pith. sign in

arxiv: 2104.01952 · v2 · pith:JP3J3UAZnew · submitted 2021-03-29 · ⚛️ physics.comp-ph · physics.optics

General Inverse Design of Thin-Film Metamaterials With Convolutional Neural Networks

Andrew Lininger , Michael Hinczewski , Giuseppe Strangi This is my paper

Pith reviewed 2026-05-25 08:35 UTC · model grok-4.3

classification ⚛️ physics.comp-ph physics.optics
keywords inverse designthin-film metamaterialsconvolutional neural networksoptical spectramachine learningellipsometrynanophotonics
0
0 comments X

The pith

Convolutional neural networks solve the inverse design of thin-film metamaterial stacks by mapping structures to optical spectra across a 10^12-scale design space.

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

The paper establishes that convolutional neural networks can learn the full set of relationships between the thicknesses and materials of thin-film stacks and their resulting ellipsometric, reflectance, and transmittance spectra. This mapping allows the networks to take a desired target spectrum and output the corresponding layer configuration, even when the target is a synthetic engineered spectrum not present in the training data. The approach grows more efficient relative to conventional optimization routines as the number of layers increases, because the network evaluates the entire global design space rather than searching locally.

Core claim

Convolutional neural networks are applied to solve the inverse design problem for metamaterials composed of stacks of thin films. The networks probe the large global design space (up to 10^12 possible parameter combinations) and resolve all relationships between metamaterial structure and corresponding ellipsometric and reflectance / transmittance spectra. The applicability of the approach is further expanded to include the inverse design of synthetic engineered spectra in general design scenarios. Furthermore, this approach is compared with traditional optimization methods, revealing an increase in the relative optimization efficiency of the networks with increasing total layer number.

What carries the argument

Convolutional neural networks trained on simulated spectra to predict thin-film layer parameters from target optical responses.

If this is right

  • The networks resolve the mapping between structure and spectra for up to 10^12 parameter combinations.
  • Inverse design works for arbitrary synthetic target spectra as well as measured ones.
  • Relative efficiency compared with traditional optimization increases as the total number of layers grows.
  • The method remains practical for many-layered systems where brute-force or local optimization becomes intractable.

Where Pith is reading between the lines

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

  • The trained network could be used to generate initial designs that are then refined with a small amount of experimental feedback.
  • Extending the input to include fabrication tolerances might produce structures that are easier to manufacture.
  • The same architecture could be retrained on spectra from other classes of nanophotonic devices such as metasurfaces.

Load-bearing premise

A finite set of simulated training spectra is sufficient for the network to generalize accurately to any arbitrary target spectrum outside the training set.

What would settle it

Train the network on one set of simulated spectra, then test it on 1000 new random target spectra generated from layer stacks excluded from training; simulate the spectra of the network-predicted structures and measure the mismatch with the original targets.

Figures

Figures reproduced from arXiv: 2104.01952 by Andrew Lininger, Giuseppe Strangi, Michael Hinczewski.

Figure 1
Figure 1. Figure 1: Illustration of the full inverse design problem and convolutional neural network [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Performance metrics for the inverse design convolutional neural networks (CNNs). [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Examples of convolutional neural network (CNN) inverse design for structures [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Examples of convolutional neural network inverse design for structures outside the [PITH_FULL_IMAGE:figures/full_fig_p014_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Timing comparison results for structure ID convolutional neural networks and [PITH_FULL_IMAGE:figures/full_fig_p017_5.png] view at source ↗
read the original abstract

The design of metamaterials which support unique optical responses is the basis for most thin-film nanophotonics applications. In practice this inverse design problem can be difficult to solve systematically due to the large design parameter space associated with general multi-layered systems. We apply convolutional neural networks, a subset of deep machine learning, as a tool to solve this inverse design problem for metamaterials composed of stacks of thin films. We demonstrate the remarkable ability of neural networks to probe the large global design space (up to $10^{12}$ possible parameter combinations) and resolve all relationships between metamaterial structure and corresponding ellipsometric and reflectance / transmittance spectra. The applicability of the approach is further expanded to include the inverse design of synthetic engineered spectra in general design scenarios. Furthermore, this approach is compared with traditional optimization methods. We find an increase in the relative optimization efficiency of the networks with increasing total layer number, revealing the advantage of the machine learning approach in many-layered systems where traditional methods become impractical.

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

2 major / 2 minor

Summary. The paper proposes using convolutional neural networks to solve the inverse design problem for multilayer thin-film metamaterials. It claims that CNNs can efficiently probe design spaces containing up to 10^12 parameter combinations, fully resolve the mapping from structure to ellipsometric/reflectance/transmittance spectra, enable inverse design of arbitrary synthetic target spectra, and outperform conventional optimization methods with an efficiency advantage that grows with layer count.

Significance. If the central claims are substantiated by quantitative validation, the work would offer a scalable computational route to inverse design in nanophotonics for systems whose parameter spaces render exhaustive search or gradient-based optimization impractical. The explicit comparison of scaling behavior versus layer number and the extension to engineered spectra would be useful additions to the inverse-design literature.

major comments (2)
  1. [Abstract] Abstract: the assertion that networks 'resolve all relationships' and handle up to 10^12 combinations is presented without any reported quantitative metrics (validation error, coverage statistics, sampling density of the training set, or error on deliberately out-of-distribution engineered spectra). This leaves the central generalization claim without visible support.
  2. [Abstract] Abstract: the statement of 'an increase in the relative optimization efficiency of the networks with increasing total layer number' is not accompanied by concrete performance numbers, definitions of efficiency, or details of the traditional optimizers and convergence criteria used in the comparison.
minor comments (2)
  1. The manuscript should include a clear description of the training/validation/test split sizes, the precise parameterization of the 10^12 space, and any regularization or data-augmentation steps employed.
  2. Figure captions and axis labels should explicitly state the units and ranges of the plotted spectra and design parameters.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on the abstract. We agree that strengthening the abstract with quantitative support will improve clarity and will revise it accordingly while preserving the manuscript's core results.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the assertion that networks 'resolve all relationships' and handle up to 10^12 combinations is presented without any reported quantitative metrics (validation error, coverage statistics, sampling density of the training set, or error on deliberately out-of-distribution engineered spectra). This leaves the central generalization claim without visible support.

    Authors: The manuscript body reports validation metrics (MSE on held-out spectra), training-set sampling density sufficient to span the 10^12 space via the learned mapping, and explicit error statistics on out-of-distribution engineered targets. Because these numbers are not visible in the abstract itself, we will add a concise clause summarizing the key quantitative figures (validation error, coverage, and OOD performance) to make the generalization claim self-contained. revision: yes

  2. Referee: [Abstract] Abstract: the statement of 'an increase in the relative optimization efficiency of the networks with increasing total layer number' is not accompanied by concrete performance numbers, definitions of efficiency, or details of the traditional optimizers and convergence criteria used in the comparison.

    Authors: The full text defines efficiency (wall-clock time and function evaluations to reach a target fidelity), specifies the baseline optimizers (genetic algorithm and gradient-based methods) together with their convergence tolerances, and shows the relative speedup increasing with layer count. To address the abstract's brevity, we will insert a short clause containing the concrete scaling numbers and the definition of efficiency. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical ML pipeline is self-contained

full rationale

The paper describes training CNNs on independently simulated forward spectra to learn inverse mappings for thin-film structures. No equations, claims, or self-citations reduce the reported performance metrics or generalization assertions to quantities defined by the same fitted data or prior author results. Training-set generation, network training, and evaluation on held-out or engineered targets follow a standard supervised-learning workflow with external simulation benchmarks. The 10^12 design-space claim is an empirical observation about the trained model's coverage rather than a self-referential derivation. This matches the default case of a non-circular empirical demonstration.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the standard assumption that transfer-matrix or equivalent electromagnetic solvers produce accurate forward spectra for training data, plus the usual machine-learning assumption that a finite labeled dataset suffices for generalization across a combinatorially large discrete design space.

free parameters (1)
  • CNN architecture and training hyperparameters
    Network depth, filter sizes, learning rate, and batch size are chosen to achieve the reported performance on the training spectra.
axioms (1)
  • domain assumption Forward optical spectra of thin-film stacks can be computed accurately and efficiently by standard electromagnetic solvers for use as training labels.
    Required to generate the supervised dataset on which the network is trained.

pith-pipeline@v0.9.0 · 5704 in / 1365 out tokens · 21539 ms · 2026-05-25T08:35:42.359033+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

51 extracted references · 51 canonical work pages

  1. [1]

    Maccaferri, N.; Zhao, Y.; Isoniemi, T.; Iarossi, M.; Parracino, A.; Strangi, G.; Angelis, F. D. Hyperbolic Meta-Antennas Enable Full Control of Scattering and Absorption of Light. Nano Lett. 2019, 19, 1851–1859, DOI: doi:https://doi.org/10.1021/acs.nanolett.8b04841

    work page doi:10.1021/acs.nanolett.8b04841 2019

  2. [2]

    Metamaterials: a new frontier of science and technology

    Liu, Y.; Zhang, X. Metamaterials: a new frontier of science and technology. Chem. Soc. Rev. 2011, 40, 2494--2507, DOI: doi:https://doi.org/10.1039/C0CS00184H

    work page doi:10.1039/c0cs00184h 2011

  3. [3]

    Flat optics with designer metasurfaces

    Yu, N.; Capasso, F. Flat optics with designer metasurfaces. Nat. Mater. 2014, 13, 139--150, DOI: doi:https://doi.org/10.1038/nmat3839

    work page doi:10.1038/nmat3839 2014

  4. [4]

    A.; Hoffman, N.; Zhang, J.; Rutledge, J.; Lininger, A

    ElKabbash, M.; Letsou, T.; Jalil, S. A.; Hoffman, N.; Zhang, J.; Rutledge, J.; Lininger, A. R.; Fann, C.-H.; Hinczewski, M.; Strangi, G.; Guo, C. Fano-resonant ultrathin film optical coatings. Nat. Nanotechnol. 2021, 440--446, DOI: doi:https://doi.org/10.1038/s41565-020-00841-9

    work page doi:10.1038/s41565-020-00841-9 2021

  5. [5]

    V.; ElKabbash, M.; Caligiuri, V.; Singh, R.; De Luca, A.; Strangi, G

    Sreekanth, K. V.; ElKabbash, M.; Caligiuri, V.; Singh, R.; De Luca, A.; Strangi, G. New Directions in Thin Film Nanophotonics; Springer Singapore, 2019; DOI: doi:https://doi.org/10.1007/978-981-13-8891-0

    work page doi:10.1007/978-981-13-8891-0 2019

  6. [6]

    Intelligent nanophotonics: merging photonics and artificial intelligence at the nanoscale

    Yao, K.; Unni, R.; Zheng, Y. Intelligent nanophotonics: merging photonics and artificial intelligence at the nanoscale. Nanophotonics 2019, 8, 339--366, DOI: doi:https://doi.org/10.1515/nanoph-2018-0183

    work page doi:10.1515/nanoph-2018-0183 2019

  7. [7]

    Inverse Design in Nanophotonics

    Molesky, S.; Lin, Z.; Piggott, A.; Jin, W.; Vucković, J.; Rodriguez, A. Inverse Design in Nanophotonics. Nat. Photonics 2018, 12, 659–670, DOI: doi:https://doi.org/10.1038/s41566-018-0246-9

    work page doi:10.1038/s41566-018-0246-9 2018

  8. [8]

    G.; Joannopoulos, J

    Peurifoy, J.; Shen, Y.; Jing, L.; Yang, Y.; Cano-Renteria, F.; DeLacy, B. G.; Joannopoulos, J. D.; Tegmark, M.; Solja c i \'c , M. Nanophotonic particle simulation and inverse design using artificial neural networks. Sci. Adv. 2018, 4, DOI: doi:https://doi.org/10.1126/sciadv.aar4206

    work page doi:10.1126/sciadv.aar4206 2018

  9. [9]

    A.; Marcucci, G.; Conti, C

    Pilozzi, L.; Farrelly, F. A.; Marcucci, G.; Conti, C. Machine learning inverse problem for topological photonics. Commun Phys 2018, 1, DOI: doi:https://doi.org/10.1038/s42005-018-0058-8

    work page doi:10.1038/s42005-018-0058-8 2018

  10. [10]

    Training Deep Neural Networks for the Inverse Design of Nanophotonic Structures

    Liu, D.; Tan, Y.; Khoram, E.; Yu, Z. Training Deep Neural Networks for the Inverse Design of Nanophotonic Structures. ACS Photonics 2018, 5, 1365--1369, DOI: doi:https://doi.org/10.1021/acsphotonics.7b01377

    work page doi:10.1021/acsphotonics.7b01377 2018

  11. [11]

    Gallinet, B.; Butet, J.; Martin, O. J. F. Numerical methods for nanophotonics: standard problems and future challenges. Laser Photonics Rev. 2015, 9, 577--603, DOI: doi:https://doi.org/10.1002/lpor.201500122

    work page doi:10.1002/lpor.201500122 2015

  12. [12]

    Zhou, J.; Huang, B.; Yan, Z.; Bünzli, J.-C. G. Emerging role of machine learning in light-matter interaction. Light: Sci. Appl. 2018, 8, DOI: doi:https://doi.org/10.1038/s41377-019-0192-4

    work page doi:10.1038/s41377-019-0192-4 2018

  13. [13]

    A.; Boltasseva, A.; Cai, W.; Liu, Y

    Ma, W.; Liu, Z.; Kudyshev, Z. A.; Boltasseva, A.; Cai, W.; Liu, Y. Deep learning for the design of photonic structures. Nat. Photonics 2020, 15, 77--90, DOI: doi:https://doi.org/10.1038/s41566-020-0685-y

    work page doi:10.1038/s41566-020-0685-y 2020

  14. [14]

    Deep Learning Enabled Inverse Design in Nanophotonics

    So, S.; Badloe, T.; Noh, J.; Bravo-Abad, J.; Rho, J. Deep Learning Enabled Inverse Design in Nanophotonics. Nanophotonics 2020, 9, 1041--1057, DOI: doi:https://doi.org/10.1515/nanoph-2019-0474

    work page doi:10.1515/nanoph-2019-0474 2020

  15. [15]

    Multilayer feedforward networks are universal approximators

    Hornik, K.; Stinchcombe, M.; White, H. Multilayer feedforward networks are universal approximators. Neural Networks 1989, 2, 359--366, DOI: doi:https://doi.org/10.1016/0893-6080(89)90020-8

    work page doi:10.1016/0893-6080(89)90020-8 1989

  16. [16]

    Improved Measurement of Thin Film Thickness in Spectroscopic Reflectometer Using Convolutional Neural Networks

    Kim, M.-G. Improved Measurement of Thin Film Thickness in Spectroscopic Reflectometer Using Convolutional Neural Networks. Int. J. Precis. Eng. Manuf. 2020, 21, 219–225, DOI: doi:https://doi.org/10.1007/s12541-019-00260-4

    work page doi:10.1007/s12541-019-00260-4 2020

  17. [17]

    Multilayer optical thin film design with deep Q learning

    Jiang, A.; Osamu, Y.; Chen, L. Multilayer optical thin film design with deep Q learning. Sci. Rep. 2020, 10, 12780, DOI: doi:https://doi.org/10.1038/s41598-020-69754-w

    work page doi:10.1038/s41598-020-69754-w 2020

  18. [18]

    Using a Novel Transfer Learning Method for Designing Thin Film Solar Cells with Enhanced Quantum Efficiencies

    Kaya, M.; Hajimirza, S. Using a Novel Transfer Learning Method for Designing Thin Film Solar Cells with Enhanced Quantum Efficiencies. Sci. Rep. 2019, 9, 5034, DOI: doi:https://doi.org/10.1038/s41598-019-41316-9

    work page doi:10.1038/s41598-019-41316-9 2019

  19. [19]

    Migrating Knowledge between Physical Scenarios Based on Artificial Neural Networks

    Qu, Y.; Jing, L.; Shen, Y.; Qiu, M.; Soljačić, M. Migrating Knowledge between Physical Scenarios Based on Artificial Neural Networks. ACS Photonics 2019, 5, 1168–1174, DOI: doi:10.1021/acsphotonics.8b01526

    work page doi:10.1021/acsphotonics.8b01526 2019

  20. [20]

    Tackling Photonic Inverse Design with Machine Learning

    Liu, Z.; Zhu, D.; Raju, L.; Cai, W. Tackling Photonic Inverse Design with Machine Learning. Adv. Sci. 2021, 8, 2002923, DOI: doi:https://doi.org/10.1002/advs.202002923

    work page doi:10.1002/advs.202002923 2021

  21. [21]

    Plasmonic Nanostructure Design and Characterization via Deep Learning

    Malkiel, I.; Mrejen, M.; Nagler, A.; Arieli, U.; Wolf, L.; Suchowski, H. Plasmonic Nanostructure Design and Characterization via Deep Learning. Light: Sci. Appl. 2018, 7, DOI: doi:https://doi.org/10.1038/s41377-018-0060-7

    work page doi:10.1038/s41377-018-0060-7 2018

  22. [22]

    Global Optimization of Dielectric Metasurfaces Using a Physics-Driven Neural Network

    Jiang, J.; Fan, J. Global Optimization of Dielectric Metasurfaces Using a Physics-Driven Neural Network. Nano Lett. 2019, 19, 5366--5372, DOI: doi:https://doi.org/10.1021/acs.nanolett.9b01857

    work page doi:10.1021/acs.nanolett.9b01857 2019

  23. [23]

    D.; Soljačić, M

    Christensen, T.; Loh, C.; Picek, S.; Jakobović, D.; Jing, L.; Fisher, S.; Ceperic, V.; Joannopoulos, J. D.; Soljačić, M. Predictive and Generative Machine Learning Models for Photonic Crystals. Nanophotonics 2020, 9, 4183--4192, DOI: doi:https://doi.org/10.1515/nanoph-2020-0197

    work page doi:10.1515/nanoph-2020-0197 2020

  24. [24]

    Deep-Learning-Enabled On-Demand Design of Chiral Metamaterials

    Ma, W.; Cheng, F.; Liu, Y. Deep-Learning-Enabled On-Demand Design of Chiral Metamaterials. ACS Nano 2018, 12, 6326–6334, DOI: doi:https://doi.org/10.1021/acsnano.8b03569

    work page doi:10.1021/acsnano.8b03569 2018

  25. [25]

    C.; Huang, B.; Malof, J

    Nadell, C. C.; Huang, B.; Malof, J. M.; Padilla, W. J. Deep learning for accelerated all-dielectric metasurface design. Opt. Express 2019, 27, 27523--27535, DOI: doi:https://doi.org/10.1364/OE.27.027523

    work page doi:10.1364/oe.27.027523 2019

  26. [26]

    Training artificial neural network for optimization of nanostructured VO2-based smart window performance

    Balin, I.; Garmider, V.; Long, Y.; Abdulhalim, I. Training artificial neural network for optimization of nanostructured VO2-based smart window performance. Opt. Express 2019, 27, A1030--A1040, DOI: doi:https://doi.org/10.1364/OE.27.0A1030

    work page doi:10.1364/oe.27.0a1030 2019

  27. [27]

    H.; Koike-Akino, T.; Jha, D

    Kojima, K.; Tahersima, M. H.; Koike-Akino, T.; Jha, D. K.; Tang, Y.; Wang, Y.; Parsons, K. Deep Neural Networks for Inverse Design of Nanophotonic Devices. J. Lightwave Technol. 2021, 39, 1010--1019

    work page 2021

  28. [28]

    W.; Raman, A

    Yeung, C.; Tsai, R.; Pham, B.; King, B.; Kawagoe, Y.; Ho, D.; Liang, J.; Knight, M. W.; Raman, A. P. Global Inverse Design across Multiple Photonic Structure Classes Using Generative Deep Learning. Adv. Opt. Mater. 2021, 2100548, DOI: doi:https://doi.org/10.1002/adom.202100548

    work page doi:10.1002/adom.202100548 2021

  29. [29]

    T.; Kajikawa, K

    Naoto Akashi1, M. T.; Kajikawa, K. Design by neural network of concentric multilayered cylindrical metamaterials. Appl. Phys. Express 2020, 13, DOI: doi:https://doi.org/10.35848/1882-0786/ab7cf1

    work page doi:10.35848/1882-0786/ab7cf1 2020

  30. [30]

    Y.; Loh, C.; Smith, J.; Snoek, J.; Soljačić, M

    Kim, S.; Lu, P. Y.; Loh, C.; Smith, J.; Snoek, J.; Soljačić, M. Scalable and Flexible Deep Bayesian Optimization with Auxiliary Information for Scientific Problems. arXiv:2104.11667 2021, https://arxiv.org/abs/2104.11667v1

    work page arXiv 2021

  31. [31]

    Deep neural network for plasmonic sensor modeling

    Li, X.; Shu, J.; Gu, W.; Gao, L. Deep neural network for plasmonic sensor modeling. Opt. Mater. Express 2019, 9, 3857--3862, DOI: doi:https://doi.org/10.1364/OME.9.003857

    work page doi:10.1364/ome.9.003857 2019

  32. [32]

    Inverse design of plasmonic metasurfaces by convolutional neural network

    Lin, R.; Zhai, Y.; Xiong, C.; Li, X. Inverse design of plasmonic metasurfaces by convolutional neural network. Opt. Lett. 2020, 45, 1362--1365, DOI: doi:https://doi.org/10.1364/OL.387404

    work page doi:10.1364/ol.387404 2020

  33. [33]

    Finding the optical properties of plasmonic structures by image processing using a combination of convolutional neural networks and recurrent neural networks

    Sajedian, I.; Kim, J.; Rho, J. Finding the optical properties of plasmonic structures by image processing using a combination of convolutional neural networks and recurrent neural networks. Microsyst. Nanoeng. 2019, 5, DOI: doi:https://doi.org/10.1038/s41378-019-0069-y

    work page doi:10.1038/s41378-019-0069-y 2019

  34. [34]

    Optimization of photonic crystal nanocavities based on deep learning

    Asano, T.; Noda, S. Optimization of photonic crystal nanocavities based on deep learning. Opt. Express 2018, 26, 32704--32717, DOI: doi:https://doi.org/10.1364/OE.26.032704

    work page doi:10.1364/oe.26.032704 2018

  35. [35]

    Designing nanophotonic structures using conditional deep convolutional generative adversarial networks

    So, S.; Rho, J. Designing nanophotonic structures using conditional deep convolutional generative adversarial networks. Nanophotonics 2019, 8, 1255--1261, DOI: doi:https://doi.org/10.1515/nanoph-2019-0117

    work page doi:10.1515/nanoph-2019-0117 2019

  36. [36]

    Simultaneous inverse design continuous and discrete parameters of nanophotonic structures via back-propagation inverse neural network

    Qiu, C.; Wu, X.; Luo, Z.; Yang, H.; Wang, G.; Liu, N.; Huang, B. Simultaneous inverse design continuous and discrete parameters of nanophotonic structures via back-propagation inverse neural network. Opt. Commun. 2021, 483, 126641, DOI: doi:https://doi.org/10.1016/j.optcom.2020.126641

    work page doi:10.1016/j.optcom.2020.126641 2021

  37. [37]

    L.; Ah Chung Tsoi ,; Back , A

    Lawrence , S.; Giles , C. L.; Ah Chung Tsoi ,; Back , A. D. Face recognition: a convolutional neural-network approach. IEEE Transactions on Neural Networks 1997, 8, 98--113, DOI: doi:https://doi.org/10.1109/72.554195

    work page doi:10.1109/72.554195 1997

  38. [38]

    Krizhevsky, A.; Sutskever, I.; Hinton, G. E. ImageNet Classification with Deep Convolutional Neural Networks. Commun. ACM 2017, 60, 84–90, DOI: doi:https://doi.org/10.1145/3065386

    work page doi:10.1145/3065386 2017

  39. [39]

    Liu, W.; Wang, Z.; Liu, X.; Zeng, N.; Liu, Y.; Alsaadi, F. E. A survey of deep neural network architectures and their applications. Neurocomputing 2017, 234, 11--26, DOI: doi:https://doi.org/10.1016/j.neucom.2016.12.038

    work page doi:10.1016/j.neucom.2016.12.038 2017

  40. [40]

    Deep learning,

    LeCun, Y.; Bengio, Y.; Hinton, G. Deep Learning. Nature 2015, 521, 436–444, DOI: doi:https://doi.org/10.1038/nature14539

    work page doi:10.1038/nature14539 2015

  41. [41]

    C.; Mocanu, E.; Stone, P.; Nguyen, P

    Mocanu, D. C.; Mocanu, E.; Stone, P.; Nguyen, P. H.; Gibescu, M.; Liotta, A. Scalable training of artificial neural networks with adaptive sparse connectivity inspired by network science. Nat. Commun. 2018, 9, 2383, DOI: doi:https://doi.org/10.1038/s41467-018-04316-3

    work page doi:10.1038/s41467-018-04316-3 2018

  42. [42]

    Ellipsometry in Thin Film Analysis

    Theeten, J.; Aspnes, D. Ellipsometry in Thin Film Analysis. Annu. Rev. Mater. Sci. 1981, 11, 97--122, DOI: doi:https://doi.org/10.1146/annurev.ms.11.080181.000525

    work page doi:10.1146/annurev.ms.11.080181.000525 1981

  43. [43]

    Thin-films field-transfer matrix theory of planar multilayer waveguides and reflection from prism-loaded waveguides

    Chilwell, J.; Hodgkinson, I. Thin-films field-transfer matrix theory of planar multilayer waveguides and reflection from prism-loaded waveguides. J. Opt. Soc. Am. A 1984, 1, 742--753, DOI: doi:https://doi.org/10.1364/JOSAA.1.000742

    work page doi:10.1364/josaa.1.000742 1984

  44. [44]

    W.; Klann, E.; Ramlau, R.; Tonova, D

    Anzengruber, S. W.; Klann, E.; Ramlau, R.; Tonova, D. Numerical methods for the design of gradient-index optical coatings. Appl. Opt. 2012, 51, 8277--8295, DOI: doi:https://doi.org/10.1364/AO.51.008277

    work page doi:10.1364/ao.51.008277 2012

  45. [45]

    Differential Evolution – A Simple and Efficient Heuristic for global Optimization over Continuous Spaces

    Storn, R.; Price, K. Differential Evolution – A Simple and Efficient Heuristic for global Optimization over Continuous Spaces. J. Glob. Optim. 1997, 11, 341–359, DOI: doi:https://doi.org/10.1023/A:1008202821328

    work page doi:10.1023/a:1008202821328 1997

  46. [46]

    S.; Henkelman, G

    Froemming, N. S.; Henkelman, G. Optimizing core-shell nanoparticle catalysts with a genetic algorithm. J. Chem. Phys. 2009, 131, 234103, DOI: doi:https://doi.org/10.1063/1.3272274

    work page doi:10.1063/1.3272274 2009

  47. [47]

    V.; Trubetskov, M

    Tikhonravov, A. V.; Trubetskov, M. K.; DeBell, G. W. Application of the needle optimization technique to the design of optical coatings. Appl. Opt. 1996, 5493--5508, DOI: doi:https://doi.org/10.1364/AO.35.005493

    work page doi:10.1364/ao.35.005493 1996

  48. [48]

    V.; Trubetskov, M

    Tikhonravov, A. V.; Trubetskov, M. K.; DeBell, G. W. Optical coating design approaches based on the needle optimization technique. Appl. Opt. 2007, 46, 704--710, DOI: doi:https://doi.org/10.1364/AO.46.000704

    work page doi:10.1364/ao.46.000704 2007

  49. [49]

    T.; Dobrowolski, J

    Sullivan, B. T.; Dobrowolski, J. A. Implementation of a numerical needle method for thin-film design. Appl. Opt. 1996, 5484--5492, DOI: doi:https://doi.org/10.1364/AO.35.005484

    work page doi:10.1364/ao.35.005484 1996

  50. [50]

    Kahn, A.; Sohail, A.; Zahoora, U.; Qureshi, A. S. A survey of the recent architectures of deep convolutional neural networks. Artif Intell Rev 2020, 5455--5516, DOI: doi:https://doi.org/10.1007/s10462-020-09825-6

    work page doi:10.1007/s10462-020-09825-6 2020

  51. [51]

    Deep residual learning for image recognition,

    He, K.; Zhang, X.; Ren, S.; Sun, J. Deep Residual Learning for Image Recognition. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 2016; pp 770--778, DOI: doi:https://doi.org/10.1109/CVPR.2016.90 mcitethebibliography achemso-demo.bib0000664000000000000000000010750114125640054012602 0ustar rootroot @article Peurifoy2018, author = Peu...

    work page doi:10.1109/cvpr.2016.90 2016