IP Library › Granted Patent US 12,555,200
Granted Patent B2
US 12,555,200 · App. 17/525,739 · Granted Feb 17, 2026

Image generation using one or more neural networks

Inventors: Xun Huang (Mountain View, CA); Zinan Lin (Pittsburgh, PA); Ming-Yu Liu (San Jose, CA)
Assignee: NVIDIA Corporation
G06T5/70G06T2207/20076G06T2207/20081G06T2207/20084
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,555,200
App. No.
17/525,739
Granted
Feb 17, 2026
Kind
B2
Abstract

Apparatuses, systems, and techniques are presented to generate one or more images. In at least one embodiment, one or more neural networks are used to generate one or more images based, at least in part, on one or noise values.

Claims (44)

1 . A processor, comprising:

one or more circuits to:

compare two or more first images generated by one or more neural networks to one or more second images by determining distances in a multi-dimensional image space between locations of the two or more first images and a location of at least one image of the one or more second images, wherein the two or more first images are generated using the one or more neural networks based, at least in part, on one or more noise values input into the one or more neural networks; and

update a neural network of the one or more neural networks based, at least in part, on a shortest distance of the determined distances.

2 . The processor of claim 1 , wherein the one or more neural networks include a denoising diffusion probabilistic model (DDPM) to generate the two or more first images.

3 . The processor of claim 2 , wherein the one or more circuits are further to train the DDPM, at least in part, by performing a sequence of diffusion steps to add respective amounts of noise to and an image of the one or more second images, and a sequence of reverse diffusion steps to attempt to remove these respective amounts of noise.

4 . The processor of claim 3 , wherein the one or more circuits are further to utilize the one or more noise values as inputs to generate the two or more first images for each respective reverse diffusion step, wherein the one or more noise values are sampled from a probability distribution.

5 . The processor of claim 1 , wherein the shortest distance of the determined distances is used as a loss value for adjusting network parameters of a denoising diffusion probabilistic model (DDPM) of the neural network.

6 . A system, comprising:

one or more processors to;

use one or more neural networks to compare two or more first images generated by one or more neural networks to one or more second images by determining distances in a multi-dimensional image space between locations of the two or more first images and a location of at least one image of the one or more second images, wherein the two or more first images are generated using the one or more neural networks based, at least in part, on one or more noise values input into the one or more neural networks; and

update a neural network of the one or more neural networks based, at least in part, on a shortest distance of the determined distances.

7 . The system of claim 6 , wherein the one or more neural networks include a denoising diffusion probabilistic model (DDPM) to generate the two or more first images.

8 . The system of claim 7 , wherein the one or more processors are further to train the DDPM, at least in part, by performing a sequence of diffusion steps to add respective amounts of noise to an image of the one or more second images, and a sequence of reverse diffusion steps to attempt to remove these respective amounts of noise.

9 . The system of claim 8 , wherein the one or more processors are further to utilize the one or more noise values as inputs to generate the two or more first images for each respective reverse diffusion step, wherein the one or more noise values are sampled from a probability distribution.

10 . The system of claim 6 , wherein a shortest distance of the determined distances is used as a loss value for adjusting network parameters of a denoising diffusion probabilistic model (DDPM) of the neural network.

11 . A method, comprising:

using one or more neural networks to compare two or more first images generated by one or more neural networks to one or more second images by determining distances in a multi-dimensional image space between locations of the two or more first images and a location of at least one image of the one or more second images, wherein the two or more first images are generated using the one or more neural networks based, at least in part, on one or more noise values input into the one or more neural networks; and

updating a neural network of the one or more neural networks based, at least in part, on a shortest distance of the determined distances.

12 . The method of claim 11 , wherein the one or more neural networks include a denoising diffusion probabilistic model (DDPM) to generate the two or more first images.

13 . The method of claim 12 , further comprising:

training the DDPM, at least in part, by performing a sequence of diffusion steps to add respective amounts of noise to an image of the one or more second images, and a sequence of reverse diffusion steps to attempt to remove these respective amounts of noise.

14 . The method of claim 13 , further comprising:

utilizing the one or more noise values as inputs to generate the two or more first images for each respective reverse diffusion step, wherein the one or more noise values are sampled from a probability distribution.

15 . The method of claim 11 , further comprising:

selecting a shortest distance of the determined distances as a loss value for adjusting network parameters of a denoising diffusion probabilistic model (DDPM) of the neural network.

16 . A non-transitory machine-readable medium having stored thereon a set of instructions, which if performed by one or more processors, cause the one or more processors to at least:

use one or more neural networks to compare two or more first images generated by one or more neural networks to one or more second images by determining distances in a multi-dimensional image space between locations of the two or more first images and a location of at least one image of the one or more second images, wherein the two or more first images are generated using the one or more neural networks based, at least in part, on one or more noise values input into the one or more neural networks and

update a neural network of the one or more neural networks based, at least in part, on a shortest distance of the determined distances.

17 . The non-transitory machine-readable medium of claim 16 , wherein the one or more neural networks include a denoising diffusion probabilistic model (DDPM) to generate the two or more first images.

18 . The non-transitory machine-readable medium of claim 17 , wherein the instructions if performed further cause the one or more processors to:

train the DDPM, at least in part, by performing a sequence of diffusion steps to add respective amounts of noise to an image of the one or more second images, and a sequence of reverse diffusion steps to attempt to remove these respective amounts of noise.

19 . The non-transitory machine-readable medium of claim 18 , wherein the instructions if performed further cause the one or more processors to:

utilize the one or more noise values as inputs to generate the two or more first images for each respective reverse diffusion step, wherein the one or more noise values are sampled from a probability distribution.

20 . The non-transitory machine-readable medium of claim 16 , wherein the instructions if performed further cause the one or more processors to: select a shortest distance of the determined distances as a loss value for adjusting network parameters of the a denoising diffusion probabilistic model (DDPM) of the neural network.

21 . A system, comprising:

one or more processors to:

use one or more neural networks to compare two or more first images generated by one or more neural networks to one or more second images by determining distances in a multi-dimensional image space between locations of the two or more first images and a location of at least one image of the one or more second images, wherein the two or more first images are generated using the one or more neural networks based, at least in part, on one or more noise values input into the one or more neural networks; and

update a neural network of the one or more neural networks based, at least in part, on a shortest distance of the determined distances; and

memory for storing the network parameters for the one or more neural networks.

22 . The system of claim 21 , wherein the one or more neural networks include a denoising diffusion probabilistic model (DDPM) to generate the two or more first images.

23 . The system of claim 22 , wherein the one or more processors are further to train the DDPM, at least in part, by performing a sequence of diffusion steps to add respective amounts of noise to an image of the one or more second images, and a sequence of reverse diffusion steps to attempt to remove these respective amounts of noise.

24 . The system of claim 23 , wherein the one or more processors are further to utilize the one or more noise values as inputs to generate the two or more first images for each respective reverse diffusion step, wherein the one or more noise values are sampled from a probability distribution.

25 . The system of claim 21 , wherein a shortest distance of the determined distances is used as a loss value for adjusting network parameters of a denoising diffusion probabilistic model (DDPM) of the neural network.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Dec 22, 2021
From: HUANG, XUN; LIN, ZINAN; LIU, MING-YU
To: NVIDIA CORPORATION
Reel/Frame 058459/0117 →
Continuity (1)
Related Publication 20230153949A1 · May 18, 2023
References Cited (114)
US 7127127B2 · Jojic · 2006 [cited by applicant]
US 10430708B1 · Hu et al. · 2019 [cited by applicant]
US 11347973B2 · Rhee · 2022 [cited by applicant]
US 11790492B1 · Ahmad · 2023 [cited by examiner]
US 12299962B2 · Kreis · 2025 [cited by applicant]
US 20170372155A1 · Odry · 2017 [cited by applicant]
US 20170372193A1 · Mailhe · 2017 [cited by applicant]
US 20190005686A1 · Liu · 2019 [cited by applicant]
US 20190251397A1 · Tremblay et al. · 2019 [cited by applicant]
US 20200020098A1 · Odry · 2020 [cited by applicant]
US 20200065626A1 · Kaufhold · 2020 [cited by applicant]
US 20200234402A1 · Schwartz · 2020 [cited by applicant]
US 20200273167A1 · Wilson · 2020 [cited by applicant]
US 20210073959A1 · Elmalem · 2021 [cited by applicant]
US 20210142177A1 · Mallya et al. · 2021 [cited by applicant]
US 20210209735A1 · Sadrizadeh · 2021 [cited by examiner]
US 20210303732A1 · Hutmacher · 2021 [cited by applicant]
US 20210360199A1 · Oz · 2021 [cited by applicant]
US 20210392296A1 · Rabinovich · 2021 [cited by applicant]
US 20220051412A1 · Gronau · 2022 [cited by applicant]
US 20220107378A1 · Dey · 2022 [cited by applicant]
US 20220189133A1 · Fuchs · 2022 [cited by applicant]
US 20220198612A1 · Weinmann · 2022 [cited by applicant]
US 20220215510A1 · Weinmann · 2022 [cited by applicant]
US 20220222781A1 · Jacob · 2022 [cited by applicant]
US 20230067841A1 · Saharia · 2023 [cited by applicant]
US 20230138380A1 · Chen · 2023 [cited by examiner]
US 20230180842A1 · Pilkington et al. · 2023 [cited by applicant]
US 20230368337A1 · Karras · 2023 [cited by applicant]
US 20240104698A1 · Nie et al. · 2024 [cited by applicant]
US 20240161248A1 · Naik · 2024 [cited by applicant]
CN 107633486A · 2018 [cited by applicant]
CN 112101083A · 2020 [cited by applicant]
WO 2023183504 · 2023 [cited by applicant]
IEEE, “IEEE Standard 754-2008 (Revision of IEEE Standard 754-1985): IEEE Standard for Floating-Point Arithmetic,” Aug. 29, 2008, 70 pages. [cited by applicant]
Engstrom, et al., “MadrLab / Robustness,” Retrieved from https://github.com/MadryLab/robustness, 2019, 7 pages. [cited by applicant]
Goodfellow, et al, “Explaining and Harnessing Adversarial Examples,” International Conference on Learning Representations, Published as a conference paper at ICLR 2015, arXiv: 1412.6572v3, Mar. 20, 2015, 11 pages. [cited by applicant]
Gowal, et al., “Uncovering the Limits of Adversarial training Against Norm-bounded Adversarial Examples,” arXiv:2010.03593v2, Oct. 27, 2020, 30 pages. [cited by applicant]
Grathwohl, et al, “Your Classifier is Secretly an Energy Based Model and you Should Treat it Like One,” ICLR, 2020, 23pages. [cited by applicant]
He, et al., “Parametric Noise Injection: Trainable Randomness to Improve Deep Neural Network Robustness against Adversarial Attack,” In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, … [cited by applicant]
Hill, et al. “Stochastic Security: Adversarial Defensse Using Long-Run Dynamics of Energy Based Models,” Published as conference pater at ICLR 2021, arXiv:2005.13525v2, Mar. 18, 2021, 18 pages. [cited by applicant]
Ho, et al., “Denoising Diffusion Probabilistic Models,” 34th Conference on Neural Information Processing Systems (NeurIPS 2020), 25 pages. [cited by applicant]
International Electrotechnical Commission, “Functional safety of electrical/electronic/programmable electronic safety-related systems,” IEC Stgandard 61508-1, Apr. 2014, 23 pages. [cited by applicant]
Sehwag, et al., “Robust Learning Meets Generative Models: Can Proxy Distributioins Improve Adversarial Robustness?”, arXiv:2104.09425v2, Sep. 27, 2021, 27 pages. [cited by applicant]
Karras, et al., “Progressive Growing of GANs for Improved Quality, Stability, and Variation,” Pub. As a conference paper at ICLR 2018, arXiv:1710.10196v3, Feb. 26, 2018, 26 pages. [cited by applicant]
Kim, et al., “DiffusionCLIP: Text-guided Image Manipulation Using Diffusion Models,” arXiv:2110.02711v1, Oct. 5, 2021, 11 pages. [cited by applicant]
Kingma, et al., “Vaariation Diffusion Models,” Jun. 12, 2021, arXiv:2107.00630v2, 24 pages. [cited by applicant]
Laidlaw, et al., “Perceptual Adversarial Robustness: Defense Against Unseen Thret Models,” Published as a conference paper at ICLR 2021, arXiv:2006.12655v3, Jan. 25, 2021, 25 pages. [cited by applicant]
LeCun, et al., “A tutorial on Energy-Based Learning,” Predicting Structured Data, 1(0):2006, 59 pages. [cited by applicant]
Li, et al., “Scalable Gradients for Stochastic Differential Equations,” Oct. 18, 2020, 25 pages. [cited by applicant]
Madry, et al., “Towards Deep Learning Models Resistant to Adversarial Attacks,” International Conference on Learning Representations, 2018, 23 pages. [cited by applicant]
P. Samangouei, et al., “Defense-GAN: Protecting classi_ers against adversarial attackes using generative models,” In Proc. Int. Conf. Learn. Represent. 2018, pp. 1-17. [cited by applicant]
Pinot, et al., “Randomization Matters How to Defend Against Strong Adversarial Attacks,” International Conference on Machine Learning, 2020, 11 pages. [cited by applicant]
Rebuffi, et al., “Fixing Data Augmentation to Improve Adversarial Robustness,” arXiv:2103.01946v2, Oct. 18, 2021, 21 pages. [cited by applicant]
Croce, Francesco, et al., “Evaluating the adversarial Robustness of Adaptive test-time defenses,” International Conference on Machine Learning, PMLR, 2022, arXiv:2202.13711v2, pp. 1-15. [cited by applicant]
Bai, Tao, et al., “Adversarial purification through representation disentanglement,” arXiv preprint arXiv:2110.07801, 2021. [cited by applicant]
U. Hwang, et al., “PuVAE: A Variational Autoencoder to Purify Adversarial Examples” in IEEE Access, vol. 7, pp. 126582-126593, 2019, doi: 10.1109/ACCESS.2019-2939352. [cited by applicant]
Dietmar Uttenweiler, et al., “Spatotemporal anisotropic diffusion filtering to improve signal-to-noise rtios and object restoration in Fluorescence microscopic image sequences,” Sep. 2002, Journal of Biomedical Optics 8… [cited by applicant]
Yin, et al., “DiffGAR: Model-Agnostic Restoration from Generative Artifacts Using Image-to-Image Diffusion Models,” Mar. 30, 2023, CSAI '22, Proceedings of the 2022 5th International Conference on Computer Science and A… [cited by applicant]
Prasath, et al., “Analysis of adaptive forward-backward diffusion flows with applications in image processing,” Sep. 2015, Inverse Problems 31, pp. 6-25. [cited by applicant]
Siwei Yu, et al., “Deep learning for denoising,” Oct. 2019, Geophysics, vol. 84, No. 8, (Nov.-Dec. 2019), pp. V333-V342. [cited by applicant]
Kevin De Haan, et al., “Deep-Learning-Based Image Reconstruction and Enhancement in Optical Microscopy”, Dec. 2019, Proceedings of the IEEE, Vo. 108, No. 1, Jan. 2020, pp. 30-46. [cited by applicant]
Society of Automotive Engineers On-Road Automated Vehicle Standards Committee, “Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles,” Standard No. J3016-201609, issued Jan… [cited by applicant]
Society of Automotice Engineers On-Road Automated Vehicle Standards Committee, “Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles,” Standard No. J3016-201806, dated Jun.… [cited by applicant]
Song, et al., “Generative Modeling by Estimating Gradients of the Data Distribution,” arXiv:1907.05600v2, Oct. 29, 2019, 23 pages. [cited by applicant]
Song, et al., “PixelDefend: Leveraging Generative Models to Understand and Defend Against Adversarial Examples,” arXiv:1710.10766v3, May 2018, published as a conference paper at ICLR 2018, 20 pages. [cited by applicant]
Song, et al., Maximum Likelihood Traiming of Score-Based Diffusion Models, Oct. 21, 2021, 24 pages. [cited by applicant]
Song, et al., “Score-Based Generative Modeling Through Stochastic Differential Equations,” International Conference on Learning Represnetations, 2021, 36 pages. [cited by applicant]
Szegedy, et al., “Intriguing properties of neural networks,” arXiv:1312.6199v4, Feb. 19, 2014, 10 pages. [cited by applicant]
Touvron, et al., “Training Data-Efficient Image Transformers & Distillation Through Attention,” International Conference on Machine Learning, 2021, 22 pages (arXiv:2012.12877v2). [cited by applicant]
Tramer, et al., “On Adaptive Attacks to Adversarial Example Defenses,” arXiv:2002.08347v2, Oct. 23, 2020, 44 pages. [cited by applicant]
Vahdat, et al., “NVAE: A Deep Hierarchical Cariational Autoencode,” Neural Informatin Processing Systems, 2020, 13 pages. [cited by applicant]
Vahdat, et al., “Score-based Genertive Modeling in Latent Space,” Dec. 2, 2021, 46 pages. [cited by applicant]
Vincent, “A Connection Between Score Matching and Denoising Autencoders,” Neutral computation 23 (7) 1661-1674, 2011, 14 pages. [cited by applicant]
Wong, et al., Fast is Better Than Free: Revisiting Adversarial Training, Published as a Conference paper at ICLR 2020, Jan. 12, 2020, 17 pages. [cited by applicant]
Wu, et al., “Adversarial Weight Perturbatioin Helps Robust Generalization,” arXiv:2004.05884v2, Oct. 13, 2020, 20 pages. [cited by applicant]
Xiao, et al., “Spatially Transformed Adversarial Examples,” arXiv:1801.02612v2, Jan. 9, 2018, 29 pages. [cited by applicant]
Yang, et al., “ME-Net: Towards Effective Adversarial Robustness with Matrix Estimation,” Proceedings of the 36th International Conference on Machine Learning, PMLR 97, arrXiv:1905.11971v1, May 2019, pp. 1-22. [cited by applicant]
Yoon, et al., “Adversarial Purification with Score-based Generative Models,” arXiv:2106.06041v1, Jun. 11, 2021, pp. 1-20. [cited by applicant]
Zagoruyko, et al., “Wide Residual Networks,” arXiv:1605.07146v1, May 23, 2016, 13 pages. [cited by applicant]
Zhang, et al., “Geometry-Aware Instance-reweighted Adversarial Training,” Oct. 5, 2020, 24 pages. [cited by applicant]
Zhang, et al., “Theoretically Principled Trade-Off between Robustness and Accuracy,” arXiv:1901.08573v3, Jun. 2019, pp. 1-31. [cited by applicant]
Lucy Chai, et al. “Ensembling with Deep Generative Views,” arXiv:2104.14551v1, Apr. 29, 2021, pp. 1-24. [cited by applicant]
Alexander Levine, et al. “Improved, Deterministic Smoothing for lv Certified Robustness,” Proceedings of the 38th International Conference on Machine Learning, PMLR 139, 2021, pp. 1-11. [cited by applicant]
Jooyoung Choi, et al., “ILVR: Conditioning Method for Denoising Diffusion Probabilistic Models,” arXiv:2108.02938v1, Aug. 2021, pp. 1-14. [cited by applicant]
Francesco Croce, et al., “Reliable Evaluation of Adversarial Robustness with an Ensemble of Diverse Parameter-free Attacks,” Proceedings of the 37th International Conference on Machine Learning, Online, PMLR 119, arXiv:… [cited by applicant]
Gavin Weiguang Ding, et al., “MMA Training: Direct Input Space Margin Maximization through Adversarial training,” Published as a conference paper at ICLR 2020, arXiv:1812.02637v4, Mar. 2020 pp. 1-28. [cited by applicant]
Chitwan Saharia, et al., “Palette: Image-to-Image Diffusion Models,” arXiv:2111.05826v1, pp. 1-27. [cited by applicant]
Richardson, et al., “Encoding in Style: a StyleGAN Encoder for Image-to-Image Trnaslation,” Proceedings of the IEEE/CVF Conference on Computer Visioin and Pattern Recognition, 2021, 10 pages. [cited by applicant]
Maximilian Augustin, et al., Adversarial Robustness on In- and Out-Distribution Improves Explainability, arXiv:2003.09461v2, Jul. 2020, pp. 1-67. [cited by applicant]
Yutong Bai, et al., “Are Transformers More Robust Than CNNs?,” 35th Conference on Neural Information Processing Systems, arXiv:2111.05464v1, Nov. 2021, pp. 1-13. [cited by applicant]
Jia Deng, et al., “ImageNet: A Large-Scale Hierarchical Image Database,” pp. 1-8. [cited by applicant]
Jerome Rony, et al., “Decoupling Direction and Norm for Efficient Gradient-Based L2 Adversarial Attacks and Defenses,” arXiv:1811.09600v3, Apr. 2019, pp. 1-13. [cited by applicant]
Simo Sarkka and Arno Solin, “Applied Stochastic Differential Equations,” May 2019, pp. 1-324. [cited by applicant]
Ian J. Goodfellow, et al., “Generative Adversarial Nets,” arXiv:1406.2661v1, Jun. 2014, pp. 1-9. [cited by applicant]
Tero Karras, et al., “Analyzing and Improving the Image Quality of StyleGAN,” CVPR 2020, pp. 8110-8119. [cited by applicant]
Siwei Lyu, “Interpretation and Generalization of Score Matching,” UAI 2009, pp. 359-366. [cited by applicant]
Hadi M. Dolatabadi, et al., “(I infinity)—Robustness and Beyond: Unleashing Efficient Adversarial Training,” arXiv:2112.00378v1, pp. 1-12. [cited by applicant]
Chenlin Meng, et al., “Image Synthesis and Editing with Stochastic Differential Equations” arXiv:2108.01073v1, Aug. 2021, pp. 1-18. [cited by applicant]
Alex Krizhevsky, “Learning Multiple Layers of Features from Tiny Images,” Apr. 8, 2009, pp. 1-60. [cited by applicant]
Zhezhi He, et al., “Parametric Noise Injection: Training Randomness to Improve Deep Neural Network Robustness against Adversarial Attck,” CVPR, pp. 588-597. [cited by applicant]
Alex Nichol, et al., “GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models,” arXiv:2112.10741v2, Dec. 2021, pp. 1-20. [cited by applicant]
Sven Gowal, et al., “Improving Robustness using Generated Data,” arXiv:2110.09468v1, Oct. 2021, Accepted at the 35th Conference on Neural Information Processing Systems, pp. 1-25. [cited by applicant]
Kaiming He, et al., “Deep Residual Learning for Image Recognition,” CVPR Paper, pp. 770-778. [cited by applicant]
Changhao Shi, et al., “Online Adversarial Purification based on Self-Supervised Learning,” Published as a conference paper at ICLR 2021, arXiv:2101.09387v1, Jan. 2021 pp. 1-15. [cited by applicant]
Andrew R. Barron, Stanford Univ., “Entropy and the Central Limit Theorem,” The Annals of Probability, 1986, vol. 14, No. 1, pp. 336-342. [cited by applicant]
Francesco Croce, et al., “RobustBench: a standardized adversarial robustness benchmark,” arXiv:2010.09670v1, Oct. 19, 2020, pp. 1-15. [cited by applicant]
Athalye, et al., “obfuscated Gradients Five a False Sense of Security: Circumventing Defenses to Adversarial Examples,” arXiv:1802.00420v4, Jul. 31, 2018, 12 pages. [cited by applicant]
Cohen, et al., “Certified Adversarial Robustness via Randomized Smoothing,” International Conference on Machine Learning, 2019, 11 pages. [cited by applicant]
Dhariwal, et al., “Diffusion Models Beat GANs on Image Synthesis,” Jun. 1, 2021, 44 pages. [cited by applicant]
Dosovitskiy, et al., “An Image is With 16x16 Words: Transformers for Image Recognition at Sale,” Oct. 22, 2020, 21 pages. [cited by applicant]
Du, et al., “Implicit Generation and Modeling with Energy Based Models,” Neural Information Processing System, 2019, 11 pages. [cited by applicant]
Office Action mailed Aug. 2, 2025 in Chinese Patent Application No. 202211406338.8, NVIDIA Corporation, 19 pages (including translation). [cited by applicant]
Jiaming Song, et al., “Denoising Diffucion Implicit Models”, arXiv:2010.02502v1, Oct. 6, 2020, pp. 1-19. [cited by applicant]