Emergent Order in Classical Data Representations on Ising Spin Models

Encoding classical data on quantum spin Hamiltonians yields ordered spin ground states which are used to discriminate data types for binary classification. The Ising Hamiltonian is a typical spin model to encode classical data onto qubits, known as the ZZ feature map. We assess the ground states of...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:arXiv.org 2023-03
Hauptverfasser: Kirk, Jorja J, Jackson, Matthew D, King, Daniel J M, Intallura, Philip, Metcalf, Mekena
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Encoding classical data on quantum spin Hamiltonians yields ordered spin ground states which are used to discriminate data types for binary classification. The Ising Hamiltonian is a typical spin model to encode classical data onto qubits, known as the ZZ feature map. We assess the ground states of the Ising Hamiltonian encoded with three separate data sets containing two classes of data. A new methodology is proposed to predict a certain data class using the ground state of the encoded Ising Hamiltonian. Ground state observables are obtained through quantum simulation on a quantum computer, and the expectation values are used to construct a classical probability distribution on the state space. Our approach is a low dimensional representation of the exponentially large feature space. The antiferromagnetic ground state is the stable ground state for the one dimensional chain lattice and the 2D square lattice. Frustration induces unique ordered states on the triangle lattice encoded with data, hinting at the possibility for an underlying phase diagram for the model. We examine order stability with data scaling and data noise.
ISSN:2331-8422