pith. sign in

arxiv: 2510.00767 · v1 · pith:KRGCWLEJnew · submitted 2025-10-01 · ⚛️ physics.optics · physics.atom-ph

Color2Struct: efficient and accurate deep-learning inverse design of structural color with controllable inference

Sichao Shan , Han Ye , Zhengmei Yang , Junpeng Hou , Zhitong Li This is my paper

Pith reviewed 2026-05-22 12:02 UTC · model grok-4.3

classification ⚛️ physics.optics physics.atom-ph
keywords structural colorinverse designdeep learningnanophotonicsreflectance spectraphysics constraintstandem networkscontrollable inference
0
0 comments X

The pith

Color2Struct uses sampling bias correction, adaptive loss weighting, and physics-guided inference to improve structural color inverse design accuracy and controllability.

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

The paper develops Color2Struct, a deep learning framework aimed at the inverse design of structural colors by learning mappings from structure parameters to optical responses. It improves upon existing tandem networks by incorporating three key techniques: sampling bias correction to handle data imbalances, adaptive loss weighting for effective training, and physics-guided inference to ensure spectral controllability during prediction. These changes result in 65% reduction in color difference errors and 48% better short-wave near-infrared reflectivity predictions for RGB primary colors. The designs are validated by fabricating nanostructures with thin-film deposition and measuring their spectra, supporting uses in display technologies and solar thermal harvesting. The method is presented as scalable and generalizable to more complex structures.

Core claim

Color2Struct is proposed as a universal framework for efficient and accurate inverse design of structural colors with controllable predictions, leveraging sampling bias correction, adaptive loss weighting, and physics-guided inference to outperform tandem networks by 65% in color difference and 48% in short-wave near-infrared reflectivity for RGB primary colors, with experimental validation on fabricated samples.

What carries the argument

The integration of sampling bias correction, adaptive loss weighting, and physics-guided inference within the Color2Struct framework to enforce physical constraints and improve prediction accuracy.

If this is right

  • Enhanced accuracy for designing RGB primary structural colors.
  • Improved reflectivity predictions in the short-wave near-infrared region.
  • Greater controllability over the output spectra of the model.
  • Practical validation through fabrication and spectral measurements of nanostructures.
  • Applicability to high-end display technologies and solar thermal energy harvesting.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar bias correction and guided inference techniques could benefit inverse design in other areas of nanophotonics like metasurface optimization.
  • The framework may allow for fewer training samples by relying on physics at inference time.
  • Extending the approach to multi-objective designs involving multiple colors or wavelengths could be tested in future work.
  • Combining this with generative models might enable discovery of novel structure geometries beyond current limits.

Load-bearing premise

The physics-guided inference at test time enforces spectral controllability while maintaining or enhancing the accuracy improvements without introducing offsetting systematic errors.

What would settle it

Measuring the actual reflectance spectra of fabricated nanostructure samples designed by Color2Struct and comparing the observed color differences and reflectivity values against those from standard tandem networks to verify the reported percentage improvements.

read the original abstract

Deep learning (DL) has revolutionized many fields such as materials design and protein folding. Recent studies have demonstrated the advantages of DL in the inverse design of structural colors, by effectively learning the complex nonlinear relations between structure parameters and optical responses, as dictated by the physical laws of light. While several models, such as tandem neural networks and generative adversarial networks, have been proposed, these methods can be biased and are difficult to scale up to complex structures. Moreover, the difficulty in incorporating physical constraints at the inference time hinders the controllability of the model-predicted spectra. In this work, we propose Color2Struct, a universal framework for efficient and accurate inverse design of structural colors with controllable predictions. By utilizing sampling bias correction, adaptive loss weighting, and physics-guided inference, Color2Struct improves the prediction of tandem networks by 65% (color difference) and 48% (short-wave near-infrared reflectivity) in designing RGB primary colors. These improvements make Color2Struct highly promising for applications in high-end display technologies and solar thermal energy harvesting. In experiments, the nanostructure samples are fabricated using a standard thin-film deposition method and their reflectance spectra are measured to validate the designs. Our work provides an efficient and highly optimized method for controllable inverse design, benefiting future explorations of more intricate structures. The proposed framework can be further generalized to a wide range of fields beyond nanophotonics.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper proposes Color2Struct, a deep-learning framework for inverse design of structural colors that incorporates sampling bias correction, adaptive loss weighting, and physics-guided inference at test time. It claims these components yield 65% improvement in color difference and 48% improvement in short-wave near-infrared reflectivity relative to tandem networks when targeting RGB primary colors, with additional experimental validation via thin-film deposition and reflectance measurements of fabricated nanostructures. The work positions the method as efficient, controllable, and generalizable beyond nanophotonics.

Significance. If the reported gains prove robust, the framework would offer a practical advance in nanophotonics inverse design by addressing bias and controllability issues that limit prior tandem and GAN approaches. The experimental fabrication step provides a concrete link to realizable devices, strengthening relevance for display and solar-thermal applications. Credit is due for attempting to combine multiple corrective techniques and for including physical validation rather than relying solely on simulation.

major comments (3)
  1. [Methods / §3.3 (physics-guided inference)] The central attribution of the 65% color-difference and 48% SWIR-reflectivity gains to the three proposed techniques (sampling bias correction, adaptive loss weighting, and physics-guided inference) cannot be verified because the manuscript provides neither an explicit mathematical formulation of the physics-guided inference step nor an ablation that isolates its effect. Without these, it remains possible that the added constraint at inference time introduces systematic shifts that the chosen scalar metrics do not penalize, as noted in the stress-test concern.
  2. [Results / §4.1 (quantitative comparison)] The comparison to the tandem-network baseline is under-determined: the manuscript does not demonstrate that the baseline uses identical data splits, hyper-parameters, or training procedures as Color2Struct. Consequently the numerical improvements cannot be confidently assigned to the new components rather than to differences in implementation details.
  3. [Results / §4.2 and Experimental Validation] No error bars, dataset sizes, or statistical significance tests accompany the headline percentage improvements or the fabrication measurements. This absence is load-bearing for the claim of reliable superiority and leaves open the possibility that the reported gains fall within experimental or training variability.
minor comments (2)
  1. [Methods] Notation for the adaptive loss weighting coefficients is introduced without a clear reference to the preceding equation that defines the base loss, making the weighting scheme harder to reproduce.
  2. [Figures] Figure captions for the fabricated-sample reflectance plots should explicitly state the number of measured devices and the wavelength range used for the SWIR metric.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. We address each major comment below and have revised the manuscript to incorporate clarifications, additional formulations, ablations, and statistical details as suggested.

read point-by-point responses

  1. Referee: [Methods / §3.3 (physics-guided inference)] The central attribution of the 65% color-difference and 48% SWIR-reflectivity gains to the three proposed techniques (sampling bias correction, adaptive loss weighting, and physics-guided inference) cannot be verified because the manuscript provides neither an explicit mathematical formulation of the physics-guided inference step nor an ablation that isolates its effect. Without these, it remains possible that the added constraint at inference time introduces systematic shifts that the chosen scalar metrics do not penalize, as noted in the stress-test concern.

    Authors: We agree that an explicit formulation and ablation study are needed to rigorously attribute the gains. In the revised manuscript we add the full mathematical description of the physics-guided inference procedure in §3.3 and report a new ablation table that isolates the incremental contribution of each component (including physics-guided inference) to the color-difference and SWIR-reflectivity metrics. This directly addresses the possibility of unpenalized systematic shifts. revision: yes

  2. Referee: [Results / §4.1 (quantitative comparison)] The comparison to the tandem-network baseline is under-determined: the manuscript does not demonstrate that the baseline uses identical data splits, hyper-parameters, or training procedures as Color2Struct. Consequently the numerical improvements cannot be confidently assigned to the new components rather than to differences in implementation details.

    Authors: We will revise §4.1 to explicitly document that the tandem-network baseline was retrained from scratch using exactly the same data splits, hyper-parameter search protocol, optimizer settings, and early-stopping criteria as Color2Struct. A supplementary table will list the shared configuration values, confirming that the reported 65 % and 48 % gains arise from the three proposed techniques rather than implementation discrepancies. revision: yes

  3. Referee: [Results / §4.2 and Experimental Validation] No error bars, dataset sizes, or statistical significance tests accompany the headline percentage improvements or the fabrication measurements. This absence is load-bearing for the claim of reliable superiority and leaves open the possibility that the reported gains fall within experimental or training variability.

    Authors: We acknowledge the need for statistical rigor. The revised manuscript will state the exact training and test set sizes, add error bars (standard deviation over five independent runs) to all quantitative metrics, and include two-sided t-test p-values for the headline improvements. For the fabricated samples we will report measurement uncertainty from repeated reflectance scans and note the number of devices measured. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical gains from added training and inference procedures

full rationale

The paper introduces Color2Struct as a DL framework that augments tandem networks via three explicit techniques—sampling bias correction, adaptive loss weighting, and physics-guided inference—and reports measured improvements (65% color difference, 48% SWIR reflectivity) on RGB primary-color designs. These gains are presented as outcomes of experimental validation that includes fabricated samples and measured spectra. No equations, derivations, or self-citations appear in the provided text that would reduce the reported metrics to quantities defined by the same fitted parameters or to a prior result by the same authors. The central claims therefore remain self-contained empirical statements rather than tautological reductions.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The abstract does not enumerate explicit free parameters or new entities; the central claim rests on the standard deep-learning assumption that a sufficiently expressive network can approximate the nonlinear structure-to-spectrum mapping when trained with the listed corrections.

axioms (1)
  • domain assumption The relationship between nanostructure parameters and optical responses is governed by physical laws of light that neural networks can learn from data.
    Invoked in the opening paragraph as the justification for using deep learning in inverse design.

pith-pipeline@v0.9.0 · 5793 in / 1351 out tokens · 62005 ms · 2026-05-22T12:02:49.159530+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

97 extracted references · 97 canonical work pages

  1. [1]

    overfitted

    Since colors are governed by nanostructures and the material properties rather than chemical dyes, structural colors offer greater durability, tunability, and stability under varying environmental conditions. For the multilayer structures, when incident light propagates in the layers, electromagnetic waves at different wavelengths undergo constructive or ...

    work page 1931

  2. [2]

    Best Result

    By applying ALW alone to FNN, we observe reductions of 4% in ∆𝐸𝑎𝑣𝑔 and ∆𝐸𝑚𝑎𝑥 on the test dataset. For the INN, the two metrics are reduced by 8% and 4%, respectively. Taking a step f urther, we combine the two treatments in Color2Struct and observe significant performance improvements, indicating a strong synergy between the two treatments. Also in Fig. 4...

    work page

  3. [3]

    & Shi, S

    Liu, Y., Zhao, T., Ju, W. & Shi, S. Materials discovery and design using machine learning. Journal of Materiomics 3, 159–177 (2017)

    work page 2017

  4. [4]

    Liu, R. et al. A predictive machine learning approach for microstructure optimization and materials design. Sci Rep 5, 11551 (2015)

    work page 2015

  5. [5]

    Wei, J. et al. Machine learning in materials science. InfoMat 1, 338–358 (2019)

    work page 2019

  6. [6]

    Tao, Q., Xu, P., Li, M. & Lu, W. Machine learning for perovskite materials design and discovery. NPJ Comput Mater 7, 23 (2021)

    work page 2021

  7. [7]

    Gubernatis, J. E. & Lookman, T. Machine learning in materials design and discovery: Examples from the present and suggestions for the future. Phys Rev Mater 2, 120301 (2018)

    work page 2018

  8. [8]

    & Kim, C

    Ramprasad, R., Batra, R., Pilania, G., Mannodi-Kanakkithodi, A. & Kim, C. Machine learning in materials informatics: recent applications and prospects. NPJ Comput Mater 3, 54 (2017)

    work page 2017

  9. [9]

    Vamathevan, J. et al. Applications of machine learning in drug discovery and development. Nat Rev Drug Discov 18, 463–477 (2019)

    work page 2019

  10. [10]

    S., Babu, C

    Dara, S., Dhamercherla, S., Jadav, S. S., Babu, C. M. & Ahsan, M. J. Machine Learning in Drug Discovery: A Review. Artif Intell Rev 55, 1947–1999 (2022)

    work page 1947

  11. [11]

    Patel, L., Shukla, T., Huang, X., Ussery, D. W. & Wang, S. Machine Learning Methods in Drug Discovery. Molecules 25, 5277 (2020)

    work page 2020

  12. [13]

    E., Torng, W

    Lo, Y.-C., Rensi, S. E., Torng, W. & Altman, R. B. Machine learning in chemoinformatics and drug discovery. Drug Discov Today 23, 1538– 1546 (2018)

    work page 2018

  13. [14]

    & Zhu, H

    Zhang, L., Tan, J., Han, D. & Zhu, H. From machine learning to deep learning: progress in machine intelligence for rational drug discovery. Drug Discov Today 22, 1680–1685 (2017)

    work page 2017

  14. [15]

    Ekins, S. et al. Exploiting machine learning for end-to-end drug discovery and development. Nat Mater 18, 435–441 (2019)

    work page 2019

  15. [16]

    A critical review of five machine learning-based algorithms for predicting protein stability changes upon mutation

    Fang, J. A critical review of five machine learning-based algorithms for predicting protein stability changes upon mutation. Brief Bioinform 21, 1285–1292 (2020)

    work page 2020

  16. [17]

    & Clementi, C

    Noé , F., De Fabritiis, G. & Clementi, C. Machine learning for protein folding and dynamics. Curr Opin Struct Biol 60, 77–84 (2020)

    work page 2020

  17. [18]

    Xie, Z.-R., Chen, J. & Wu, Y. Predicting Protein–protein Association Rates using Coarse-grained Simulation and Machine Learning. Sci Rep 7, 46622 (2017)

    work page 2017

  18. [19]

    Machine learning in protein structure prediction

    AlQuraishi, M. Machine learning in protein structure prediction. Curr Opin Chem Biol 65, 1–8 (2021)

    work page 2021

  19. [20]

    Distance-based protein folding powered by deep learning

    Xu, J. Distance-based protein folding powered by deep learning. Proceedings of the National Academy of Sciences 116, 16856–16865 (2019)

    work page 2019

  20. [21]

    & Cheng, J

    Jo, T., Hou, J., Eickholt, J. & Cheng, J. Improving Protein Fold Recognition by Deep Learning Networks. Sci Rep 5, 17573 (2015)

    work page 2015

  21. [22]

    Abramson, J. et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 630, 493–500 (2024)

    work page 2024

  22. [23]

    & Elofsson, A

    Bryant, P., Pozzati, G. & Elofsson, A. Improved prediction of protein- protein interactions using AlphaFold2. Nat Commun 13, 1265 (2022)

    work page 2022

  23. [24]

    Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021)

    work page 2021

  24. [25]

    Xu, X. et al. Photonic Perceptron Based on a Kerr Microcomb for High‐ Speed, Scalable, Optical Neural Networks. Laser Photon Rev 14, (2020)

    work page 2020

  25. [26]

    & Ozcan, A

    Mengu, D., Luo, Y., Rivenson, Y. & Ozcan, A. Analysis of Diffractive Optical Neural Networks and Their Integration With Electronic Neural Networks. IEEE Journal of Selected Topics in Quantum Electronics 26, 1–14 (2020)

    work page 2020

  26. [27]

    & Englund, D

    Hamerly, R., Bernstein, L., Sludds, A., Soljačić, M. & Englund, D. Large- Scale Optical Neural Networks Based on Photoelectric Multiplication. Phys Rev X 9, 021032 (2019)

    work page 2019

  27. [28]

    Fu, T. et al. Optical neural networks: progress and challenges. Light Sci Appl 13, 263 (2024)

    work page 2024

  28. [29]

    Williamson, I. A. D. et al. Reprogrammable Electro-Optic Nonlinear Activation Functions for Optical Neural Networks. IEEE Journal of Selected Topics in Quantum Electronics 26, 1–12 (2020)

    work page 2020

  29. [30]

    Zuo, Y. et al. All-optical neural network with nonlinear activation functions. Optica 6, 1132 (2019)

    work page 2019

  30. [31]

    Lin, X. et al. All-optical machine learning using diffractive deep neural networks. Science (1979) 361, 1004–1008 (2018)

    work page 1979

  31. [33]

    & Guo, L

    Wang, H. & Guo, L. J. NEUTRON: Neural particle swarm optimization for material-aware inverse design of structural color. iScience 25, 104339 (2022)

    work page 2022

  32. [34]

    Liu, C., Maier, S. A. & Li, G. Genetic-Algorithm-Aided Meta-Atom Multiplication for Improved Absorption and Coloration in Nanophotonics. ACS Photonics 7, 1716–1722 (2020)

    work page 2020

  33. [35]

    Ren, S. et al. Inverse deep learning methods and benchmarks for artificial electromagnetic material design. Nanoscale 14, 3958–3969 (2022)

    work page 2022

  34. [36]

    & Guo, L

    Ma, T., Tobah, M., Wang, H. & Guo, L. J. Benchmarking deep learning- based models on nanophotonic inverse design problems. Opto- Electronic Science 1, 210012–210012 (2022)

    work page 2022

  35. [37]

    & Guo, L

    Han, X., Fan, Z., Liu, Z., Li, C. & Guo, L. J. Inverse design of metasurface optical filters using deep neural network with high degrees of freedom. InfoMat 3, 432–442 (2021)

    work page 2021

  36. [38]

    & Zang, J

    Huang, Z., Liu, X. & Zang, J. The inverse design of structural color using machine learning. Nanoscale 11, 21748–21758 (2019)

    work page 2019

  37. [39]

    & Liu, L

    Ma, L., Hu, K., Wang, C., Yang, J.-Y. & Liu, L. Prediction and Inverse Design of Structural Colors of Nanoparticle Systems via Deep Neural Network. Nanomaterials 11, 3339 (2021)

    work page 2021

  38. [40]

    & Zheng, G

    Fu, R., Chen, K., Li, Z., Yu, S. & Zheng, G. Metasurface-based nanoprinting: principle, design and advances. Opto-Electronic Science 1, 220011–220011 (2022)

    work page 2022

  39. [41]

    Ji, C. et al. Engineering Light at the Nanoscale: Structural Color Filters and Broadband Perfect Absorbers. Adv Opt Mater 5, (2017)

    work page 2017

  40. [42]

    & Tian, J

    Yang, B., Cheng, H., Chen, S. & Tian, J. Structural colors in metasurfaces: principle, design and applications. Mater Chem Front 3, 750–761 (2019)

    work page 2019

  41. [43]

    Xu, T. et al. Structural Colors: From Plasmonic to Carbon Nanostructures. Small 7, 3128–3136 (2011)

    work page 2011

  42. [44]

    Wang, D. et al. Structural color generation: from layered thin films to optical metasurfaces. Nanophotonics 12, 1019–1081 (2023)

    work page 2023

  43. [45]

    A., Donev, E

    Fu, Y., Tippets, C. A., Donev, E. U. & Lopez, R. Structural colors: from natural to artificial systems. WIREs Nanomedicine and Nanobiotechnology 8, 758–775 (2016)

    work page 2016

  44. [46]

    Xuan, Z. et al. Artificial Structural Colors and Applications. The Innovation 2, 100081 (2021)

    work page 2021

  45. [47]

    & Yoshioka, S

    Kinoshita, S. & Yoshioka, S. Structural Colors in Nature: The Role of Regularity and Irregularity in the Structure. ChemPhysChem 6, 1442– 1459 (2005)

    work page 2005

  46. [48]

    & Miyazaki, J

    Kinoshita, S., Yoshioka, S. & Miyazaki, J. Physics of structural colors. Reports on Progress in Physics 71, 076401 (2008)

    work page 2008

  47. [49]

    Chang, S., Guo, X. & Ni, X. Optical Metasurfaces: Progress and Applications. Annu Rev Mater Res 48, 279–302 (2018)

    work page 2018

  48. [50]

    & Wang, J

    Shao, L., Zhuo, X. & Wang, J. Advanced Plasmonic Materials for Dynamic Color Display. Advanced Materials 30, (2018)

    work page 2018

  49. [51]

    Wang, Y. et al. Stepwise-Nanocavity-Assisted Transmissive Color Filter Array Microprints. Research 2018, (2018)

    work page 2018

  50. [52]

    Zhao, Y. et al. Artificial Structural Color Pixels: A Review. Materials 10, 944 (2017)

    work page 2017

  51. [53]

    Zhao, Y., Xie, Z., Gu, H., Zhu, C. & Gu, Z. Bio-inspired variable structural color materials. Chem Soc Rev 41, 3297 (2012)

    work page 2012

  52. [54]

    & Zhao, Y

    Zhang, Z., Chen, Z., Shang, L. & Zhao, Y. Structural Color Materials from Natural Polymers. Adv Mater Technol 6, (2021)

    work page 2021

  53. [55]

    Flauraud, V., Reyes, M., Paniagua-Domí nguez, R., Kuznetsov, A. I. & Brugger, J. Silicon Nanostructures for Bright Field Full Color Prints. ACS Photonics 4, 1913–1919 (2017)

    work page 1913

  54. [56]

    Tan, S. J. et al. Plasmonic Color Palettes for Photorealistic Printing with Aluminum Nanostructures. Nano Lett 14, 4023–4029 (2014)

    work page 2014

  55. [57]

    U., Schnoering, G., Damak, M., Poulikakos, D

    Hail, C. U., Schnoering, G., Damak, M., Poulikakos, D. & Eghlidi, H. A Plasmonic Painter’s Method of Color Mixing for a Continuous Red– Green–Blue Palette. ACS Nano 14, 1783–1791 (2020)

    work page 2020

  56. [58]

    Zang, X. et al. Polarization Encoded Color Image Embedded in a Dielectric Metasurface. Advanced Materials 30, (2018)

    work page 2018

  57. [59]

    H., Yoo, Y

    Ko, J. H., Yoo, Y. J., Kim, Y. J., Lee, S. & Song, Y. M. Flexible, Large‐ Area Covert Polarization Display Based on Ultrathin Lossy Nanocolumns on a Metal Film. Adv Funct Mater 30, (2020)

    work page 2020

  58. [60]

    Song, M. et al. Color display and encryption with a plasmonic polarizing metamirror. Nanophotonics 7, 323–331 (2018)

    work page 2018

  59. [62]

    L., de Groot, C

    Dai, P., Sun, K., Muskens, O. L., de Groot, C. H. & Huang, R. Inverse design of a vanadium dioxide based dynamic structural color via conditional generative adversarial networks. Opt Mater Express 12, 3970 (2022)

    work page 2022

  60. [63]

    Dai, P. et al. Accurate inverse design of Fabry–Perot-cavity-based color filters far beyond sRGB via a bidirectional artificial neural network. Photonics Res 9, B236 (2021)

    work page 2021

  61. [64]

    Abd Elaziz, M. et al. Advanced metaheuristic optimization techniques in applications of deep neural networks: a review. Neural Comput Appl 33, 14079–14099 (2021)

    work page 2021

  62. [65]

    Tsakyridis, A. et al. Photonic neural networks and optics-informed deep learning fundamentals. APL Photonics 9, (2024)

    work page 2024

  63. [66]

    Hegde, R. S. Deep learning: a new tool for photonic nanostructure design. Nanoscale Adv 2, 1007–1023 (2020)

    work page 2020

  64. [67]

    Heidari, A., Navimipour, N. J. & Unal, M. Applications of ML/DL in the management of smart cities and societies based on new trends in information technologies: A systematic literature review. Sustain Cities Soc 85, 104089 (2022)

    work page 2022

  65. [68]

    Xu, S. et al. Deep-learning-powered photonic analog-to-digital conversion. Light Sci Appl 8, 66 (2019)

    work page 2019

  66. [69]

    & Mesgari, M

    Kaveh, M. & Mesgari, M. S. Application of Meta-Heuristic Algorithms for Training Neural Networks and Deep Learning Architectures: A Comprehensive Review. Neural Process Lett 55, 4519–4622 (2023)

    work page 2023

  67. [70]

    Hadi, M. U. et al. Large Language Models: A Comprehensive Survey of its Applications, Challenges, Limitations, and Future Prospects. Preprint at https://doi.org/10.36227/techrxiv.23589741.v6 (2024)

    work page doi:10.36227/techrxiv.23589741.v6 2024

  68. [71]

    Meyer, J. G. et al. ChatGPT and large language models in academia: opportunities and challenges. BioData Min 16, 20 (2023)

    work page 2023

  69. [73]

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

    work page 2024

  70. [74]

    Chen, M. et al. Evaluating Large Language Models Trained on Code. (2021)

    work page 2021

  71. [75]

    Zhao, W. X. et al. A Survey of Large Language Models. (2023)

    work page 2023

  72. [76]

    Wei, J. et al. Emergent Abilities of Large Language Models. (2022)

    work page 2022

  73. [77]

    & Rajkumar, R

    Ganesan, P., Rajini, V. & Rajkumar, R. I. Segmentation and edge detection of color images using CIELAB color space and edge detectors. in INTERACT-2010 393–397 (IEEE, 2010). doi:10.1109/INTERACT.2010.5706186

    work page doi:10.1109/interact.2010.5706186 2010

  74. [78]

    & Fleiss, T

    Connolly, C. & Fleiss, T. A study of efficiency and accuracy in the transformation from RGB to CIELAB color space. IEEE Transactions on Image Processing 6, 1046–1048 (1997)

    work page 1997

  75. [79]

    Weatherall, I. L. & Coombs, B. D. Skin Color Measurements in Terms of CIELAB Color Space Values. Journal of Investigative Dermatology 99, 468–473 (1992)

    work page 1992

  76. [80]

    & Lee, C.-G

    Son, D.-K., Cho, E.-B., Moon, I.-K., Park, Y.-S. & Lee, C.-G. Development of an Illumination Measurement Device for Color Distribution Based on a CIE 1931 XYZ Sensor. J Opt Soc Korea 15, 44–51 (2011)

    work page 1931

  77. [81]

    & others

    Pope, R., Douglas, S., Chowdhery, A. & others. Efficiently scaling transformer inference. Proceedings of Machine Learning and Systems 5, 606–624 (2023)

    work page 2023

  78. [82]

    R., Ghasemi, M., Hassanzadeh, A

    Fallah, H. R., Ghasemi, M., Hassanzadeh, A. & Steki, H. The effect of annealing on structural, electrical and optical properties of nanostructured ITO films prepared by e-beam evaporation. Mater Res Bull 42, 487–496 (2007)

    work page 2007

  79. [83]

    & Hash, D

    Meyyappan, M., Delzeit, L., Cassell, A. & Hash, D. Carbon nanotube growth by PECVD: a review. Plasma Sources Sci Technol 12, 205–216 (2003)

    work page 2003

  80. [84]

    A., Melamed, M

    Kamentsky, L. A., Melamed, M. R. & Derman, H. Spectrophotometer: New Instrument for Ultrarapid Cell Analysis. Science (1979) 150, 630– 631 (1965)

    work page 1979

Showing first 80 references.