Scalable and Energy-Efficient Optical Computing Architectures
B Carpinlioglu1, U Tegin1
1 Electric and Electronic Engineering, Koç University, Istanbul, Turkey
Seminar: S12 — Optical Computing and Neural Networks
Monday, 6 July 2026 · 16:00 – 16:30
Abstract
Optics is a promisingly strong candidate for next-generation artificial intelligence applications owing to its inherent speed, parallelism and energy efficiency. In this talk, I will present our recent results on photonic neural networks spanning free-space [1, 2], fiber [3-6], and integrated optics [7] on the platform level and feed-forward, recurrent, and spiking models on the architectural level. I will describe how harnessing physical phenomena such as rogue waves [8], spatiotemporal nonlinear propagation in multimode fibers [9], and fiber laser cavities [10] serves as computing substrates for large-scale learning problems. Lastly, I will discuss a low-cost passive ultrafast imaging technique [11] that can increase the throughput of optical computing platforms.
References
- B Çarpınlıoğlu and U Teğin, Commun. Phys. 8, 349 (2025)
- F N Kılınç and U Teğin, arXiv: 2601.07574 (2026)
- A M I Muda and U Teğin, Opt. Express 33, 7852 (2025)
- B U. Kesgin and U Teğin, Nanophotonics 14, 2723 (2025)
- F Yüce, B Çarpınlıoğlu and U. Teğin, Opt. Lett. 51, 1239 (2026)
- B U Kesgin, F Yüce and U Teğin, Opt. Lett. 50, 5254 (2025)
- A M I Muda and U Teğin, arXiv: 2604.21301 (2026)
- B U Kesgin, G Y Durdu and U Teğin, arXiv: 2512.24983 (2025)
- D Eşlik, B U Kesgin and U Teğin, arXiv: 2602.19246 (2026)
- D Eşlik, B U Kesgin, F N Kılınç and U Teğin, arXiv: 2602.09519 (2026)
- D Eşlik, B U Kesgin and U Teğin, arXiv: 2604.27898 (2026)