Signed Convolution in Photonics with Phase-Change Materials using Mixed-Polarity Bitstreams
As AI continues to grow in importance, in order to reduce its carbon footprint and utilization of computer resources, numerous alternatives are under investigation to improve its hardware building blocks. In particular, in convolutional neural networks (CNNs), the convolution function represents the most important operation and one of the best targets for optimization. A new approach to convolution had recently emerged using optics, phase-change materials (PCMs) and stochastic computing, but is thus far limited to unsigned operands. In this paper, we propose an extension in which the convolutional kernels are signed, using mixed-polarity bitstreams. We present a proof of validity for our method, while also showing that, in simulation and under similar operating conditions, our approach is less affected by noise than the common approach in the literature.
Citation
If you build on this work, please cite the paper using the entry below. The BibTeX can be copied to clipboard with the button at the top of this page.
@article{raphael20248068657de02c561ee1bdf3eafd27cedcf52ab6d6,
title = {Signed Convolution in Photonics with Phase-Change Materials using Mixed-Polarity Bitstreams},
author = {Raphael Cardoso and Clément Zrounba and M.F. Abdalla and Paul Jiménez and Mauricio Gomes de Queiroz and B. Charbonnier and Fabio Pavanello and Ian O'Connor and S. L. Beux},
journal = {Asia and South Pacific Design Automation Conference},
year = {2024},
doi = {10.1109/ASP-DAC58780.2024.10473952}
} Acknowledgements
This work was supported in part by the Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grants programme and by the Fonds de recherche du Québec — Nature et technologies (FRQNT).