RQC Colloquium

38th RQC Colloquium

• 講演者

  山本 直樹 教授
  (慶應義塾大学)

• 日程

  2025年10月21日（火）16:00-17:00 (4:00p.m.-5:00p.m.)（JST）

• 開催場所

  Hybrid (Zoom・和光地区 C00 本部棟 2階 大会議室)

• 講演タイトル

  Quantum Algorithmic Primitive for Quantum Machine Learning

• お問合せ

  rqc_colloquium_inquiry[at]ml.riken.jp

• 登録フォーム

  https://krs2.riken.jp/m/rqc_registration_form

講演概要
In this talk, I will begin with the topic of a quantum machine learning problem where we are interested in classifying the label of an unknown “quantum data” (specifically, the phase of a quantum ground state). The optimal measurement strategy for this problem needs full knowledge on the target state; hence I will show a circumventing method based on partial state tomography [1]. Yet we may have a route of coherently using the data to devise the optimal measurement. The promising method is the Density matrix exponentiation (DME),which is a general procedure that converts an unknown quantum state into the Hamiltonian evolution. The issue of DME is that it is proven to require O(1/epsilon) state copies in error epsilon. I’ll show a method [2] that goes beyond this lower bound and achieves O(log(1/epsilon)) or O(1) state copies, by using non-physical processes. This can realize a general-purpose quantum algorithm for property estimation, that achieves exponential circuit-depth reductions over existing protocols across various tasks; I will present quantum principal component analysis, quantum emulator, and entropy calculation, as examples.

References:
[1] Tanji, Yano, Yamamoto, Quantum phase classification via quantum hypothesis testing, arXiv:2504.04101, 2025
[2] Wada, Kato, Harada, Yamamoto, State-to-Hamiltonian conversion with a few copies, arXiv:2509.14791, 2025

Flyer: 38th RQC Colloquium Flyer