Quantum Neural Networks. https://arxiv.org/abs/2610.23415

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

MIP*=RE

We show that the class MIP* of languages that can be decided by a classical verifier interacting with multiple all-powerful quantum provers sharing entanglement is equal to the class RE of recursively enumerable languages. Our proof builds upon the quantum low-degree test of (Natarajan and Vidick, FOCS 2018) and the classical low-individual degree test of (Ji, et al., 2020) by integrating recent developments from (Natarajan and Wright, FOCS 2019) and combining them with the recursive compression framework of (Fitzsimons et al., STOC 2019). An immediate byproduct of our result is that there is an efficient reduction from the Halting Problem to the problem of deciding whether a two-player nonlocal game has entangled value $1$ or at most $1/2$. Using a known connection, undecidability of the entangled value implies a negative answer to Tsirelson's problem: we show, by providing an explicit example, that the closure $C_{qa}$ of the set of quantum tensor product correlations is strictly included in the set $C_{qc}$ of quantum commuting correlations. Following work of (Fritz, Rev. Math. Phys. 2012) and (Junge et al., J. Math. Phys. 2011) our results provide a refutation of Connes' embedding conjecture from the theory of von Neumann algebras.

quant-ph↗

arXiv2026

Molecular Ground State Simulation by Subspace Restriction and Hund's Rule

The Subspace Restriction Scheme (SRS) is presented to address the qubit and resource constraints of molecular simulation on near-term quantum hardware. SRS restricts a molecular Hamiltonian to a selected determinant subspace before qubit encoding, so a $D$-dimensional subspace requires only $q=\lceil\log_2D\rceil$ qubits. We apply SRS to the established seniority-resolved Multi-Hund Subspaces $\mathrm{MHS}(Ω)$ and develop compact encodings for these sectors. At fixed seniority and filling fraction, MHS asymptotically uses about half as many qubits as a compact particle-conserving encoding. For the $\mathrm{H}_{22}$ chain, the register shrinks from $44$ to $20$ qubits. Compression can nevertheless increase Pauli-operator density and does not always reduce measurement cost. In noiseless VQE benchmarks, shallow circuits reach within $1.6$~mHa of the exact subspace minima using only $4$--$30$ CNOT gates. However, the corresponding subspace minima for LiH, HF, OH, and CH remain $4.39$--$9.73$~mHa above FCI. Full-space UCCSD reaches chemical accuracy but requires hundreds to thousands of CNOT gates. SRS therefore provides a controlled resource--accuracy trade-off when register width is the binding constraint and the target state is well represented within a selected seniority sector.

quant-ph↗

arXiv2026

Predicting Dynamics from Flows of the Eigenstate Thermalization Hypothesis

Analytical treatments of far-from-equilibrium quantum dynamics are few, even in well-thermalizing systems. The celebrated eigenstate thermalization hypothesis (ETH) provides a post hoc ansatz for the matrix elements of observables in the eigenbasis of a thermalizing Hamiltonian, given various response functions of those observables as input. However, the ETH cannot predict these response functions. We introduce a procedure, dubbed the statistical Jacobi approximation (SJA), to update the ETH ansatz after a perturbation to the Hamiltonian and predict perturbed response functions. The Jacobi algorithm diagonalizes the perturbation through a sequence of two-level rotations. The SJA implements these rotations statistically assuming the ETH throughout the diagonalization procedure, and generates integrodifferential flow equations for various form factors in the ETH ansatz. We approximately solve these flow equations using statistical information about Jacobi rotations as input, and predict both quench dynamics and autocorrelators in the thermal state of the perturbed Hamiltonian. The predicted dynamics agree with exact numerics in both random matrix models and spin chains, outperforming second-order time-dependent perturbation theory.

quant-ph↗