Pith. sign in

REVIEW 3 major objections 3 minor 63 references

Hardware-Efficient Photonic Tensor Core: Accelerating Deep Neural Networks with Structured Compression

T0 review · 3 major / 3 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read A fabricated photonic tensor core that enforces block-circulant weight compression through its circuit topology runs image classifiers with up to 74.91% fewer trainable parameters and accuracy within 1.41–3.65% of full-precision digital…

desk verdict A genuine fabricated photonic tensor core with useful experimental results, but the headline efficiency claims rest on a spectral-folding scheme that appears to conflate wavelength periodicity with independent weighting. read the letter →

arxiv 2502.01670 v2 pith:62E7NPW5 submitted 2025-02-01 cs.AR cs.ETcs.LG

Shupeng Ning , Hanqing Zhu , Chenghao Feng , Jiaqi Gu , David Z. Pan , Ray T. Chen This is my paper · ORCID
classification cs.ARcs.ETcs.LG
keywords opticalneuralnetworkphotonictensorcoreblock-circulantmatricesstructuredcompressionhardware-awaretrainingmicroringresonatorwavelengthdivisionmultiplexingpowerefficiency
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

This paper aims to show that a neural network's weight matrices can be compressed into block-circulant form and that this structure can be enforced directly by the physical wiring of a photonic chip, rather than by software alone. The authors fabricate an order-4 CirPTC and use it to run convolutional image-processing and classification tasks on SVHN, CIFAR-10, and COVID-QU-Ex. They report up to 74.91% reduction in trainable parameters, active modulators, and memory with accuracy drops of only 1.41–3.65% compared to full-precision digital GEMM-based DNNs. They also project that a scaled and spectrally folded version would reach 17.13 TOPS/W, a 3.56x power-efficiency gain over uncompressed microring-based optical neural networks. The broader point is that structured compression and photonic hardware can be co-designed so that the circuit topology itself is the compression mechanism.

What carries the argument

The central object is the block-circulant matrix (BCM), a matrix made of P x Q square circulant blocks, where each block is generated by cyclically shifting one primary vector, so independent parameters drop from MN to MN/l. The CirPTC is an l-order crossbar of add-drop microring resonators arranged so that each ring's wavelength assignment follows the circulant pattern; input vectors are encoded by broadband Mach-Zehnder modulators and wavelengths are summed by photodetectors on the column bus. The differentiable PIC estimator (DPE) is a surrogate model fitted to measured chip responses, making the non-ideal analog forward pass differentiable for backpropagation. Together these pieces let structured compression, WDM-based multiply-accumulate, and hardware-aware training reinforce one another.

What would settle it

Build the scaled 48x48 CirPTC with spectral folding r=4 and high-speed modulators, then measure end-to-end TOPS/W and weight resolution; if the folded free-spectral-range crosstalk pushes the effective weight resolution below 6 bits, or the modulator energy exceeds 0.35 pJ per symbol, the projected 17.13 TOPS/W and 3.56x improvement over uncompressed MRR ONNs will not materialize.

Watch

Extended reading notes

Core claim

The central claim is that block-circulant structured compression, previously used to shrink DNNs in digital electronics, can be embodied in photonic hardware with no extra computation: an M x N weight matrix is partitioned into l x l circulant blocks, each fully determined by one length-l primary vector, and the CirPTC crossbar maps those vectors to physical wavelengths and ring-resonator switches arranged in the same block-circulant pattern. Because compression is topological, the chip needs only M x N/l active modulators and static switches, and the authors' hardware-aware training with a differentiable PIC estimator compensates for crosstalk, thermal drift, truncation, and noise. Experimental classification on three datasets matches or approaches full-precision GEMM digital baselines, and the benchmark analysis predicts 3.56x power-efficiency improvement after scaling to 48 x 48 with r=4 spectral folding using high-speed modulators.

Load-bearing premise

The projected 17.13 TOPS/W and 3.56x gain assume the thermo-optic prototype can be swapped for 10 GHz carrier-depletion or MOSCAP modulators with the assumed energy and area, and that spectral folding with r=4 keeps the ring resonators sharp enough (quality factor near 2.49e5) for 6-bit weight resolution; the measured chip itself runs at tens of kilohertz.

Editorial extensions

If this is right

  • If the accuracy results hold, block-circulant photonic tensor cores can replace GEMM-style ONNs with a compact crossbar that uses a fraction of the modulators and DACs.
  • The 74.91% parameter reduction should carry over to larger DNNs with FC, convolutional, recurrent, and attention layers, since their core computation can be reformulated as matrix-vector products.
  • The DPE training framework gives a path to deploy other analog photonic chips without iterative on-chip calibration, because it embeds measured nonidealities into the training loop.
  • With high-speed carrier-depletion or MOSCAP modulators and spectral folding, the projected 17.13 TOPS/W would make the CirPTC competitive with electronic accelerators on power efficiency.

Reading between the lines

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

  • The authors leave implicit that the same topology-level compression could reduce the DAC/ADC bottleneck, not just the compute array, because fewer active weight channels mean fewer high-speed electrical interfaces to drive.
  • The one-shot calibration property suggests that scaling to larger crossbars could be easier than for mesh-based ONNs, but this needs validation on a chip with more than 4 wavelengths before it can be relied on.
  • A testable extension would be to apply the CirPTC to a transformer block by block-circulant-compressing the Q/K/V and output projection matrices; the paper's argument implies this should work, but no experiment here shows it.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 3 minor

Summary. The paper presents a block-circulant photonic tensor core (CirPTC) that enforces structured compression directly through the circuit topology of an MRR crossbar, with serial MRR weight banks and MZM input encoding. The authors fabricate an order-4 CirPTC and report on-chip convolution with normalized RMSE 0.0243, end-to-end classification accuracies of 80.04% on CIFAR-10, 88.08% on SVHN, and 92.6% on COVID-QU-Ex, and up to a 74.91% reduction in trainable parameters against full GEMM baselines. They also introduce a hardware-aware training framework based on a differentiable PIC estimator (DPE) calibrated to measured device behavior. The Discussion projects that a scaled, spectrally folded 48x48 CirPTC would reach 5.48 TOPS/mm2 and 17.13 TOPS/W, a 3.56x power-efficiency improvement over uncompressed MRR-based ONNs.

Significance. The experimental core is valuable: the fabricated chip demonstrates that block-circulant structure can be imposed by photonic topology, that one-shot cascaded calibration is feasible, and that hardware-aware training recovers accuracy close to full-precision digital GEMM baselines. The reported normalized RMSE, the three classification results, and the DPE-based transfer of measured nonidealities into training are concrete strengths that stand independently of the projected efficiency numbers. If the experimental claims are accepted, the work provides a plausible path toward compact photonic tensor cores for structured-compressed DNNs. However, the headline efficiency claims are not established: the spectral-folding mechanism has an internal consistency problem, and the main text contains inconsistent computation-density figures. The paper is publishable in revised form, but the projected 3.56x improvement should not be presented as a supported result in its current form.

major comments (3)
  1. [Discussion, 'Spectral folding'] The r-fold spectral-folding proposal is internally inconsistent with the FSR-periodicity of the MRRs. An add-drop MRR's amplitude transmission is periodic in frequency with period equal to the FSR; for wavelengths separated by integer multiples of the FSR the detuning from resonance is identical, so a single MRR cannot apply r independent weights to r folded wavelength channels. Since the serial weight MRRs used for encoding are likewise periodic filters, folding the input onto the same crossbar at r FSR-spaced wavelengths replicates the same weight pattern for all folds rather than creating r·N independent columns. Thus the claimed M×(r·N) BCM operation is not realized by the described hardware, and the resulting throughput, 5.48 TOPS/mm2 and 17.13 TOPS/W (3.56x) estimates are unsupported unless a non-periodic spectral-shaping component or per-FSR weight bank is specified.
  2. [Introduction vs. Discussion, 'Benchmark analysis'] The computing-density figures for what appears to be the same 48×48 CirPTC are inconsistent: the Introduction reports 5.84 TOPS/mm2, while the Discussion reports 4.85 TOPS/mm2 for a 48×48 CirPTC at 10 GHz and 5.48 TOPS/mm2 for the r=4 folded configuration. The paper should reconcile these numbers or state the exact area and throughput assumptions behind each; as it stands, the efficiency claims cannot be independently verified.
  3. [Discussion, 'Benchmark analysis'] The 10 GHz projection replaces the experimentally demonstrated thermo-optic MZMs with carrier-depletion or MOSCAP MZMs and assumes 0.35 pJ per symbol, 3 mW per MRR static power, and ADC/TIA energies from external references; it also relies on a required MRR Q of 2.49×10^5 for 6-bit weight resolution at M=48. These are plausible scaling assumptions, but none are experimentally characterized here, and the paper should explicitly mark the 17.13 TOPS/W and 3.56x figures as projections contingent on these assumptions, with sensitivity analysis (e.g., to Q, modulator energy, and static power) rather than presenting them as expected performance.
minor comments (3)
  1. [Operation Principle, Eq. (1)] The matrix dimensions M×N, N×M, and the block counts P×Q are used inconsistently; for example, the text says 'M×N BCM' while the CirPTC is described as an N×M crossbar. Please standardize the indexing.
  2. [Operation Mechanism of CirPTC] It is not immediately clear why an order-4 4×4 BCM requires 16 crossbar MRRs in addition to the serial weight MRRs; a sentence explaining that the crossbar contains one static MRR per BCM element rather than per independent parameter would help reconcile the 16-MRR array with the 74.91% active-modulator reduction claim.
  3. [Results, 'CirPTC-based ONN for classification'] The sentence stating that crosstalk and noise accumulate along the forward path of the network is qualitative; please add a pointer to the supplementary figure or data that quantifies this depth-dependent degradation.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the experimental accuracy and parameter-reduction claims rest on measured chip data and external block-circulant theory, while the efficiency projections are explicit extrapolations from external device parameters rather than fitted or self-referential outcomes.

full rationale

The paper's central experimental results are self-contained empirical outcomes: on-chip convolution with 0.0243 normalized RMSE, measured classification accuracies of 92.6% on COVID-QU-Ex, 80.04% on CIFAR-10, and 88.08% on SVHN, and a sub-1% accuracy drop attributable to DPE-based hardware-aware training. These accuracy figures come from chip measurements, not from the DPE surrogate; the surrogate is calibrated to measured device transmission curves and is used to make offline training hardware-aware, so it is not fitted to the reported accuracies. The 74.91% parameter reduction is a direct mathematical property of block-circulant matrices (MN/l independent parameters) derived from external prior work (CircNN and low-displacement-rank theory), not a fitted input renamed as a prediction. Self-citations [25,36,50] supply review context, prior device demonstrations, and comparative architecture discussion, and they are accompanied by external references such as [45,47,53,54]; none is load-bearing for the accuracy or efficiency claims. The projected 17.13 TOPS/W and 3.56x improvement follow arithmetically from explicit modeling assumptions and externally sourced device parameters (e.g., 0.35 pJ/symbol MOSCAP MZM, 39 mW ADC), and the paper states these are expected after appropriate scaling and spectral folding, not measured results. Whether spectral folding can in fact provide r independent weights at FSR-spaced wavelengths is a physics/engineering validity question, not a circularity in the derivation chain. No load-bearing step reduces by construction to its own inputs, so the circularity score is 0.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

The ledger reflects that the experimental core rests on standard im2col and the domain assumption that block-circulant compression preserves accuracy. The efficiency projection rests on extra assumptions: high-speed modulators, spectral-folding behavior, and power constants from references. No new physical entities such as particles, forces, or dimensions are introduced. CirPTC, DPE, and spectral folding are new designs or models rather than postulated entities.

free parameters (3)
  • Block size l (order of CirPTC) = 4
    Selected as the optimal compromise between pre-training accuracy and chip footprint in the 'Design of CirPTC' section. The compression ratio, parameter reduction, and hardware savings all depend on this hyperparameter.
  • DPE surrogate model parameters = Not reported
    The differentiable PIC estimator is fitted to a lookup table of measured vector combinations in the 'Training Framework' section, so the DPE contains parameters fit to on-chip data. The fit is calibration rather than a physical free parameter, but it influences the reported robustness gains.
  • Projected device power constants (0.35 pJ/symbol MZM, 3 mW/MRR, ADC 39/194 mW, TIA 0.65 pJ/bit) = Taken from cited references
    The 17.13 TOPS/W and 3.56x projections depend on these assumed component power figures from references rather than on measurements from the fabricated CirPTC.
assumptions (6)
  • domain assumption DNN weight matrices can be constrained to block-circulant form with acceptable accuracy loss.
    Invoked throughout 'Operation Principle' and 'ONN with Structured Compression'; supported by citations to CircNN [19] and low-displacement-rank theory [37], but not re-derived in this paper.
  • standard math Convolution can be exactly converted to matrix-vector products via im2col, and BCM compression applies to the resulting weight matrix.
    Used in 'ONN with Structured Compression' and Fig. 1(a); a standard transformation cited to [38,39].
  • domain assumption The fitted transmission curves of MZMs and MRRs accurately predict on-chip behavior during inference.
    The paper fits device models to measurements in Fig. 2 and uses these fits to build the DPE and benchmark model. If the fits do not generalize, the hardware-aware training benefit is reduced.
  • ad hoc to paper Projected high-speed operation can be achieved by replacing thermo-optic modulators with carrier-depletion or MOSCAP MZMs while retaining the same crossbar topology and error behavior.
    Introduced in 'Benchmark analysis' and 'Spectral folding' to compute 17.13 TOPS/W; no experimental evidence on the fabricated chip supports this replacement.
  • ad hoc to paper Spectral folding with r=4 preserves 6-bit weight resolution without additional crosstalk beyond modeled overlap.
    The efficiency improvement depends on this scaling approach in the 'Spectral folding' section, which is analyzed numerically but not experimentally demonstrated.
  • domain assumption Weights and inputs can be treated as nonnegative either by positive normalization or by splitting matrices into positive and negative parts without changing trained model accuracy.
    The CirPTC implements amplitude-only multiplication; the paper describes two methods for full-range weights, and the classification experiments rely on this being trainable.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hardware-Efficient Photonic Tensor Core: Accelerating Deep Neural Networks with Structured Compression." pith.science (2026). https://pith.science/paper/62E7NPW5

@misc{pith2026250201670,
  author       = {Pith},
  title        = {Pith review of: Hardware-Efficient Photonic Tensor Core: Accelerating Deep Neural Networks with Structured Compression},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/62E7NPW5}},
  note         = {Machine review of arXiv:2502.01670}
}
read the original abstract

The rapid growth in computing demands, particularly driven by artificial intelligence applications, has begun to exceed the capabilities of traditional electronic hardware. Optical computing offers a promising alternative due to its parallelism, high computational speed, and low power consumption. However, existing photonic integrated circuits are constrained by large footprints, costly electro-optical interfaces, and complex control mechanisms, limiting the practical scalability of optical neural networks (ONNs). To address these limitations, we introduce a block-circulant photonic tensor core for a structure-compressed optical neural network (StrC-ONN) architecture. The structured compression technique substantially reduces both model complexity and hardware resources without sacrificing the versatility of neural networks, and achieves accuracy comparable to uncompressed models. Additionally, we propose a hardware-aware training framework to compensate for on-chip nonidealities to improve model robustness and accuracy. Experimental validation through image processing and classification tasks demonstrates that our StrC-ONN achieves a reduction in trainable parameters of up to 74.91%,while still maintaining competitive accuracy levels. Performance analyses further indicate that this hardware-software co-design approach is expected to yield a 3.56 times improvement in power efficiency. By reducing both hardware requirements and control complexity across multiple dimensions, this work explores a new pathway toward practical and scalable ONNs, highlighting a promising route to address future computational efficiency challenges.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 54 canonical work pages

  1. [1]

    Deep learning,

    Y. LeCun, Y. Bengio, and G. Hinton, "Deep learning," Nature 521, 436-444 (2015)

    work page 2015

  2. [2]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren et al., "Deep residual learning for image recognition," in Proceedings of the IEEE conference on computer vision and pattern recognition(2016), pp. 770-778

    work page 2016

  3. [3]

    You only look once: Unified, real -time object detection,

    J. Redmon, "You only look once: Unified, real -time object detection," in Proceedings of the IEEE conference on computer vision and pattern recognition(2016)

    work page 2016

  4. [4]

    Imagenet classification with deep convolutional neural networks,

    A. Krizhevsky, I. Sutskever , and G. E. Hinton, "Imagenet classification with deep convolutional neural networks," Advances in neural information processing systems 25 (2012). 18

    work page 2012

  5. [5]

    End to end learning for self-driving cars,

    M. Bojarski, "End to end learning for self-driving cars," arXiv preprint arXiv:1604.07316 (2016)

    arXiv 2016

  6. [6]

    Sequence to Sequence Learning with Neural Networks,

    I. Sutskever, "Sequence to Sequence Learning with Neural Networks," arXiv preprint arXiv:1409.3215 (2014)

    arXiv 2014

  7. [7]

    Bert: Pre -training of deep bidirectional transformers for language understanding,

    J. Devlin, "Bert: Pre -training of deep bidirectional transformers for language understanding," arXiv preprint arXiv:1810.04805 (2018)

    arXiv 2018

  8. [8]

    Unlike the Von Neumann architecture, PTCs can execute an MVM operation within a single clock cycle. Defining an operation as either the multiplication or addition, the throughput of an N × M CirPTC can be evaluated by the number of operations per second (OPS), which is quantified as: OPS 2 opMN f (2) where fop represents the operational rate . Clearly, en...

    work page

Show all 63 references
  1. [9]

    Cardiologist -level arrhythmia detection and classification in ambulatory electrocardiograms using a deep neural network,

    A. Y. Hannun, P. Rajpurkar, M. Haghpanahi et al., "Cardiologist -level arrhythmia detection and classification in ambulatory electrocardiograms using a deep neural network," Nat. Med. 25, 65-69 (2019)

    2019

  2. [10]

    Large -scale pancreatic cancer detection via non -contrast CT and deep learning,

    K. Cao, Y. Xia, J. Yao et al., "Large -scale pancreatic cancer detection via non -contrast CT and deep learning," Nat. Med. 29, 3033-3043 (2023)

    2023

  3. [11]

    Bart: Denoising sequence -to-sequence pre-training for natural language generation, translation, and comprehension,

    M. Lewis, "Bart: Denoising sequence -to-sequence pre-training for natural language generation, translation, and comprehension," arXiv preprint arXiv:1910.13461 (2019)

    2019 arXiv

  4. [12]

    Language models are few -shot learners,

    T. Brown, B. Mann, N. Ryder et al., "Language models are few -shot learners," Advances in neural information processing systems 33, 1877-1901 (2020)

    2020

  5. [13]

    Harnessing the power of llms in practice: A survey on chatgpt and beyond,

    J. Yang, H. Jin, R. Tang et al., "Harnessing the power of llms in practice: A survey on chatgpt and beyond," ACM Transactions on Knowledge Discovery from Data 18, 1-32 (2024)

    2024

  6. [14]

    Optimizing FPGA-based accelerator design for deep convolutional neural networks,

    C. Zhang, P. Li, G. Sun et al., "Optimizing FPGA-based accelerator design for deep convolutional neural networks," in Proceedings of the 2015 ACM/SIGDA international symposium on field-programmable gate arrays(2015), pp. 161- 170

    2015

  7. [15]

    A high performance FPGA -based accelerator for large -scale convolutional neural networks,

    H. Li, X. Fan, L. Jiao et al., "A high performance FPGA -based accelerator for large -scale convolutional neural networks," in 2016 26th International Conference on Field Programmable Logic and Applications (FPL) (IEEE2016), pp. 1-9

    2016

  8. [16]

    In -datacenter performance analysis of a tensor processing unit,

    N. P. Jouppi, C. Young, N. Patil et al., "In -datacenter performance analysis of a tensor processing unit," in Proceedings of the 44th annual international symposium on computer architecture(2017), pp. 1-12

    2017

  9. [17]

    More than moore,

    M. M. Waldrop, "More than moore," Nature 530, 144-148 (2016)

    2016

  10. [18]

    Towards atomic and close-to-atomic scale manufacturing,

    F. Fang, N. Zhang, D. Guo et al., "Towards atomic and close-to-atomic scale manufacturing," Int. J. Extreme Manuf. 1, 012001 (2019)

    2019

  11. [19]

    Reinventing high performance computing: challenges and opportunities,

    D. Reed, D. Gannon, and J. Dongarra, "Reinventing high performance computing: challenges and opportunities," arXiv preprint arXiv:2203.02544 (2022)

    2022 arXiv

  12. [20]

    Circnn: accelerating and compressing deep neural networks using block -circulant weight matrices,

    C. Ding, S. Liao, Y. Wang et al., "Circnn: accelerating and compressing deep neural networks using block -circulant weight matrices," in Proceedings of the 50th Annual IEEE/ACM International Symposium on Microarchitecture(2017), pp. 395-408

    2017

  13. [21]

    Equivalent-accuracy accelerated neural-network training using analogue memory,

    S. Ambrogio, P. Narayanan, H. Tsai et al., "Equivalent-accuracy accelerated neural-network training using analogue memory," Nature 558, 60-67 (2018)

    2018

  14. [22]

    Analogue signal and image processing with large memristor crossbars,

    C. Li, M. Hu, Y. Li et al., "Analogue signal and image processing with large memristor crossbars," Nat. Electron. 1, 52-59 (2018)

    2018

  15. [23]

    Deep learning with coherent nanophotonic circuits,

    Y. Shen, N. C. Harris, S. Skirlo et al., "Deep learning with coherent nanophotonic circuits," Nat . photonics 11, 441- 446 (2017)

    2017

  16. [24]

    Neuromorphic photonic networks using silicon photonic weight banks,

    A. N. Tait, T. F. De Lima, E. Zhou et al., "Neuromorphic photonic networks using silicon photonic weight banks," Sci. Rep. 7, 7430 (2017)

    2017

  17. [25]

    Parallel convolutional processing using an integrated photonic tensor core,

    J. Feldmann, N. Youngblood, M. Karpov et al., "Parallel convolutional processing using an integrated photonic tensor core," Nature 589, 52-58 (2021)

    2021

  18. [26]

    Photonic -electronic integrated circuits for high -performance computing and ai 19 accelerators,

    S. Ning, H. Zhu, C. Feng et al., "Photonic -electronic integrated circuits for high -performance computing and ai 19 accelerators," J. Lightwave Technol. (2024)

    2024

  19. [27]

    11 TOPS photonic convolutional accelerator for optical neural networks,

    X. Xu, M. Tan, B. Corcoran et al., "11 TOPS photonic convolutional accelerator for optical neural networks," Nature 589, 44-51 (2021)

    2021

  20. [28]

    Large -scale photonic chiplet Taichi empowers 160 -TOPS/W artificial general intelligence,

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

    2024

  21. [29]

    All -optical spiking neurosynaptic networks with self -learning capabilities,

    J. Feldmann, N. Youngblood, C. D. Wright et al., "All -optical spiking neurosynaptic networks with self -learning capabilities," Nature 569, 208-214 (2019)

    2019

  22. [30]

    Compressing the CNN architecture for in -air handwritten Chinese character recognition,

    J. Gan, W. Wang, and K. Lu, "Compressing the CNN architecture for in -air handwritten Chinese character recognition," Pattern Recognit. Lett. 129, 190-197 (2020)

    2020

  23. [31]

    Deep compression: Compressing deep neural networks with pruning, trained quantization and huffman coding,

    S. Han, H. Mao, and W. J. Dally, "Deep compression: Compressing deep neural networks with pruning, trained quantization and huffman coding," arXiv preprint arXiv:1510.00149 (2015)

    2015 arXiv

  24. [32]

    Pruning convolutional neural networks for resource efficient inference,

    P. Molchanov, S. Tyree, T. Karras et al., "Pruning convolutional neural networks for resource efficient inference," arXiv preprint arXiv:1611.06440 (2016)

    2016 arXiv

  25. [33]

    Speeding up convolutional neural networks with low rank expansions,

    M. Jaderberg, A. Vedaldi, and A. Zisserman, "Speeding up convolutional neural networks with low rank expansions," arXiv preprint arXiv:1405.3866 (2014)

    2014 arXiv

  26. [34]

    Convolutional neural networks with low -rank regularization,

    C. Tai, T. Xiao, Y. Zhang et al., "Convolutional neural networks with low -rank regularization," arXiv preprint arXiv:1511.06067 (2015)

    2015 arXiv

  27. [35]

    Efficient recurrent neural networks using structured matrices in fpgas,

    Z. Li, S. Wang, C. Ding et al., "Efficient recurrent neural networks using structured matrices in fpgas," arXiv preprint arXiv:1803.07661 (2018)

    2018 arXiv

  28. [36]

    Massively parallel amplitude -only Fourier neural network,

    M. Miscuglio, Z. Hu, S. Li et al., "Massively parallel amplitude -only Fourier neural network," Optica 7, 1812-1819 (2020)

    2020

  29. [37]

    A compact butterfly-style silicon photonic–electronic neural chip for hardware-efficient deep learning,

    C. Feng, J. Gu, H. Zhu et al., "A compact butterfly-style silicon photonic–electronic neural chip for hardware-efficient deep learning," Acs Photonics 9, 3906-3916 (2022)

    2022

  30. [38]

    Theoretical properties for neural networks with weight matrices of low displacement rank,

    L. Zhao, S. Liao, Y. Wang et al., "Theoretical properties for neural networks with weight matrices of low displacement rank," in international conference on machine learning(PMLR2017), pp. 4082-4090

  31. [39]

    High performance convolutional neural networks for document processing,

    K. Chellapilla, S. Puri, and P. Simard, "High performance convolutional neural networks for document processing," in Tenth international workshop on frontiers in handwriting recognition(Suvisoft2006)

  32. [40]

    Caffe: Convolutional architecture for fast feature embedding,

    Y. Jia, E. Shelhamer, J. Donahue et al., "Caffe: Convolutional architecture for fast feature embedding," in Proceedings of the 22nd ACM international conference on Multimedia(2014), pp. 675-678

    2014

  33. [41]

    Efficient training and design of photonic neural network through neuroevolution,

    T. Zhang, J. Wang, Y. Dan et al., "Efficient training and design of photonic neural network through neuroevolution," Opt. Express 27, 37150-37163 (2019)

    2019

  34. [42]

    Deep physical neural networks trained with backpropagation,

    L. G. Wright, T. Onodera, M. M. Stein et al., "Deep physical neural networks trained with backpropagation," Nature 601, 549-555 (2022)

    2022

  35. [43]

    Hybrid training of optical neural networks,

    J. Spall, X. Guo, and A. I. Lvovsky, "Hybrid training of optical neural networks," Optica 9, 803-811 (2022)

    2022

  36. [44]

    Fully forward mode training for optical neural networks,

    Z. Xue, T. Zhou, Z. Xu et al., "Fully forward mode training for optical neural networks," Nature 632, 280-286 (2024)

    2024

  37. [45]

    Single -chip photonic deep neural network with forward -only training,

    S. Bandyopadhyay, A. Sludds, S. Krastanov et al., "Single -chip photonic deep neural network with forward -only training," Nat. Photonics 18, 1335-1343 (2024)

    2024

  38. [46]

    Experimental extraction of effective refractive index and thermo- optic coefficients of silicon-on-insulator waveguides using interferometers,

    S. Dwivedi, A. Ruocco, M. Vanslembrouck et al., "Experimental extraction of effective refractive index and thermo- optic coefficients of silicon-on-insulator waveguides using interferometers," J. Lightwave Technol. 33, 4471-4477 (2015)

    2015

  39. [47]

    Van, Optical microring resonators: theory, techniques, and applications (CRC Press, 2016)

    V. Van, Optical microring resonators: theory, techniques, and applications (CRC Press, 2016). 20

    2016

  40. [48]

    Delocalized photonic deep learning on the internet’s edge,

    A. Sludds, S. Bandyopadhyay, Z. Chen et al., "Delocalized photonic deep learning on the internet’s edge," Science 378, 270-276 (2022)

    2022

  41. [49]

    Si microring resonator crossbar array for on-chip inference and training of the optical neural network,

    S. Ohno, R. Tang, K. Toprasertpong et al., "Si microring resonator crossbar array for on-chip inference and training of the optical neural network," Acs Photonics 9, 2614-2622 (2022)

    2022

  42. [50]

    Realization of an integrated photonic platform for coherent photo - electric processing,

    N. Youngblood, S. R. Kari, N. Nobile et al., "Realization of an integrated photonic platform for coherent photo - electric processing," (2023)

    2023

  43. [51]

    SqueezeLight: Towards scalable optical neural networks with multi -operand ring resonators,

    J. Gu, C. Feng, Z. Zhao et al., "SqueezeLight: Towards scalable optical neural networks with multi -operand ring resonators," in 2021 Design, Automation & Test in Europe Conference & Exhibition (DATE)(IEEE2021), pp. 238-243

    2021

  44. [52]

    Covid-qu-ex dataset,

    A. Tahir, M. E. Chowdhury, Y. Qiblawey et al., "Covid-qu-ex dataset," Kaggle (2022)

    2022

  45. [53]

    APSUNY process design kit (PDKv3. 0): O, C and L band silicon photonics component libraries on 300mm wafers,

    E. Timurdogan, Z. Su, R.-J. Shiue et al., "APSUNY process design kit (PDKv3. 0): O, C and L band silicon photonics component libraries on 300mm wafers," in Optical Fiber Communication Conference(Optica Publishing Group2019), p. Tu2A. 1

  46. [54]

    Optimization and demonstration of a large-bandwidth carrier-depletion silicon optical modulator,

    J. Wang, C. Qiu, H. Li et al., "Optimization and demonstration of a large-bandwidth carrier-depletion silicon optical modulator," J. Lightwave Technol. 31, 4119-4125 (2013)

    2013

  47. [55]

    Heterogeneously integrated iii–v/si mos capacitor mach–zehnder modulator,

    T. Hiraki, T. Aihara, K. Hasebe et al., "Heterogeneously integrated iii–v/si mos capacitor mach–zehnder modulator," Nat. Photonics 11, 482-485 (2017)

    2017

  48. [56]

    Kihara, S

    M. Kihara, S. Ono, and P. Eskelinen, Digital clocks for synchronization and communications (Artech House, 2003)

    2003

  49. [57]

    Post -fabrication trimming of silicon photonic ring resonators at wafer - scale,

    H. Jayatilleka, H. Frish, R. Kumar et al., "Post -fabrication trimming of silicon photonic ring resonators at wafer - scale," J. Lightwave Technol. 39, 5083-5088 (2021)

    2021

  50. [58]

    https://www.alphacoreinc.com/adc -analog-to-digital-converters/ (accessed 2024-10-20)

    ADC (Analog -to-Digital converters) −Alphacore. https://www.alphacoreinc.com/adc -analog-to-digital-converters/ (accessed 2024-10-20)

    2024

  51. [59]

    A 53 -Gbit/s optical receiver frontend with 0.65 pJ/bit in 28 -nm bulk-CMOS,

    L. Szilagyi, J. Pliva, R. Henker et al., "A 53 -Gbit/s optical receiver frontend with 0.65 pJ/bit in 28 -nm bulk-CMOS," IEEE J. Solid-State Circuits 54, 845-855 (2018)

    2018

  52. [60]

    Photonic multiply-accumulate operations for neural networks,

    M. A. Nahmias, T. F. De Lima, A. N. Tait et al., "Photonic multiply-accumulate operations for neural networks," IEEE J. Sel. Top. Quantum Electron. 26, 1-18 (2019)

    2019

  53. [61]

    High-Q and high finesse silicon microring resonator,

    J. Nijem, A. Naiman, R. Zektzer et al., "High-Q and high finesse silicon microring resonator," Opt. Express 32, 7896- 7906 (2024)

    2024

  54. [62]

    High -Q silica microdisk optical resonators with large wedge angles on a silicon chip,

    G. Li, P. Liu, X. Jiang et al., "High -Q silica microdisk optical resonators with large wedge angles on a silicon chip," Photonics Res. 3, 279-282 (2015)

    2015

  55. [63]

    Dry -etched ultrahigh-Q silica microdisk resonators on a silicon chip,

    J. Gu, J. Liu, Z. Bai et al., "Dry -etched ultrahigh-Q silica microdisk resonators on a silicon chip," Photonics Res. 9, 722-725 (2021)

    2021

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.