Lightspeed Binary Neural Networks using Optical Phase-Change Materials

Lightspeed Binary Neural Networks using Optical Phase-Change Materials

Shahroodi, Taha (TU Delft Computer Engineering)
Cardoso, Rafaela (École Centrale de Lyon)
Zahedi, M.Z. (TU Delft Computer Engineering)
Wong, J.S.S.M. (TU Delft Computer Engineering)
Bosio, Alberto (École Centrale de Lyon)
O'Connor, Ian (École Centrale de Lyon)
Hamdioui, S. (TU Delft Quantum & Computer Engineering)

Quantum & Computer Engineering

2023

This paper investigates the potential of a compute-in-memory core based on optical Phase Change Materials (oPCMs) to speed up and reduce the energy consumption of the Matrix-Matrix-Multiplication operation. The paper also proposes a new data mapping for Binary Neural Networks (BNNs) tailored for our oPCM core. The preliminary results show a significant latency improvement irrespective of the evaluated network structure and size. The improvement varies from network to network and goes up to ~1053x.

Binary Neural Network
Optical PCM
Computation-In-Memory

http://resolver.tudelft.nl/uuid:6cc6ec26-54ee-4a64-b48c-2ef88fdfea1d

https://doi.org/10.23919/DATE56975.2023.10137229

IEEE, Piscataway

2023-12-02

979-8-3503-9624-9

Proceedings of the 2023 Design, Automation & Test in Europe Conference & Exhibition (DATE)

DATE 2023, 2023-04-17 → 2023-04-19, Antwerp, Belgium

Green Open Access added to TU Delft Institutional Repository 'You share, we take care!' - Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.

Institutional Repository

conference paper

© 2023 Taha Shahroodi, Rafaela Cardoso, M.Z. Zahedi, J.S.S.M. Wong, Alberto Bosio, Ian O'Connor, S. Hamdioui