RQC Seminar

74th RQC Seminar

• Speaker

  Mr. Harvey Cao
  ( Imperial College London )

• Date

  16:00-17:00, August 10, 2023 (Thursday)

• Venue

  Hybrid(ZOOM・ Wako Main Research 3F 345-347 Seminar Room(C01)）

• Title

  Unsupervised learning of quantum scars using intrinsic dimension

• Inquiries

  rqc_info[at]ml.riken.jp

Abstract
Quantum many-body scarring is a phenomenon describing the weak ergodicity breaking that arises from an intruiging interplay between integrability and chaos in quantum many-body systems. In this regime the predictions made by the Eigenstate Thermalization Hypothesis no longer remain valid across the full spectrum. Instead, an isolated sector of atypical, non-thermal eigenstates exhibits slow relaxation dynamics which retain memory of the system's initial configuration, known as quantum many-body scars. Within the past decade, dimensionality reduction techniques from machine learning have been extensively applied for extracting universal properties of complex quantum systems, such as learning phases in an unsupervised setting. In this work we focus directly on the intrinsic dimension, a topological quantity which describes the number of variables required for minimal representation of a data set. We apply a recent method for estimation of intrinsic dimension using discrete observables typically arising in quantum simulation experiments to analyze the PXP model, an effective spin system with constraints imposed by nearest-neighbour interactions. Our scheme provides an experimentally accessible way of learning scar states in an unsupervised manner, as well as a novel perspective on the use of dimensionality reduction without the projection of data onto lower dimensions.