arXiv subjects

Siheon Park

Publications and source records attributed to Siheon Park.

7 recordsLinked to original sources

arXiv2026

Improved sample complexity bound for sample-based Lindbladian simulation

We establish improved sample-complexity bounds for sample-based Lindbladian simulation based on the Wave Matrix Lindbladization (WML) algorithm. For a jump operator $L$ with dimension $d$, we derive an explicit non-asymptotic sample complexity bound $n_d^*(t,\varepsilon) \le \left( \frac{2d+3}{8} \right) \|L\|_\infty^2 \left( \frac{t^2}{\varepsilon} \right)$, holding for simulation time $t$ and error $\varepsilon$. This refines the dimension dependence of the best previously known bound, $O(d^2 t^2/\varepsilon)$, from [Go et al., Quantum Sci. Tech. 10, 045058 (2025)]. Remarkably, we show that this dimensional overhead can be entirely avoided when $\| L\|_\infty^2 = O(1/d)$, a condition satisfied with high probability for random Lindblad operators, yielding a typical-case sample complexity of $O(t^2/\varepsilon)$. On the other hand, in the worst case, we show that WML necessarily requires $\Omega(dt^2/\varepsilon)$ samples by constructing an explicit example with a rank-one Lindblad operator. Our results reveal a sharp dichotomy between typical and adversarial sample complexities in Lindbladian simulation, thereby strengthening the theoretical foundations of sample-based quantum algorithms.

quant-ph↗

arXiv2026

Code-agnostic bosonic noise suppression with hybrid rotations

Physical-level noise on traveling bosonic modes remains a critical bottleneck for scalable quantum information processing. We show that for any single-mode bosonic code (qumode) corrupted by thermal or Gaussian displacement noise at loss rate $\mu$ and amplification $G$, a hybrid continuous-discrete-variable (CV-DV) interferometer using a single qubit ancilla and two controlled-Fourier (CF) gates sandwiching the noise channel suppresses its effects from linear to quadratic scaling. This is achieved without active error correction or destructive measurements of the encoded state, maintaining high success probabilities $\geq 0.5$ when $\mu G \leq 0.5$. When supplemented with multiple ancillas, the protocol converts photon loss into coherent Fock-damping, and thermal or displacement noise into a mixture of Fock-diagonal noise. The protocol is entirely code-agnostic. For the special case of $2^K$-fold rotation-symmetric bosonic codes, it simplifies to conventional error detection and projection with $K$ ancillas. Suppression with simple gates and few ancillas demonstrates a clear hardware-efficient advantage over previously proposed ``bypass'' schemes, where quantum information transferred to the DV ancillas is readily corrupted by ancilla noise. Finally, we extend the protocol to a qutrit DV ancilla. This demonstrates resilience to both CV noise and composite DV damping noise, achieving a truly hybrid noise suppression scheme that operates effectively even on CV encodings lacking a well-defined photon-number parity syndrome.

quant-ph↗

arXiv2025

Ballistic bosonic noise suppression with hybrid qumode-qubit rotation gates

Noise suppression is of paramount importance for reliable quantum information processing and computation. We show that for any single-mode bosonic code (qumode) corrupted by thermal~noise at rate~$\eta$ and mean \mbox{excitation}~$\bar{n}$, a hybrid continuous-discrete-variable~(CV-DV) interferometer using only a single qubit ancilla~(DV) and two controlled~Fourier~(CF) gates sandwiching the noise channel suppresses its effects to $\mathcal{O}(\eta^2)$ \emph{without} any active error correction or destructive measurements of the encoded state and with high success probabilities~$>0.5$ if~$\eta(1+\bar{n})<0.5$. This suppression scheme works by conditionally monitoring the photon-number parities after the interferometer. Bosonic codes with two logical states of the same photon-number parity (like-parity codes) are \emph{completely resilient} to DV amplitude- and phase-damping ancilla noise. For such codes, the interferometer simplifies to the use of a qumode rotation gate and a \emph{single} CF~gate. This presents a clear advantage of our CF-gate-based error suppression scheme over previously-proposed ``bypass'' protocols, where qubit information transferred to the DV mode is readily corrupted by damping~noise. Finally, we present a simple extension to direct communication of qumode states between two parties over a noisy channel using a preshared DV entangled state, by implementing a CF gate in the first laboratory and its inverse in the other. Such a communication protocol achieves a similar fidelity performance at the same success rate as the single-party case, but with greater resilience to the ancilla noise than DV~teleportation. Resource-efficient multi-qubit codes that depend on a few essential long-range interactions can benefit from it.

quant-ph↗

arXiv2025

Variational Quantum Approximated Spectral Clustering

Clustering is a fundamental task for analyzing unlabeled data based solely on its underlying distribution. Spectral clustering is a clustering method that represents a dataset as a graph and uses the relationships between data points. However, classical spectral clustering methods incur high computational costs that can scale cubically with the dataset size-as is typical for approaches that involve solving eigenvalue problems. In this work, we propose Variational Quantum Approximated Spectral Clustering (VQASC), which extends quantum distance-based classifier models to the clustering framework. Our approach uses efficient quantum circuit designs whose depth scales sub-quadratically with dataset size, enabling the computation of weighted sums over various matrix representations of an undirected graph. Furthermore, we adopt an empirical risk formulation to reduce the impact of local minima arising from parameterized quantum circuits, and we validate our approach through simulations on real-world datasets.

quant-ph↗

arXiv2024

Sample-based Hamiltonian and Lindbladian simulation: Non-asymptotic analysis of sample complexity

Density matrix exponentiation (DME) is a quantum algorithm that processes multiple copies of a program state $\sigma$ to realize the Hamiltonian evolution $e^{-i \sigma t}$. Wave matrix Lindbladization (WML) similarly processes multiple copies of a program state $\psi_L$ in order to realize a Lindbladian evolution. Both algorithms are prototypical sample-based quantum algorithms and can be used for various quantum information processing tasks, including quantum principal component analysis, Hamiltonian simulation, and Lindbladian simulation. In this work, we present detailed sample complexity analyses for DME and sample-based Hamiltonian simulation, as well as for WML and sample-based Lindbladian simulation. In particular, we prove that the sample complexity of DME is no larger than $4t^2/\varepsilon$ for evolution time $t$ and imprecision level $\varepsilon$ quantified by the normalized diamond distance. We also establish a fundamental lower bound on the sample complexity of sample-based Hamiltonian simulation, which matches our DME sample complexity bound up to a constant multiplicative factor. Additionally, we prove that the sample complexity of WML is no larger than $3t^2d^2/\varepsilon$, where $d$ is the dimension of the space on which the Lindblad operator acts nontrivially, and we prove a lower bound of $10^{-4} t^2/\varepsilon$ on the sample complexity of sample-based Lindbladian simulation. These results prove that WML is optimal for sample-based Lindbladian simulation whenever the Lindblad operator acts nontrivially on a constant-sized system. Finally, we point out that the DME sample complexity analysis in [Kimmel et al., npj Quantum Information 3, 13 (2017)] and the WML sample complexity analysis in [Patel and Wilde, Open Systems \& Information Dynamics 30, 2350010 (2023)] appear to be incomplete, highlighting the need for the results presented here.

quant-ph↗

arXiv2024

Optimizing Quantum Convolutional Neural Network Architectures for Arbitrary Data Dimension

Quantum convolutional neural networks (QCNNs) represent a promising approach in quantum machine learning, paving new directions for both quantum and classical data analysis. This approach is particularly attractive due to the absence of the barren plateau problem, a fundamental challenge in training quantum neural networks (QNNs), and its feasibility. However, a limitation arises when applying QCNNs to classical data. The network architecture is most natural when the number of input qubits is a power of two, as this number is reduced by a factor of two in each pooling layer. The number of input qubits determines the dimensions (i.e. the number of features) of the input data that can be processed, restricting the applicability of QCNN algorithms to real-world data. To address this issue, we propose a QCNN architecture capable of handling arbitrary input data dimensions while optimizing the allocation of quantum resources such as ancillary qubits and quantum gates. This optimization is not only important for minimizing computational resources, but also essential in noisy intermediate-scale quantum (NISQ) computing, as the size of the quantum circuits that can be executed reliably is limited. Through numerical simulations, we benchmarked the classification performance of various QCNN architectures when handling arbitrary input data dimensions on the MNIST and Breast Cancer datasets. The results validate that the proposed QCNN architecture achieves excellent classification performance while utilizing a minimal resource overhead, providing an optimal solution when reliable quantum computation is constrained by noise and imperfections.

quant-ph↗

arXiv2023

Variational Quantum Approximate Support Vector Machine with Inference Transfer

A kernel-based quantum classifier is the most practical and influential quantum machine learning technique for the hyper-linear classification of complex data. We propose a Variational Quantum Approximate Support Vector Machine (VQASVM) algorithm that demonstrates empirical sub-quadratic run-time complexity with quantum operations feasible even in NISQ computers. We experimented our algorithm with toy example dataset on cloud-based NISQ machines as a proof of concept. We also numerically investigated its performance on the standard Iris flower and MNIST datasets to confirm the practicality and scalability.

quant-ph↗