arXiv subjects

Takashi Imamichi

Publications and source records attributed to Takashi Imamichi.

6 recordsLinked to original sources

arXiv2025

Qiskit Machine Learning: an open-source library for quantum machine learning tasks at scale on quantum hardware and classical simulators

We present Qiskit Machine Learning (ML), a high-level Python library that combines elements of quantum computing with traditional machine learning. The API abstracts Qiskit's primitives to facilitate interactions with classical simulators and quantum hardware. Qiskit ML started as a proof-of-concept code in 2019 and has since been developed to be a modular, intuitive tool for non-specialist users while allowing extensibility and fine-tuning controls for quantum computational scientists and developers. The library is available as a public, open-source tool and is distributed under the Apache version 2.0 license.

quant-ph↗

arXiv2021

Approximate Solutions of Combinatorial Problems via Quantum Relaxations

Combinatorial problems are formulated to find optimal designs within a fixed set of constraints. They are commonly found across diverse engineering and scientific domains. Understanding how to best use quantum computers for combinatorial optimization is to date an open problem. Here we propose new methods for producing approximate solutions for the maximum cut problem and its weighted version, which are based on relaxations to local quantum Hamiltonians. These relaxations are defined through commutative maps, which in turn are constructed borrowing ideas from quantum random access codes. We establish relations between the spectra of the relaxed Hamiltonians and optimal cuts of the original problems, via two quantum rounding protocols. The first one is based on projections to random magic states. It produces average cuts that approximate the optimal one by a factor of least 0.555 or 0.625, depending on the relaxation chosen, if given access to a quantum state with energy between the optimal classical cut and the maximal relaxed energy. The second rounding protocol is deterministic and it is based on estimation of Pauli observables. The proposed quantum relaxations inherit memory compression from quantum random access codes, which allowed us to test the performances of the methods presented for 3-regular random graphs and a design problem motivated by industry for sizes up to 40 nodes, on superconducting quantum processors.

quant-ph↗

arXiv2020

Scheduling of Operations in Quantum Compiler

When scheduling quantum operations, a shorter overall execution time of the resulting schedule yields a better throughput and higher fidelity output. In this paper, we demonstrate that quantum operation scheduling can be interpreted as a special type of job-shop problem. On this basis, we provide its formulation as Constraint Programming while taking into account commutation between quantum operations. We show that this formulation improves the overall execution time of the resulting schedules in practice through experiments with a real quantum compiler and quantum circuits from two common benchmark sets.

quant-ph↗

arXiv2019

Efficient evaluation of quantum observables using entangled measurements

The advent of cloud quantum computing has led to the rapid development of quantum algorithms. In particular, it is necessary to study variational quantum-classical hybrid algorithms, which are executable on noisy intermediate-scale quantum (NISQ) computers. Evaluations of observables appear frequently in the variational quantum-classical hybrid algorithms for NISQ computers. By speeding up the evaluation of observables, it is possible to realize a faster algorithm and save resources of quantum computers. Grouping of observables with separable measurements has been conventionally used, and the grouping with entangled measurements has also been proposed recently by several teams. In this paper, we show that entangled measurements enhance the efficiency of evaluation of observables, both theoretically and experimentally by taking into account the covariance effect, which may affect the quality of evaluation of observables. We also propose using a part of entangled measurements for grouping to keep the depth of extra gates constant. Our proposed method is expected to be used in conjunction with other related studies. We hope that entangled measurements would become crucial resources, not only for joint measurements but also for quantum information processing.

quant-ph↗

arXiv2019

Optimization of Quantum Circuit Mapping using Gate Transformation and Commutation

This paper addresses quantum circuit mapping for Noisy Intermediate-Scale Quantum (NISQ) computers. Since NISQ computers constraint two-qubit operations on limited couplings, an input circuit must be transformed into an equivalent output circuit obeying the constraints. The transformation often requires additional gates that can affect the accuracy of running the circuit. Based upon a previous work of quantum circuit mapping that leverages gate commutation rules, this paper shows algorithms that utilize both transformation and commutation rules. Experiments on a standard benchmark dataset confirm the algorithms with more rules can find even better circuit mappings compared with the previously-known best algorithms.

quant-ph↗

arXiv2018

Profile-guided memory optimization for deep neural networks

Recent years have seen deep neural networks (DNNs) becoming wider and deeper to achieve better performance in many applications of AI. Such DNNs however require huge amounts of memory to store weights and intermediate results (e.g., activations, feature maps, etc.) in propagation. This requirement makes it difficult to run the DNNs on devices with limited, hard-to-extend memory, degrades the running time performance, and restricts the design of network models. We address this challenge by developing a novel profile-guided memory optimization to efficiently and quickly allocate memory blocks during the propagation in DNNs. The optimization utilizes a simple and fast heuristic algorithm based on the two-dimensional rectangle packing problem. Experimenting with well-known neural network models, we confirm that our method not only reduces the memory consumption by up to $49.5\%$ but also accelerates training and inference by up to a factor of four thanks to the rapidity of the memory allocation and the ability to use larger mini-batch sizes.

cs.DC↗