Ultrafast pulse retrieval with one-shot unsupervised deep learning

Jesse W. Wilson

doi:10.1117/12.2677885

5 October 2023 Ultrafast pulse retrieval with one-shot unsupervised deep learning

Jesse W. Wilson

Proceedings Volume 12681, Ultrafast Nonlinear Imaging and Spectroscopy XI; 1268105 (2023) https://doi.org/10.1117/12.2677885
Event: SPIE Optical Engineering + Applications, 2023, San Diego, California, United States

Abstract

Recently, much attention has been drawn to deep learning models, namely deep Convolutional Neural Networks (CNNs) trained in a supervised fashion. Supervised learning is time consuming, requires a large dataset where the answer is already known, and often fails to generalize to new cases with patterns or noise statistics that were not foreseen in the training data. Unsupervised learning, on the other hand, does not use prior knowledge of the answer, can adapt to new data at hand, and can learn from single examples. A CNN can solve inverse problems in this way by learning to produce a result that, when passed through a forward model, recovers the experimental measurement. For example, a CNN can be trained in a supervised fashion to recover the complex field of a pulse from its Frequency-Resolved Optical Gating (FROG) measurement, by training it on thousands of examples of FROG measurements (the network inputs) and their corresponding optical pulses (the target outputs). In unsupervised learning, the network learns to transform a FROG measurement into an optical pulse, that, when passed through a forward model, recovers the original FROG measurement. We find this succeeds even on single FROG measurements, opening up new possibilities for analyzing measurements that fall outside the foreseeable distribution of training data.

(2023) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Jesse W. Wilson "Ultrafast pulse retrieval with one-shot unsupervised deep learning", Proc. SPIE 12681, Ultrafast Nonlinear Imaging and Spectroscopy XI, 1268105 (5 October 2023); https://doi.org/10.1117/12.2677885

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
5 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Education and training

Ultrafast phenomena

Machine learning

Inverse optics

Transient nonlinear optics

Deep learning

Electric fields

Show All Keywords

Keywords/Phrases

Search In:

Publication Years