If you would like to
attend in person but do not have an INL access card, please let me know by
11am on Wednesday so that I can register you as a daily visitor.
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
Time: Wednesday, 23rd April 2025 14:00 (WEST)
Location: Vision Room
Speaker:
Daphne Wang (Quandela)
Title: Photonic-native convolutional neural networks
Abstract
Classical convolutional neural networks have revolutionized the field of machine learning, especially when dealing with images. The key to their success is their inductive bias: their structure mimics the image processing mechanism inside the brain. In the
field of quantum machine learning, analogues of such models have been investigated [1,2], each with their pros and cons. In this work, we consider a photonic-native convolutional neural network similar to the one of [2] consisting of passive linear optical
circuits and simple adaptative measurements. We then benchmark the performance and resource requirements of our photonic models against their classical counterparts.
[1] Iris Cong, Soonwon Choi, and Mikhail D. Lukin. "Quantum convolutional neural networks."
Nature Physics 15: 1273–1278 (2019).
[2] Léo Monbroussou et al. "Subspace preserving quantum convolutional neural network architectures."
Quantum Science and Technology 10: 025050 (2024).
–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––
Note. Information about this and other upcoming meetings (including slides for past talks) can be found at
https://www.ruisoaresbarbosa.com/qlocseminar
Note. Please let me know if you would like to stop receiving these announcements.