Femtosecond pulse parameter estimation from photoelectron momenta using machine learning

Tomasz Szołdra; Marcelo F. Ciappina; Nicholas Werby; Philip H. Bucksbaum; Maciej Lewenstein; Jakub Zakrzewski; Andrew S. Maxwell

doi:10.1088/1367-2630/acee19

Femtosecond pulse parameter estimation from photoelectron momenta using machine learning

Tomasz Szołdra, Marcelo F. Ciappina, Nicholas Werby, Philip H. Bucksbaum, Maciej Lewenstein, Jakub Zakrzewski, Andrew S. Maxwell^*

^*Corresponding author for this work

Physics

Research output: Contribution to journal › Article › peer-review

Abstract

Deep learning models have provided huge interpretation power for image-like data. Specifically, convolutional neural networks (CNNs) have demonstrated incredible acuity for tasks such as feature extraction or parameter estimation. Here we test CNNs on strong-field ionization photoelectron spectra, training on theoretical data sets to 'invert' experimental data. Pulse characterization is used as a 'testing ground', specifically we retrieve the laser intensity, where 'traditional' measurements typically lead to 20% uncertainty. We report on crucial data augmentation techniques required to successfully train on theoretical data and return consistent results from experiments, including accounting for detector saturation. The same procedure can be repeated to apply CNNs in a range of scenarios for strong-field ionization. Using a predictive uncertainty estimation, reliable laser intensity uncertainties of a few percent can be extracted, which are consistently lower than those given by traditional techniques. Using interpretability methods can reveal parts of the distribution that are most sensitive to laser intensity, which can be directly associated with holographic interferences. The CNNs employed provide an accurate and convenient ways to extract parameters, and represent a novel interpretational tool for strong-field ionization spectra.

Original language	English
Article number	083039
Journal	New Journal of Physics
Volume	25
Issue number	8
DOIs	https://doi.org/10.1088/1367-2630/acee19
State	Published - 1 Aug 2023

Keywords

convolutional neural networks
femtosecond pulse characterization
machine learning
strong-field ionization

Access to Document

10.1088/1367-2630/acee19

Cite this

@article{4f335408ad0a400284c3f6de8fd6b4b3,

title = "Femtosecond pulse parameter estimation from photoelectron momenta using machine learning",

abstract = "Deep learning models have provided huge interpretation power for image-like data. Specifically, convolutional neural networks (CNNs) have demonstrated incredible acuity for tasks such as feature extraction or parameter estimation. Here we test CNNs on strong-field ionization photoelectron spectra, training on theoretical data sets to 'invert' experimental data. Pulse characterization is used as a 'testing ground', specifically we retrieve the laser intensity, where 'traditional' measurements typically lead to 20% uncertainty. We report on crucial data augmentation techniques required to successfully train on theoretical data and return consistent results from experiments, including accounting for detector saturation. The same procedure can be repeated to apply CNNs in a range of scenarios for strong-field ionization. Using a predictive uncertainty estimation, reliable laser intensity uncertainties of a few percent can be extracted, which are consistently lower than those given by traditional techniques. Using interpretability methods can reveal parts of the distribution that are most sensitive to laser intensity, which can be directly associated with holographic interferences. The CNNs employed provide an accurate and convenient ways to extract parameters, and represent a novel interpretational tool for strong-field ionization spectra.",

keywords = "convolutional neural networks, femtosecond pulse characterization, machine learning, strong-field ionization",

author = "Tomasz Szo{\l}dra and Ciappina, {Marcelo F.} and Nicholas Werby and Bucksbaum, {Philip H.} and Maciej Lewenstein and Jakub Zakrzewski and Maxwell, {Andrew S.}",

note = "Publisher Copyright: {\textcopyright} 2023 The Author(s). Published by IOP Publishing Ltd on behalf of the Institute of Physics and Deutsche Physikalische Gesellschaft.",

year = "2023",

month = aug,

day = "1",

doi = "10.1088/1367-2630/acee19",

language = "英语",

volume = "25",

journal = "New Journal of Physics",

issn = "1367-2630",

publisher = "IOP Publishing Ltd.",

number = "8",

}

TY - JOUR

T1 - Femtosecond pulse parameter estimation from photoelectron momenta using machine learning

AU - Szołdra, Tomasz

AU - Ciappina, Marcelo F.

AU - Werby, Nicholas

AU - Bucksbaum, Philip H.

AU - Lewenstein, Maciej

AU - Zakrzewski, Jakub

AU - Maxwell, Andrew S.

PY - 2023/8/1

Y1 - 2023/8/1

N2 - Deep learning models have provided huge interpretation power for image-like data. Specifically, convolutional neural networks (CNNs) have demonstrated incredible acuity for tasks such as feature extraction or parameter estimation. Here we test CNNs on strong-field ionization photoelectron spectra, training on theoretical data sets to 'invert' experimental data. Pulse characterization is used as a 'testing ground', specifically we retrieve the laser intensity, where 'traditional' measurements typically lead to 20% uncertainty. We report on crucial data augmentation techniques required to successfully train on theoretical data and return consistent results from experiments, including accounting for detector saturation. The same procedure can be repeated to apply CNNs in a range of scenarios for strong-field ionization. Using a predictive uncertainty estimation, reliable laser intensity uncertainties of a few percent can be extracted, which are consistently lower than those given by traditional techniques. Using interpretability methods can reveal parts of the distribution that are most sensitive to laser intensity, which can be directly associated with holographic interferences. The CNNs employed provide an accurate and convenient ways to extract parameters, and represent a novel interpretational tool for strong-field ionization spectra.

AB - Deep learning models have provided huge interpretation power for image-like data. Specifically, convolutional neural networks (CNNs) have demonstrated incredible acuity for tasks such as feature extraction or parameter estimation. Here we test CNNs on strong-field ionization photoelectron spectra, training on theoretical data sets to 'invert' experimental data. Pulse characterization is used as a 'testing ground', specifically we retrieve the laser intensity, where 'traditional' measurements typically lead to 20% uncertainty. We report on crucial data augmentation techniques required to successfully train on theoretical data and return consistent results from experiments, including accounting for detector saturation. The same procedure can be repeated to apply CNNs in a range of scenarios for strong-field ionization. Using a predictive uncertainty estimation, reliable laser intensity uncertainties of a few percent can be extracted, which are consistently lower than those given by traditional techniques. Using interpretability methods can reveal parts of the distribution that are most sensitive to laser intensity, which can be directly associated with holographic interferences. The CNNs employed provide an accurate and convenient ways to extract parameters, and represent a novel interpretational tool for strong-field ionization spectra.

KW - convolutional neural networks

KW - femtosecond pulse characterization

KW - machine learning

KW - strong-field ionization

UR - http://www.scopus.com/inward/record.url?scp=85169583073&partnerID=8YFLogxK

U2 - 10.1088/1367-2630/acee19

DO - 10.1088/1367-2630/acee19

M3 - 文章

AN - SCOPUS:85169583073

SN - 1367-2630

VL - 25

JO - New Journal of Physics

JF - New Journal of Physics

IS - 8

M1 - 083039

ER -

Femtosecond pulse parameter estimation from photoelectron momenta using machine learning

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this