QRC Seminars - Prof. Jens Eisert

QRC Seminars - Quantum Research Center

Prof. Jens Eisert

Freie Universität Berlin

12th July 2021 - 4.00pm - 5.00pm (GST)



Learning classical and quantum dynamical laws from data


Traditionally, physical laws are being formulated in a largely
heuristic fashion and subsequently their predictions are empirically
explored. Generations of physics students are being told that
Hamiltonians govern the dynamics of physical systems both in the
quantum and classical realm. While this is perfectly right, much less
is said on how these Hamiltonian are determined or characterized in
the first place. Often, a-priori knowledge of some sort is available,
but then the question emerges of how one can be sure that the actual
Hamiltonian is close to the anticipated one based on physical
reasoning. This issue seems particularly pressing for complex systems
involving many degrees of freedom, or for systems in the quantum
technologies where high precision is imperative.

These basic yet profound insights motivate efforts to learn
Hamiltonians - or directly physical laws - from data. In the first
part of this talk, we will be concerned with new ways of learning
classical dynamical laws from data. We move on to learn
instances of quantum Hamiltonians from data, and show how
superconducting devices as experimented with by the Google AI team can
be characterized to unprecedented precision. We will see how one
can set up a tensor network based and machine learning inspired way of
learning quantum many-body Hamiltonians from dynamical data. If
time allows, I will mention aspects of rigorously minded
quantum-assisted machine learning and of the recovery of quantum
processes from data. In an outlook, we will discuss further
perspectives of data-driven approaches in identifying physical laws
from data.