Pith. sign in

REVIEW 3 major objections 8 minor 56 references

Seamless Optical Cloud Computing across Edge-Metro Network for Generative AI

T0 review · 3 major / 8 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper proposes a cloud computing architecture where edge nodes send data and model weights as light over a metro network to an AWGR-based optical computing center, claiming energy efficiency of 118.6 mW/TOPS and demonstrating image…

desk verdict Credible optical-computing proof-of-concept with a real experimental core, but the abstract overclaims the generative-AI validation since only the first convolution layer ran optically. read the letter →

arxiv 2412.12126 v2 pith:TXPGXRKP submitted 2024-12-04 cs.DC cs.CVcs.LGeess.IVeess.SP

Sizhe Xing , Aolong Sun , Chengxi Wang , Yizhi Wang , Boyu Dong , Junhui Hu , Xuyu Deng , An Yan
show 13 more authors
Yingjun Liu Fangchen Hu Zhongya Li Ouhan Huang Junhao Zhao Yingjun Zhou Ziwei Li Jianyang Shi Xi Xiao Richard Penty Qixiang Cheng Nan Chi Junwen Zhang
This is my paper · ORCID
classification cs.DCcs.CVcs.LGeess.IVeess.SP
keywords opticalcloudcomputingphotonicarrayedwaveguidegratinggenerativeAIenergyefficiencyedge-metronetworkneuralconvolutional
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that generative AI cloud computing can run on light rather than on electronic servers: edge nodes encode both their input data and the neural-network weights onto an optical frequency comb, send them through an existing edge-metro fiber network to a cloud optical computing center, and receive convolution results back. The authors report an energy efficiency of 118.6 mW/TOPS for a single optical processing unit, about two orders of magnitude lower than a leading electronic GPU, and demonstrate handwritten digit recognition plus several image-to-image generation tasks (sketch-to-object, season transfer, semantic segmentation, depth estimation). The system also performs communication and computation in the same optical link, transmitting images at 50 Gbps over 80 km of fiber during the experiments. The paper is explicit that, in the generative AI demonstrations, only the first convolutional layer was executed optically while all remaining layers were simulated at matching bit precision. If the architecture scales to many parallel optical processing units, it would reduce the energy and physical security burden of model weights in AI clouds.

What carries the argument

The central object is the AWGR-based optical processing unit. The AWGR's cyclic wavelength routing property maps input ports and wavelengths to output ports so that the summed optical intensity at each output is exactly a convolution of the input vector with a kernel; the kernel's positive and negative components ride on two adjacent free spectral ranges and are separated by microring filters before balanced photodetection. Because the input data and weights travel on the same frequency comb from the edge node, the model weights never have to be stored or uploaded to the cloud, which the paper identifies as the physical-layer security advantage.

What would settle it

Run the full pix2pix and CycleGAN networks with every convolutional layer executed by multiple synchronized OPUs over the metro link, and compare the generated images against the paper's noise-injected 7-bit simulation. If accumulated optical noise, inter-OPU crosstalk, or timing misalignment pushes SSIM below 0.92, FID above 39.8, or LPIPS above 0.23, then the reported image-generation quality does not transfer to the complete optical architecture.

Watch

Extended reading notes

Core claim

The central claim is that a wavelength-routing arrayed waveguide grating (AWGR) can act as a remote, in-network optical convolution engine, and that this engine is sufficient to run the convolution-heavy front end of generative models across a metro network. By loading positive and negative parts of a convolution kernel onto two free spectral ranges of the AWGR and subtracting them with a balanced photodetector, a single optical processing unit (OPU) performs a convolution in one clock slot, operating at 3.6 TOPS with 118.6 mW/TOPS. The authors experimentally show 7-bit computational accuracy at 10 Gbaud and use the system to achieve 88% accuracy on MNIST digit classification, and to generate images with pix2pix and CycleGAN tasks at quality comparable to 7-bit electronic computing, measured by SSIM 0.92, FID 39.8, and LPIPS 0.23.

Load-bearing premise

The results depend on the assumption that adding Gaussian noise at 7-bit precision in simulation, followed by fine-tuning on part of the data, faithfully represents what the remaining convolutional layers would do if they were actually executed on many optical processing units in parallel; only the first layer was run optically in the experiment.

Editorial extensions

If this is right

  • Multiple independent OPUs can process different convolution kernels of the same network layer in parallel, so a generative model can be split across an array of optical computing nodes.
  • An optical cloud computing center could be deployed inside metro networks, using existing fiber and wavelength-division multiplexing to serve many edge nodes without moving weights or data into a central server farm.
  • Scaling the AWGR from 8 ports to 64 ports would raise the projected peak computing rate to roughly 148.8 TOPS per OPU, more than 40 times the demonstrated 3.6 TOPS, since computational rate grows quadratically with port count while power grows linearly.
  • If the reported 7-bit optical precision holds, the architecture is compatible with the precision requirements of current CNN-based generative models, which the paper demonstrates through its SSIM, FID, and LPIPS comparisons.
  • Since the optical comb source sits at the edge, the cloud-side OPUs avoid laser power consumption, which supports denser packing of computing chips in the optical data center.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial extension: the paper's power accounting treats the laser as part of the transceiver rather than the OPU; a full end-to-end comparison that includes comb generation, modulation, and all electronic control overhead would clarify how much of the two-order-of-magnitude advantage survives at system level.
  • Editorial extension: the security benefit is physical rather than cryptographic, so a natural next question is whether an attacker who can observe the optical outputs from the network could reconstruct the weights or inputs; the paper does not analyze this side-channel.
  • Editorial extension: the demonstrations are convolution-dominated CNNs; applying the same remote-OPU idea to transformer-based generative models would require an optical matrix-vector multiplier or attention mechanism, which the AWGR convolver does not directly provide.
  • Editorial extension: the noise-injected fine-tuning recipe, if it holds for full multi-OPU execution, could serve as a general method for deploying photonic accelerators inside larger end-to-end deep learning models.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 8 minor

Summary. The paper proposes an optical cloud computing architecture in which input data and model weights are modulated onto an optical frequency comb at edge nodes, transmitted over an edge-metro network, and processed in the cloud by AWGR-based optical processing units (OPUs). The authors experimentally demonstrate 7-bit arithmetic at 10 Gbaud, communication over 80 km, MNIST classification at 88%, and a first-layer optical convolution for generative image tasks. They report an energy efficiency of 118.6 mW/TOPS and claim a two-order-of-magnitude reduction relative to electronic cloud computing, with image generation demonstrated using pix2pix and CycleGAN. The central caveat, stated in the main text and Methods, is that only the first convolutional layer was executed optically; all subsequent layers were simulated with injected 7-bit quantization noise and the model was fine-tuned on a portion of the data. The power-efficiency number is also a component-based estimate rather than a measured wall-plug figure.

Significance. If the claims were fully supported, the architecture would be a notable step toward delocalized optical accelerators, building on the cyclic routing properties of AWGRs and extending prior delocalized photonic learning work. The paper has clear strengths: the experimental setup uses commercially available components, the source code is made available, and the measured RMSE of 0.0304 for a single optical convolution layer provides a concrete error benchmark. However, the significance is currently limited by the mismatch between the abstract's claim of experimental validation for generative AI and the actual scope of the optical experiments. The energy-efficiency claim also needs to be recast as an estimate. The system concept and the single-layer optical validation are valuable, but the full generative-AI and efficiency claims are not yet supported.

major comments (3)
  1. [Experimentally Demonstration of Image-generation Tasks; Methods: The architecture of the large-scale model] The abstract states that 'it is experimentally validated that this architecture can perform various complex generative AI models through parallel computing to achieve image generation tasks,' but the main text says 'only the calculations for the first convolutional layer were conducted in the optical domain during the experiment, while all other operations were performed on a computer.' The Methods further explain that the remaining layers are modeled by adding Gaussian white noise with a specified bit quantization level, and that the model is fine-tuned on a portion of the data after deployment. The reported SSIM, FID, and LPIPS values therefore apply to a network in which 14 of 15 convolutional layers are simulated, not to the full optical architecture. The measured first-layer RMSE of 0.0304 does not bound error accumulation through residual blocks, skip connections, normalization layers, AWGR crosstalk, MZM nonlinearity, thermal or polarization drift, EDFA noise, or inter-OPU synchronization skew. Fine-tuning on real data can absorb part of the simulation-experiment mismatch, so the image-generation metrics do not validate the full system. Please either revise the abstract and Results to state explicitly that the generative results are simulation-based after the first layer, or provide experiments that execute additional layers optically.
  2. [Power Efficiency; Methods: Power consumption] The abstract and Discussion describe 118.6 mW/TOPS as an 'experimental validation,' but the Methods derive this number from a component power budget using assumed values for wall-plug efficiency (eta=0.3), photodetector responsivity (R=0.65 A/W), bias voltages, and DAC/ADC control electronics. No direct measurement of the operating OPU's wall-plug power is reported. The comparison with the Nvidia H200 at 20.6 W/TOPS also uses different system boundaries: the OPU budget excludes edge-side lasers and appears to count only computing-related components. Please relabel 118.6 mW/TOPS as an estimated component-level budget, specify exactly which components are included and excluded, and avoid the phrase 'experimental validation' for this number unless a direct power measurement is provided.
  3. [Computing speed; Power Efficiency] The statement that 'power consumption scales linearly with component size, while the maximum computational rate scales quadratically' is asserted without supporting measurements or a validated physical model for AWGR port scaling. The paper itself acknowledges that precision degrades with accumulated noise and that SNR is influenced by the number of wavelengths received, so the quadratic-scaling efficiency gain is not established by the present experiments. If this is intended as a projection, it should be explicitly labeled as such and the assumptions should be stated.
minor comments (8)
  1. [Abstract and throughout] The units 'TOPs' and 'TOPS' are used inconsistently; please standardize to 'TOPS' throughout.
  2. [Equations (1)-(4)] The notation switches from optical intensity I_q in Eq. (2) to E_q in Eq. (4) after balanced detection; please clarify the relationship between these quantities.
  3. [Reference list] Reference 42 is an incomplete citation ('optical_fiber_communication-libre.pdf') and should be replaced with a full bibliographic entry.
  4. [Fig. 5e] The caption for Fig. 5e does not define the metric plotted on the axes; the reader cannot tell whether the comparison is SSIM, FID, LPIPS, or another quantity.
  5. [Supplementary Note numbering] The main text refers to Supplementary Notes 3, 4, 7, and 8, but the numbering appears inconsistent with the descriptions of the generative-AI training and power-budget details; please align the references.
  6. [Methods: Details of optical cloud computing experiment] The calibration paragraph mentions 'the signal is first modulated at the -3dB point' and later says 'the signal is asymmetrical' where 'asymmetric' appears to be a typo for 'symmetric'; please correct.
  7. [Power consumption] The variable subscripts for power are inconsistent (e.g., P_laser, P_TEC, P_MZM, P_MRM, P_Pd); please define all symbols in one place and use them consistently.
  8. [Abstract] The abstract states 'an operational rate of 10 GHz' but the experiments are performed at 10 Gbaud; please use consistent terminology for baud rate.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation found: the OPU equations are standard AWGR routing properties, the energy-efficiency figure is an accounting calculation, and the generative-AI limitation is an acknowledged experimental scope restriction rather than a self-referential prediction.

full rationale

The paper's central derivations are not circular. Equations (1)-(4) describe the cyclic wavelength routing of an AWGR and the balanced photodetection that realizes convolution; these are standard device-physics identities and do not encode any of the claimed performance outcomes such as 118.6 mW/TOPS, 88% MNIST accuracy, or the SSIM/FID/LPIPS values. The energy-efficiency number is a component-power accounting calculation (426.92 mW / 3.6 TOPS) and is not fitted to the reported accuracy or image-quality metrics. The generative-AI validation is the closest candidate for circularity: the paper states, "Due to experimental limitations, only the calculations for the first convolutional layer were conducted in the optical domain during the experiment, while all other operations were performed on a computer," and the Methods describe adding "Gaussian white noise with a specified bit quantization level" to simulate the chip, followed by "fine-tuning the model using a portion of the data." However, this is an acknowledged experimental limitation, not a definitional identity. The reported image metrics are emergent outputs of a trained network after a real optical first layer with measured RMSE 0.0304; they are not equal by construction to the injected noise level. Fine-tuning on a portion of data is calibration that weakens external validity, but the paper does not claim to predict those metrics from the noise model. No load-bearing self-citation chain is present: the AWGR routing rule is standard, and the cited prior work is used for scaling context rather than to forbid alternatives. The main concerns are experimental scope and overstatement, not circular derivation.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim rests on standard photonic hardware plus two load-bearing modeling assumptions: digital simulation stands in for unexecuted optical layers, and a component-power model stands in for a measured energy budget. No new physical entities are postulated.

free parameters (4)
  • Wall-plug efficiency eta = 0.3
    Assumed energy conversion efficiency from electrical to optical power for the laser and EDFAs; not measured in the experiment. Directly scales the headline 118.6 mW/TOPS calculation.
  • Photodetector power model = 3.9 mW per PD (R=0.65 A/W, V_bias=2 V, P_opt=3 mW)
    Chosen component parameters used to estimate PD power consumption; not directly measured.
  • DAC/ADC control electronics power = 320.12 mW for 8 DACs and 6 ADCs in the computing module
    Estimated control electronics power, not measured in the experiment; included in the 426.92 mW computation-part budget.
  • AWGR thermal stabilizer power = 600 uW
    Taken from a cited prior work (ref 55), not measured in this setup.
assumptions (5)
  • domain assumption AWGR cyclic wavelength routing relation (Eq. 1) and impulse-function filtering (Eq. 2)
    The convolution result depends on ideal AWGR routing with negligible crosstalk and a comb linewidth narrow enough to treat filtering as an impulse.
  • domain assumption Positive and negative kernel parts over two FSRs are cleanly separated by microrings and subtracted by a balanced photodetector (Eq. 4)
    Assumes the microring filters fully separate the two FSRs and that the balanced detector has no significant imbalance.
  • ad hoc to paper Noise-injected digital simulation at matched bit precision represents the remaining optical layers
    The full generative models are evaluated by simulating all layers after the first at the hardware precision; the paper explicitly states only the first layer was optical.
  • ad hoc to paper Power consumption scales linearly with component count while computing rate scales quadratically with AWGR ports
    Used to project 148.8 TOPS for a 64-port AWGR; no noise or crosstalk analysis is provided for the larger device.
  • domain assumption Storing weights only at the edge provides physical-layer computing security
    The security claim assumes no other exfiltration paths, such as output-side inference attacks or side channels, and that edge storage is physically isolated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Seamless Optical Cloud Computing across Edge-Metro Network for Generative AI." pith.science (2026). https://pith.science/paper/TXPGXRKP

@misc{pith2026241212126,
  author       = {Pith},
  title        = {Pith review of: Seamless Optical Cloud Computing across Edge-Metro Network for Generative AI},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TXPGXRKP}},
  note         = {Machine review of arXiv:2412.12126}
}
read the original abstract

The rapid advancement of generative artificial intelligence (AI) in recent years has profoundly reshaped modern lifestyles, necessitating a revolutionary architecture to support the growing demands for computational power. Cloud computing has become the driving force behind this transformation. However, it consumes significant power and faces computation security risks due to the reliance on extensive data centers and servers in the cloud. Reducing power consumption while enhancing computational scale remains persistent challenges in cloud computing. Here, we propose and experimentally demonstrate an optical cloud computing system that can be seamlessly deployed across edge-metro network. By modulating inputs and models into light, a wide range of edge nodes can directly access the optical computing center via the edge-metro network. The experimental validations show an energy efficiency of 118.6 mW/TOPs (tera operations per second), reducing energy consumption by two orders of magnitude compared to traditional electronic-based cloud computing solutions. Furthermore, it is experimentally validated that this architecture can perform various complex generative AI models through parallel computing to achieve image generation tasks.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

56 extracted references · 43 canonical work pages

  1. [1]

    Zhou, Y ., Guo, C., Wang, X., Chang, Y . & Wu, Y . A Survey on Data Augmentation in Large Model Era. Preprint at http://arxiv.org/abs/2401.15422 (2024)

    arXiv 2024

  2. [2]

    Wang, M., Fu, W., He, X., Hao, S. & Wu, X. A Survey on Large -Scale Machine Learning. IEEE Trans. Knowl. Data Eng. 1–1 (2020) doi:10.1109/TKDE.2020.3015777

    arXiv 2020

  3. [3]

    Jiang, P., Ergu, D., Liu, F., Cai, Y . & Ma, B. A Review of Yolo Algorithm Developments. Procedia Computer Science 199, 1066–1073 (2022)

    work page 2022

  4. [4]

    Ramesh, A. et al. Zero-Shot Text-to-Image Generation. in Proceedings of the 38th International Conference on Machine Learning 8821–8831 (PMLR, 2021)

    work page 2021

  5. [5]

    Min, B. et al. Recent Advances in Natural Language Processing via Large Pre - trained Language Models: A Survey. ACM Comput. Surv. 56, 1–40 (2024)

    work page 2024

  6. [6]

    Chang, Y . et al. A Survey on Evaluation of Large Language Models. ACM Trans. Intell. Syst. Technol. 15, 1–45 (2024)

    work page 2024

  7. [7]

    Fei, N. et al. Towards artificial general intelligence via a multimodal foundation model. Nat Commun 13, 3094 (2022)

    work page 2022

  8. [8]

    Alouffi, B. et al. A Systematic Literature Review on Cloud Computing Security: Threats and Mitigation Strategies. IEEE Access 9, 57792–57807 (2021)

    work page 2021

Show all 56 references
  1. [9]

    Basu, S. et al. Cloud computing security challenges & solutions-A survey. in 2018 IEEE 8th Annual Computing and Communication Workshop and Conference (CCWC) 347–356 (2018). doi:10.1109/CCWC.2018.8301700

    2018

  2. [10]

    Ale, L., Zhang, N., King, S. A. & Chen, D. Empowering generative AI through mobile edge computing. Nat Rev Electr Eng 1, 478–486 (2024)

    2024

  3. [11]

    & Buyya, R

    Varghese, B. & Buyya, R. Next generation c loud computing: New trends and research directions. Future Generation Computer Systems 79, 849–861 (2018)

    2018

  4. [12]

    & Sun, J

    Ren, S., He, K., Girshick, R. & Sun, J. Faster R -CNN: Towards Real-Time Object Detection with Region Proposal Networks. IEEE Trans. Pattern Anal. Mach. Intell. 39, 1137–1149 (2017)

    2017

  5. [13]

    & Lin, C

    An, J., Ding, W. & Lin, C. ChatGPT: tackle the growing carbon footprint of generative AI. Nature 615, 586–586 (2023)

    2023

  6. [14]

    BBC News https://www.bbc.com/news/articles/cj5ll89dy2mo (2024)

    Electricity grids creak as AI demands soar. BBC News https://www.bbc.com/news/articles/cj5ll89dy2mo (2024)

    2024

  7. [15]

    & Zhao, W

    Kang, W., Huang, Y ., Zhang, X., Zhou, Y . & Zhao, W. Skyrmion-Electronics: An Overview and Outlook. Proceedings of the IEEE 104, 2040–2061 (2016)

    2016

  8. [16]

    Waldrop, M. M. The chips are down for Moore’s law. Nature News 530, 144 (2016)

    2016

  9. [17]

    Moore’s Law Forever? Science 299, 210–211 (2003)

    Lundstrom, M. Moore’s Law Forever? Science 299, 210–211 (2003)

    2003

  10. [18]

    & van der Wiel, W

    Jaeger, H., Noheda, B. & van der Wiel, W. G. Toward a formal theory for computing machines made out of whatever physics offers. Nat Commun 14, 4911 (2023)

    2023

  11. [19]

    Sun, A. et al. Edge-guided inverse design of digital metamaterial- based mode multiplexers for high -capacity multi- dimensional optical interconnect. Nat Commun 16, 2372 (2025)

    2025

  12. [20]

    Shen, Y . et al. Deep learning with coherent nanophotonic circuits. Nature Photon 11, 441–446 (2017)

    2017

  13. [21]

    Ashtiani, F., Geers, A. J. & Aflatouni, F. An on-chip photonic deep neural network for image classification. Nature 606, 501–506 (2022)

    2022

  14. [22]

    Awaysheh, F. M. et al. Security by Design for Big Data Frameworks Over Cloud Computing. IEEE Transactions on Engineering Management 69, 3676–3693 (2022)

    2022

  15. [23]

    Huang, C. et al. A silicon photonic–electronic neural network for fibre nonlinearity compensation. Nat Electron 4, 837–844 (2021)

    2021

  16. [24]

    Shu, H. et al. Microcomb-driven silicon photonic systems. Nature 605, 457–463 (2022)

    2022

  17. [25]

    Zhou, H. et al. Photonic matrix multiplication lights up photonic accelerator and beyond. Light Sci Appl 11, 30 (2022)

    2022

  18. [26]

    Chen, Y . et al. All-analog photoelectronic chip for high-speed vision tasks. Nature (2023) doi:10.1038/s41586-023-06558-8

    2023 doi

  19. [27]

    11 TOPS photonic convolutional accelerator for optical neural networks

    Xu, X. 11 TOPS photonic convolutional accelerator for optical neural networks. 14

  20. [28]

    Xu, Z. et al. Large-scale photonic chiplet Taichi empowers 160-TOPS/W artificial general intelligence. Science 384, 202–209 (2024)

    2024

  21. [29]

    Zhang, W. et al. A system-on-chip microwave photonic processor solves dynamic RF interference in real time with picosecond latency. Light Sci Appl 13, 14 (2024)

    2024

  22. [30]

    & Zhang, J

    Xing, S., Li, G., Li, Z., Chi, N. & Zhang, J. Study of Efficient Photonic Chromatic Dispersion Equalization Using MZI-Based Coherent Optical Matrix Multiplication. (2022)

    2022

  23. [31]

    Sludds, A. et al. Delocalized photonic deep learning on the internet’s edge. Science 378, 270–276 (2022)

    2022

  24. [32]

    Filipovich, M. J. et al. Silicon photonic architecture for training deep neural networks with direct feedback alignment. Optica 9, 1323 (2022)

    2022

  25. [33]

    Bai, B. et al. Microcomb-based integrated photonic processing unit. Nat Commun 14, 66 (2023)

    2023

  26. [34]

    Brückerhoff-Plückelmann, F. et al. Event-driven adaptive optical neural network. Sci. Adv. 9, eadi9127 (2023)

    2023

  27. [35]

    Chen, M. et al. I/O-efficient iterative matrix inversion with photonic integrated circuits. Nat Commun 15, 5926 (2024)

    2024

  28. [36]

    Feng, H. et al. Integrated lithium niobate microwave photonic processing engine. Nature (2024) doi:10.1038/s41586-024-07078-9

    2024 doi

  29. [37]

    Wang, Y . et al. Asymmetrical estimator for training encapsulated deep photonic neural networks. Nat Commun 16, 2143 (2025)

    2025

  30. [38]

    Zhou, T. et al. Large-scale neuromorphic optoelectronic computing with a reconfigurable diffractive processing unit. Nat. Photonics 15, 367–373 (2021)

    2021

  31. [39]

    Fu, T. et al. Photonic machine learning with on-chip diffractive optics. Nat Commun 14, 70 (2023)

    2023

  32. [40]

    Senior, J. M. & Jamro, M. Y . Optical Fiber Communications: Principles and Practice. (Pearson Education, 2009)

    2009

  33. [41]

    Essiambre, R. -J. & Tkach, R. W. Capacity Trends and Limits of Optical Communication Networks. Proceedings of the IEEE 100, 1035–1055 (2012)

    2012

  34. [42]

    optical_fiber_communication-libre.pdf

  35. [43]

    Zhang, S. et al. Redundancy-free integrated optical convolver for optical neural networks based on arrayed waveguide grating. Nanophotonics 13, 19–28 (2024)

    2024

  36. [44]

    Cheng, J. et al. Direct Optical Convolution Computing Based on Arrayed Waveguide Grating Router. Laser & Photonics Reviews 2301221 (2024) doi:10.1002/lpor.202301221

    2024 doi

  37. [45]

    Pappas, C. et al. A 160 TOPS Multi-dimensional AWGR-based accelerator for Deep Learning. in Optical Fiber Communication Conference (OFC) 2024 Th4C.3 (Optica Publishing Group, San Diego California, 2024). doi:10.1364/OFC.2024.Th4C.3

    2024 doi

  38. [46]

    Proietti, R. et al. Scalable Optical Interconnect Architecture Using AWGR-Based TONAK LION Switch With Limited Number of Wavelengths. Journal of Lightwave Technology 31, 4087–4097 (2013)

    2013

  39. [47]

    & Van Noorden, R

    Stokel-Walker, C. & Van Noorden, R. What ChatGPT and generative AI mean for science. Nature 614, 214–216 (2023)

    2023

  40. [48]

    https://www.science.org/doi/10.1126/sciadv.adi5339 doi:10.1126/sciadv.adi5339

    Slow-light silicon modulator with 110 -GHz bandwidth. https://www.science.org/doi/10.1126/sciadv.adi5339 doi:10.1126/sciadv.adi5339

    doi

  41. [49]

    Shi, Y . et al. Avalanche photodiode with ultrahigh gain–bandwidth product of 1,033 GHz. Nat. Photon. 18, 610–616 (2024)

    2024

  42. [50]

    Since the light sources are positioned at the user end, OPUs in the optical computing center consume even less power

    By contrast, considering only the power consumption of the computing units, an AWGR-based OPU exhibits a power efficiency of just 118.6 mW/TOPS, which is significantly lower than that of electronic chips. Since the light sources are positioned at the user end, OPUs in the opti...

  43. [51]

    NVIDIA https://www.nvidia.com/en-us/data - center/h200/

    NVIDIA H200 Tensor Core GPU. NVIDIA https://www.nvidia.com/en-us/data - center/h200/. Method Comb generation and control In this experiment, a Continuous Wave (CW) wavelength- tunable laser (TSP-400- E0018) with an output power of 16 dBm is utilized to generate the optical com...

  44. [52]

    Zhang, Y . et al. 240 Gb/s optical transmission based on an ultrafast silicon microring modulator. Photon. Res. 10, 1127 (2022)

    2022

  45. [53]

    Hu, F. et al. Beyond 300Gbps Silicon Microring Modulator with AI Acceleration. 15

  46. [54]

    & Efros, A

    Isola, P., Zhu, J.- Y ., Zhou, T. & Efros, A. A. Image -To-Image Translation With Conditional Adversarial Networks

  47. [55]

    & Efros, A

    Zhu, J.-Y ., Park, T., Isola, P. & Efros, A. A. Unpaired Image-to-Image Translation Using Cycle -Consistent Adversarial Networks. in 2017 IEEE International Conference on Computer Vision (ICCV) 2242–2251 (IEEE, Venice, 2017). doi:10.1109/ICCV .2017.244

    2017 doi

  48. [56]

    Zanetto, F. et al. WDM-Based Silicon Photonic M ulti-Socket Interconnect Architecture With Automated Wavelength and Thermal Drift Compensation. J. Lightwave Technol. 38, 6000–6006 (2020). Data Availability The authors declare that the main data supporting the findings of this ...

    2020

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.