Calibration-Robust Wavelength-Multiplexed Coherent Photonic Tensor Core for Optical Neural Inference
Keywords:
optical computing; photonic tensor core; coherent photonics; wavelength-division multiplexing; optical neural networks; calibrationAbstract
Optical neural inference can exploit the bandwidth and parallelism of photonic integrated circuits, but practical tensor cores must remain accurate under wavelength crosstalk, phase quantization, insertion-loss imbalance, and detector noise. Here we propose a calibration-robust coherent photonic tensor core that combines wavelength-division multiplexed inputs with a programmable Mach-Zehnder interferometer mesh and balanced coherent readout. The architecture maps a real-valued matrix-vector multiplication to a complex optical transform followed by signed electronic accumulation and a lightweight precompensation loop. In a 16-channel numerical evaluation with 6-bit phase control, 0.25 dB channel gain variation, and -28 dB nearest-neighbor crosstalk, calibration reduced the normalized matrix-vector error from 3.1 x 10^-2 to 7.8 x 10^-3. A synthetic 10-class optical-feature benchmark retained 96.4% inference accuracy at 24 dB optical SNR after noise-aware training. System-level energy accounting predicted 0.42 pJ per multiply-accumulate operation at 20 Gbaud, with the dominant consumption arising from modulators and transimpedance amplification rather than passive wave propagation. These results identify error budgets and calibration targets for scalable intelligent optical computing systems
References
(1) Shen, Y.; Harris, N. C.; Skirlo, S.; Prabhu, M.; Baehr-Jones, T.; Hochberg, M.; Sun, X.; Zhao, S.; Larochelle, H.; Englund, D.; Soljacic, M. Deep Learning with Coherent Nanophotonic Circuits. Nat. Photonics 2017, 11, 441-446. https://doi.org/10.1038/nphoton.2017.93
(2) Clements, W. R.; Humphreys, P. C.; Metcalf, B. J.; Kolthammer, W. S.; Walmsley, I. A. Optimal Design for Universal Multiport Interferometers. Optica 2016, 3, 1460-1465. https://doi.org/10.1364/OPTICA.3.001460
(3) Miller, D. A. B. Self-Configuring Universal Linear Optical Component. Photonics Res. 2013, 1, 1-15. https://doi.org/10.1364/PRJ.1.000001
(4) Feldmann, J.; Youngblood, N.; Karpov, M.; Gehring, H.; Li, X.; Stappers, M.; Le Gallo, M.; Fu, X.; Lukashchuk, A.; Raja, A. S.; Liu, J.; Wright, C. D.; Sebastian, A.; Kippenberg, T. J.; Pernice, W. H. P.; Bhaskaran, H. Parallel Convolution Processing Using an Integrated Photonic Tensor Core. arXiv 2020, arXiv:2002.00281.
(5) Bogaerts, W.; Perez, D.; Capmany, J.; Miller, D. A. B.; Poon, J.; Englund, D.; Morichetti, F.; Melloni, A. Programmable Photonic Circuits. Nature 2020, 586, 207-216.
(6) Shastri, B. J.; Tait, A. N.; Ferreira de Lima, T.; Pernice, W. H. P.; Bhaskaran, H.; Wright, C. D.; Prucnal, P. R. Photonics for Artificial Intelligence and Neuromorphic Computing. Nat. Photonics 2021, 15, 102-114.
(7) Tait, A. N.; Ferreira de Lima, T.; Zhou, E.; Wu, A. X.; Nahmias, M. A.; Shastri, B. J.; Prucnal, P. R. Neuromorphic Photonic Networks Using Silicon Photonic Weight Banks. Sci. Rep. 2017, 7, 7430.
(8) Harris, N. C.; Steinbrecher, G. R.; Prabhu, M.; Lahini, Y.; Mower, J.; Bunandar, D.; Chen, C.; Wong, F. N. C.; Baehr-Jones, T.; Hochberg, M.; Lloyd, S.; Englund, D. Quantum Transport Simulations in a Programmable Nanophotonic Processor. Nat. Photonics 2017, 11, 447-452. https://doi.org/10.1038/nphoton.2017.95
(9) Reck, M.; Zeilinger, A.; Bernstein, H. J.; Bertani, P. Experimental Realization of Any Discrete Unitary Operator. Phys. Rev. Lett. 1994, 73, 58-61.
(10) Nahmias, M. A.; Tait, A. N.; Shastri, B. J.; de Lima, T. F.; Prucnal, P. R. Photonic Multiply-Accumulate Operations for Neural Networks. IEEE J. Sel. Top. Quantum Electron. 2020, 26, 1-18.
