SAMOP 2023 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
MO: Fachverband Molekülphysik
MO 3: Interaction with Strong or Short Laser Pulses I (joint session A/MO)
MO 3.3: Vortrag
Montag, 6. März 2023, 11:45–12:00, F303
Transfer learning and visualization of a convolutional neural network for recognition of the internuclear distance in a molecule from electron momentum distributions — •Nikolay Shvetsov-Shilovski and Manfred Lein — Leibniz Universität Hannover
We use a convolutional neural network (CNN) to retrieve the internuclear distance in the two-dimensional H2+ molecule ionized by an intense few-cycle laser pulse from the photoelectron momentum distributions [1]. We study the effect of the carrier-envelope phase on the retrieval of the internuclear distance with a CNN [2]. By using the transfer learning technique, we make our CNN applicable to momentum distributions obtained at the parameters it was not explicitly trained for. We compare the CNN with alternative approaches that are shown to have very limited transferability. Finally, we use the occlusion sensitivity technique to extract features of the momentum distributions that allow a CNN to predict the internuclear distance.
[1] N. I. Shvetsov-Shilovski and M. Lein, Phys. Rev. A, 105, L021102 (2022).
[2] N. I. Shvetsov-Shilovski and M. Lein, submitted to Phys. Rev. A, arXiv:2211.01210.