arXiv subjects

Alena S. Mastiukova

Publications and source records attributed to Alena S. Mastiukova.

4 recordsLinked to original sources

arXiv2025

Regularizing quantum loss landscapes by noise injection

The difficulty of training variational quantum algorithms and quantum machine learning models is well established. In particular, quantum loss landscapes are often highly non-convex and dominated by poor local minima. While this renders their training NP-hard in general, efficient heuristics that work well for typical instances may still exist. Here, we propose a protocol that uses a targeted noise injection to smooth and regularize quantum loss landscapes. It works by exponentially suppressing the high-frequency components in the Fourier expansion of the quantum loss function. The protocol can be efficiently implemented both in hardware and in simulations. We observe significant and robust improvements of solution quality across various problem types. Our method can be combined with existing techniques mitigating the local minima, such as the quantum natural gradient optimizer, and adds to the toolbox of methods for optimizing quantum loss functions.

quant-ph↗

arXiv2023

Quantum-inspired optimization for wavelength assignment

Problems related to wavelength assignment (WA) in optical communications networks involve allocating transmission wavelengths for known transmission paths between nodes that minimize a certain objective function, for example, the total number of wavelengths. Playing a central role in modern telecommunications, this problem belongs to NP-complete class for a general case, so that obtaining optimal solutions for industry relevant cases is exponentially hard. In this work, we propose and develop a quantum-inspired algorithm for solving the wavelength assignment problem. We propose an advanced embedding procedure for this problem into the quadratic unconstrained binary optimization (QUBO) form having an improvement in the number of iterations with price-to-pay being a slight increase in the number of variables ("spins"). Then we compare a quantum-inspired technique for solving the corresponding QUBO form against classical heuristic and industrial combinatorial solvers. The obtained numerical results indicate on an advantage of the quantum-inspired approach in a substantial number of test cases against the industrial combinatorial solver that works in the standard setting. Our results pave the way to the use of quantum-inspired algorithms for practical problems in telecommunications and open a perspective for the further analysis of the employ of quantum computing devices.

quant-ph↗

arXiv2023

Suppressing decoherence in quantum state transfer with unitary operations

Decoherence is the fundamental obstacle limiting the performance of quantum information processing devices. The problem of transmitting a quantum state (known or unknown) from one place to another is of great interest in this context. In this work, by following the recent theoretical proposal [Opt. Eng. {\bf 59}, 061625 (2020)] we study an application of quantum state-dependent pre- and post-processing unitary operations for protecting the given (multi-qubit) quantum state against the effect of decoherence acting on all qubits. We observe the increase in the fidelity of the output quantum state both in a quantum emulation experiment, where all protecting unitaries are perfect, and in a real experiment with a cloud-accessible quantum processor, where protecting unitaries themselves are affected by the noise. We expect the considered approach can be useful for analyzing capabilities of quantum information processing devices in transmitting known quantum states. We also demonstrate an applicability of the developed approach for suppressing decoherence in the process of distributing a two-qubit state over remote physical qubits of a quantum processor.

quant-ph↗

arXiv2022

Multiclass classification using quantum convolutional neural networks with hybrid quantum-classical learning

Multiclass classification is of great interest for various applications, for example, it is a common task in computer vision, where one needs to categorize an image into three or more classes. Here we propose a quantum machine learning approach based on quantum convolutional neural networks for solving the multiclass classification problem. The corresponding learning procedure is implemented via TensorFlowQuantum as a hybrid quantum-classical (variational) model, where quantum output results are fed to the softmax activation function with the subsequent minimization of the cross entropy loss via optimizing the parameters of the quantum circuit. Our conceptional improvements here include a new model for a quantum perceptron and an optimized structure of the quantum circuit. We use the proposed approach to solve a 4-class classification problem for the case of the MNIST dataset using eight qubits for data encoding and four ancilla qubits; previous results have been obtained for 3-class classification problems. Our results show that accuracies of our solution are similar to classical convolutional neural networks with comparable numbers of trainable parameters. We expect that our finding provide a new step towards the use of quantum neural networks for solving relevant problems in the NISQ era and beyond.

quant-ph↗