BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:On the construction of useful quantum kernels
DTSTART:20230329T123000Z
DTEND:20230329T133000Z
DTSTAMP:20230528T053846Z
UID:indico-event-1251853@indico.cern.ch
DESCRIPTION:The representation of data is of paramount importance for mach
ine learning methods. Kernel methods are used to enrich the feature repres
entation\, allowing better generalization. Quantum kernels implement effic
iently complex transformation encoding classical data in the Hilbert space
of a quantum system\, resulting in even exponential speedup. However\, we
need prior knowledge of the data to choose an appropriate parametric quan
tum circuit that can be used as quantum embedding.We propose an algorithm
that automatically selects the best quantum embedding through a combinator
ial optimization procedure that modifies the structure of the circuit\, ch
anging the generators of the gates\, their angles (which depend on the dat
a points)\, and the qubits on which the various gates act. Since combinato
rial optimization is computationally expensive\, we have introduced a crit
erion based on the exponential concentration of kernel matrix coefficients
around the mean to immediately discard an arbitrarily large portion of so
lutions that are believed to perform poorly.Contrary to the gradient-based
optimization (e.g. trainable quantum kernels)\, our approach is not affec
ted by the barren plateau by construction. We have used both artificial an
d real-world datasets to demonstrate the increased performance of our appr
oach with respect to randomly generated PQC. We have also compared the eff
ect of different optimization algorithms\, including greedy local search\,
simulated annealing\, and genetic algorithms\, showing that the algorithm
choice largely affects the result.About the speaker Massimiliano Incudin
i is PhD student at the Dep. of Computer Science\, University of Verona. I
n his studies\, he is focused on the development of quantum kernels for re
al-world applications.Collaborators The presentation is based on the join
t work with Alessandra Di Pierro and Francesco Martini available at arxiv
.org/abs/2209.11144.\n\nhttps://indico.cern.ch/event/1251853/
LOCATION:CERN Online
URL:https://indico.cern.ch/event/1251853/
END:VEVENT
END:VCALENDAR