Quantum computation at the edge of chaos
Tomohiro Hashizume, Zhengjun Wang, Frank Schlawin, Dieter Jaksch
Abstract
A key challenge in classical machine learning is to mitigate overparameterization by selecting sparse solutions. We translate this concept to the quantum domain, introducing quantum sparsity as a principle based on minimizing quantum information shared across multiple parties. This allows us to address fundamental issues in quantum data processing and convergence issues such as the barren plateau problem in Variational Quantum Algorithm (VQA). We propose a practical implementation of this principle using the topological Entanglement Entropy (TEE) as a cost function regularizer. A non-negative TEE is associated with states with a sparse structure in a suitable basis, while a negative TEE signals untrainable chaos. The regularizer, therefore, guides the optimization along the critical edge of chaos that separates these regimes. We link the TEE to structural complexity by analyzing quantum states encoding functions of tunable smoothness, deriving a quantum Nyquist-Shannon sampling theorem that bounds the resource requirements and error propagation in VQA. Numerically, our TEE regularizer demonstrates significantly improved convergence and precision for complex data encoding and ground-state search tasks. This work establishes quantum sparsity as a design principle for robust and efficient VQAs.
AI Impact Assessments
(3 models)Scientific Impact Assessment: "Quantum computation at the edge of chaos"
1. Core Contribution
The paper introduces "quantum sparsity" as a design principle for variational quantum algorithms (VQAs), analogizing the classical machine learning strategy of regularization to penalize overparameterization. The central proposal is to use the topological entanglement entropy (TEE) — specifically the tripartite information — as a cost function regularizer that constrains optimization to the "edge of chaos," a regime between trainable but expressive parameter space and the chaotic, barren-plateau regime.
The paper makes three interconnected claims: (i) TEE serves as a diagnostic for the onset of information scrambling in parameterized quantum circuits, (ii) a quantum Nyquist-Shannon sampling theorem (QNSST) bounds the resources needed for amplitude encoding, and (iii) TEE regularization practically improves VQA convergence. The conceptual framing — linking barren plateaus to information scrambling and proposing a regularizer grounded in quantum information geometry — is creative and offers a fresh perspective on a well-studied problem.
2. Methodological Rigor
Analytical results: The exact MPS representation of amplitude-encoded sine functions (Appendix A) is rigorous and elegantly derived. The resulting threshold for faithful encoding is a clean result. The circuit depth scaling argument (Eq. 2, ) relies on a recursive construction with several assumptions (e.g., similar spectral structure across subdomains, number of discontinuities independent of ), which limits its generality but provides useful intuition.
TEE analysis of Weierstrass functions: Using the Weierstrass function as a tunable-complexity test case is clever and well-motivated. The transition from non-negative to negative TEE as fractal dimension increases (Fig. 2b) is convincing. However, the connection between TEE and classical sparsity (the threshold) is derived under strong assumptions (uniform amplitudes, diagonal-dominant reduced density operators) and may not generalize cleanly to realistic quantum states.
Numerical benchmarks: The two VQA benchmarks (turbulent flow encoding, AF-2D-NNH ground state search) are performed on qubits with layers, which is quite small. While the results show improved convergence (1-2 orders of magnitude in cost function value), the statistical analysis is based on 100 realizations, and the distributions shown in Figs. 3d-e reveal significant overlap between regularized and unregularized cases. The exponentially decaying schedule introduces hyperparameters () that require tuning, somewhat undermining the claimed generality.
Gradient variance analysis (Fig. 3b): The gradient analysis of shows cliff-like behavior rather than exponential decay, which is promising but only demonstrated for specific circuit architectures and moderate system sizes ().
3. Potential Impact
The barren plateau problem is arguably the most critical obstacle to practical VQA deployment, making any progress highly valued. If the TEE regularizer scales favorably, it could become a standard component of VQA optimization pipelines. The paper's contribution is particularly valuable because:
However, the practical impact depends on scalability beyond . The measurement overhead for the regularizer ( to ) is polynomial but adds non-trivial cost, especially when combined with the need for gradient evaluation via parameter-shift rules.
4. Timeliness & Relevance
This work addresses a timely need. The barren plateau problem remains a central challenge as the quantum computing community attempts to demonstrate practical advantage with VQAs. Recent work (Larocca et al., Nature Reviews Physics 2025, cited here) has systematized understanding of barren plateaus, and this paper adds a constructive mitigation strategy. The "edge of chaos" framing also connects to recent interest in complexity theory and quantum scrambling dynamics.
The QNSST contribution is timely given growing interest in quantum algorithms for scientific computing (CFD, PDE solving), where amplitude encoding of classical data is a bottleneck.
5. Strengths & Limitations
Strengths:
Limitations:
Missing elements: No noise analysis (relevant for NISQ applications), no comparison with tensor-network-based approaches for the same problems, and no wall-clock time comparisons accounting for regularizer overhead.
Summary
This is a creative and well-crafted paper that proposes an intellectually appealing framework connecting quantum information scrambling, sparsity, and VQA trainability. The analytical contributions (exact MPS, QNSST) are solid, and the numerical demonstrations, while limited in scale, show promising trends. The main question is whether the approach scales — both computationally and in terms of practical advantage — to problem sizes where barren plateaus truly manifest. The work opens interesting directions but falls short of providing the rigorous guarantees needed to definitively resolve the barren plateau problem.
Generated Apr 20, 2026
Comparison History (47)
Paper 1 addresses the barren plateau problem, a fundamental and widely recognized bottleneck in variational quantum algorithms across all hardware platforms. Its introduction of quantum sparsity and topological entanglement entropy regularization offers a novel, broad-impact solution bridging quantum information and machine learning. Paper 2, while methodologically rigorous and mathematically deep, focuses specifically on neutral atom routing architectures, giving it a narrower scope of application and potentially less overall scientific impact compared to the algorithmic advancements in Paper 1.
Paper 2 has higher likely impact due to broader scope and applicability: it proposes a general design principle (“quantum sparsity”) addressing a widely felt bottleneck in NISQ-era algorithms (barren plateaus) and offers an implementable regularizer (TEE) with numerical evidence across tasks. It also connects to learning theory via a quantum sampling theorem, potentially influencing VQAs, quantum machine learning, and complexity/chaos studies. Paper 1 is rigorous and timely but more specialized (thermal state prep via collision models) with narrower cross-field reach.
Paper 1 addresses the barren plateau problem, a critical bottleneck in Variational Quantum Algorithms. By introducing quantum sparsity and a TEE regularizer to improve algorithmic convergence, it offers high real-world applicability in the rapidly growing field of quantum machine learning. While Paper 2 presents a valuable fundamental refinement for mQED and van der Waals interactions, Paper 1's methodological innovation and potential to enable robust near-term quantum computing applications grant it a significantly broader and more immediate scientific impact.
Paper 2 addresses the barren plateau problem in VQAs—a central challenge in quantum computing—by introducing a novel 'quantum sparsity' principle using topological entanglement entropy as a regularizer. It bridges concepts from classical ML, quantum information theory, and topological order, deriving a quantum Nyquist-Shannon theorem. Its breadth of impact (quantum computing, ML, complexity theory), practical applicability to VQA design, and timeliness give it higher potential impact than Paper 1, which, while rigorous and elegant, provides more specialized results connecting spin Hamiltonians to quantum Fisher information.
Paper 2 offers a general, rigorous structural characterization of continuous-time non-Markovian dynamics under physically central constraints (complete positivity and non-signalling), yielding an integro-differential extension of GKSL and a multi-time correlation formalism that avoids a regression theorem. This is broadly applicable across open quantum systems, quantum optics, control, metrology, and noise modeling, and directly targets realistic non-Markovian environments. Paper 1 is timely for VQAs and may be impactful in quantum ML, but its scope is narrower and depends more on heuristic/empirical regularization choices.
Paper 2 introduces a novel conceptual framework connecting quantum sparsity, topological entanglement entropy, and the edge of chaos to address the barren plateau problem in VQAs. It bridges concepts from classical ML (sparsity), topological physics (TEE), and complexity theory, offering broader interdisciplinary impact. The quantum Nyquist-Shannon sampling theorem and the practical TEE regularizer provide both theoretical depth and practical utility. Paper 1, while solid, presents an incremental extension of adaptive variational methods to multi-constraint state preparation, with narrower scope and less conceptual novelty.
Paper 1 introduces a novel theoretical framework connecting quantum sparsity, topological entanglement entropy, and the barren plateau problem in variational quantum algorithms. It derives a quantum Nyquist-Shannon sampling theorem and provides a practical regularization technique with numerical validation. This addresses a fundamental and widely recognized challenge (barren plateaus) in quantum computing with broad implications across quantum machine learning and algorithm design. Paper 2, while experimentally valuable in identifying a specific TLS-related readout failure mechanism in transmon qubits, has a narrower scope focused on a specific hardware defect characterization.
Paper 1 introduces a novel theoretical framework connecting quantum sparsity, topological entanglement entropy, and the barren plateau problem in variational quantum algorithms. It provides both theoretical contributions (quantum Nyquist-Shannon sampling theorem) and practical tools (TEE regularizer) with broad applicability across quantum computing and machine learning. Paper 2, while valuable, reports a specific experimental observation about TLS-induced readout failures—an incremental addition to the known catalog of TLS-related decoherence mechanisms. Paper 1's breadth of impact, novelty, and potential to influence VQA design principles give it higher estimated impact.
Paper 1 is more novel and potentially broader in impact: it introduces “quantum sparsity” as a general design principle for VQAs, proposes an unusual and theoretically grounded regularizer based on topological entanglement entropy to steer training to the edge of chaos, and derives a quantum Nyquist–Shannon-type sampling theorem linking function smoothness to resource/error scaling. This combination of conceptual framework, theoretical results, and demonstrated training improvements could influence VQA design across optimization, learning, and simulation. Paper 2 is valuable but closer to existing adaptive-ansatz themes, with narrower scope.
Paper 1 demonstrates a concrete, dramatic reduction in the physical qubit requirements for breaking RSA-2048 by an order of magnitude using QLDPC codes, which has immediate and profound implications for both quantum computing hardware development and cryptographic security. This result directly impacts national security policy, post-quantum cryptography migration timelines, and quantum hardware engineering targets. Paper 2 introduces interesting theoretical concepts connecting quantum sparsity and barren plateaus via topological entanglement entropy, but VQA improvements are incremental in a crowded field with uncertain practical quantum advantage. Paper 1's specificity and real-world urgency give it greater impact.
Paper 1 addresses the barren plateau problem, a major fundamental bottleneck in Variational Quantum Algorithms, by introducing a novel algorithmic framework based on quantum sparsity and topological entanglement entropy. Its theoretical contributions, including a quantum Shannon-Nyquist theorem, offer broad applicability across quantum machine learning and optimization. In contrast, Paper 2, while highly innovative in proposing the first magnonic GKP state preparation, focuses on a specific physical realization of existing error-correction states, making its immediate impact narrower compared to the algorithmic advancements of Paper 1.
Paper 2 likely has higher impact due to its broad, timely relevance to near-term quantum computing: it targets major bottlenecks in variational quantum algorithms (barren plateaus, trainability) with a potentially practical regularization method (TEE-based) and claims both theoretical results (quantum sparsity principle, sampling theorem) and numerical improvements. Its ideas could influence quantum ML, optimization, and quantum information. Paper 1 is methodologically deep and novel for superconducting circuit foundations, but its immediate applicability is narrower and mainly impacts circuit-QED theory rather than multiple fast-moving application areas.
Paper 1 introduces a novel, broadly applicable design principle (“quantum sparsity”) for variational quantum algorithms, with a concrete regularizer (TEE) aimed at a central, timely bottleneck (barren plateaus) and supported by theory (sampling/resource bounds) plus numerical evidence across tasks. This combination of conceptual innovation, methodological linkage to complexity, and direct relevance to near-term quantum computing suggests wide cross-field uptake (QML, VQAs, optimization, quantum information). Paper 2 is rigorous and foundational for superconducting circuit quantization, but its impact is narrower and more incremental within a mature framework.
Paper 1 addresses the barren plateau problem, arguably the most critical bottleneck in near-term quantum computing (NISQ) and Variational Quantum Algorithms. By introducing quantum sparsity and a TEE regularizer, it offers a fundamental solution with broad implications across quantum machine learning, optimization, and chemistry. Paper 2 presents a highly innovative spectroscopy technique for quantum materials, but its impact is more localized to condensed matter physics and materials characterization. Therefore, Paper 1 demonstrates broader potential impact across the rapidly growing field of quantum computing.
Paper 2 addresses the barren plateau problem, a critical and widespread bottleneck in quantum machine learning and Variational Quantum Algorithms. By introducing quantum sparsity and using topological Entanglement Entropy as a regularizer, it offers a fundamental algorithmic improvement. This has broader, field-wide implications for quantum computing compared to Paper 1, which proposes a specialized, albeit innovative, nanoscale spectroscopy technique for specific quantum materials.
Paper 1 addresses the fundamental barren plateau problem in Variational Quantum Algorithms (VQAs) by introducing quantum sparsity and a novel topological Entanglement Entropy regularizer. The derivation of a quantum Nyquist-Shannon sampling theorem and the concept of optimizing at the 'edge of chaos' offer broad, foundational theoretical advancements. Paper 2, while important, focuses on the more niche area of adversarial robustness in rotationally equivariant quantum models, which has a narrower scope compared to resolving general trainability and convergence issues in QML.
Paper 1 addresses the critical barren plateau problem in variational quantum algorithms by introducing a novel quantum sparsity principle grounded in topological entanglement entropy. It bridges concepts from classical machine learning, quantum information, and topological physics, deriving a quantum Nyquist-Shannon theorem and demonstrating practical numerical improvements. Its breadth of impact spans quantum computing, machine learning, and information theory, with immediate practical applications to VQA design. Paper 2, while mathematically rigorous in characterizing exceptional point hierarchies, addresses a more specialized topic in non-Hermitian physics with narrower immediate applications.
Paper 2 likely has higher impact due to a clear experimental advance with near-term architectural relevance: on-demand strong interaction from a weakly coupled qubit, enabling deterministic Fock-state generation and SWAP in high-Q cavities—core primitives for bosonic quantum computing. The work is methodologically rigorous (hardware demonstration, timing, fidelity characterization) and directly applicable to scalable superconducting platforms. Paper 1 is conceptually novel and potentially broad, but appears more theoretical/algorithmic with impact contingent on wider validation across VQA settings and practical adoption.
Paper 2 likely has higher impact: it resolves a recent open question by showing 2-Forrelation is solvable with highly restricted IQP (commuting) circuits, and strengthens oracle separations related to BQP vs PH—core complexity-theory results with broad relevance across quantum computation and theoretical CS. The contributions appear methodologically rigorous (explicit constructions, complexity consequences, new Fourier growth bounds, key algebraic identity) and timely. Paper 1 is innovative and potentially valuable for NISQ/VQA training, but its impact depends more on empirical robustness and adoption; its theoretical claims (TEE as regularizer, “edge of chaos”) may be less universally foundational.
Paper 1 addresses the barren plateau problem, a critical bottleneck in near-term quantum machine learning, by introducing a novel TEE regularizer. Its practical application to improving Variational Quantum Algorithms offers immediate, broad impact across quantum chemistry and optimization. Paper 2 provides an elegant theoretical proof for query complexity in partial search, but has a narrower scope and less immediate real-world applicability.