Thomas Müller
Jan Novák
Skip Abstract Section
Skip Supplemental Material SectionAbstract
We propose neural control variates (NCV) for unbiased variance reduction in parametric Monte Carlo integration. So far, the core challenge of applying the method of control variates has been finding a good approximation of the integrand that is cheap to integrate. We show that a set of neural networks can face that challenge: a normalizing flow that approximates the shape of the integrand and another neural network that infers the solution of the integral equation. We also propose to leverage a neural importance sampler to estimate the difference between the original integrand and the learned control variate. To optimize the resulting parametric estimator, we derive a theoretically optimal, variance-minimizing loss function, and propose an alternative, composite loss for stable online training in practice. When applied to light transport simulation, neural control variates are capable of matching the state-of-the-art performance of other unbiased approaches, while providing means to develop more performant, practical solutions. Specifically, we show that the learned light-field approximation is of sufficient quality for high-order bounces, allowing us to omit the error correction and thereby dramatically reduce the noise at the cost of negligible visible bias.
Supplemental Material
- Martín Abadi, Ashish Agarwal, Paul Barham, Eugene Brevdo, Zhifeng Chen, Craig Citro, Greg S. Corrado, Andy Davis, Jeffrey Dean, et al. 2015. TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems. http://tensorflow.org/Google Scholar
- Roland Assaraf and Michel Caffarel. 1999. Zero-Variance Principle for Monte Carlo Algorithms. Phys. Rev. Lett. 83 (Dec 1999), 4682--4685. Issue 23. Google ScholarCross Ref
- Andrea Barth, Christoph Schwab, and Nathaniel Zollinger. 2011. Multi-level Monte Carlo Finite Element method for elliptic PDEs with Stochastic Coefficients. Numer. Math. 119, 1 (2011), 123--161. Google ScholarDigital Library
- Philippe Bekaert, Philipp Slusallek, Ronald Cools, Vlastimil Havran, and Hans-Peter Seidel. 2003. A custom designed Density Estimation Method for Light Transport. MPI-I-2003-4-004 (April 2003).Google Scholar
- Laurent Belcour, Cyril Soler, Kartic Subr, Nicolas Holzschuch, and Fredo Durand. 2013. 5D Covariance Tracing for Efficient Defocus and Motion Blur. ACM Trans. Graph. 32, 3, Article Article 31 (July 2013), 18 pages. Google ScholarDigital Library
- Benedikt Bitterli. 2016. Rendering resources. https://benedikt-bitterli.me/resources/.Google Scholar
- Mark Broadie and Paul Glasserman. 1998. Risk Management and Analysis, Volume 1: Measuring and Modelling Financial Risk. Wiley, New York, Chapter Simulation for option pricing and risk management, 173--208.Google Scholar
- Tian Qi Chen, Yulia Rubanova, Jesse Bettencourt, and David Duvenaud. 2018. Neural Ordinary Differential Equations. arXiv:1806.07366 (June 2018).Google Scholar
- Petrik Clarberg and Tomas Akenine-Möller. 2008. Exploiting Visibility Correlation in Direct Illumination. Computer Graphics Forum 27, 4 (2008), 1125--1136. Google ScholarDigital Library
- Ken Dahm and Alexander Keller. 2018. Learning Light Transport the Reinforced Way. In Monte Carlo and Quasi-Monte Carlo Methods, Art B. Owen and Peter W. Glynn (Eds.). Springer International Publishing, 181--195.Google Scholar
- Laurent Dinh, David Krueger, and Yoshua Bengio. 2014. NICE: Non-linear Independent Components Estimation. arXiv:1410.8516 (Oct. 2014).Google Scholar
- Laurent Dinh, Jascha Sohl-Dickstein, and Samy Bengio. 2016. Density Estimation using Real NVP. arXiv:1605.08803 (March 2016).Google Scholar
- Shaohua Fan, Stephen Chenney, Bo Hu, Kam-Wah Tsui, and Yu-Chi Lai. 2006. Optimizing Control Variate Estimators for Rendering. Computer Graphics Forum 25, 3 (2006), 351--358.Google ScholarCross Ref
- Iliyan Georgiev, Zackary Misso, Toshiya Hachisuka, Derek Nowrouzezahrai, Jaroslav Křivánek, and Wojciech Jarosz. 2019. Integral formulations of volumetric transmittance. ACM Transactions on Graphics (Proceedings of SIGGRAPH Asia) 38, 6 (Nov. 2019). https://doi.org/10/dffnGoogle Scholar
- Mathieu Germain, Karol Gregor, Iain Murray, and Hugo Larochelle. 2015. MADE: Masked Autoencoder for Distribution Estimation. In International Conference on Machine Learning. 881--889.Google Scholar
- Michael B. Giles. 2008. Improved Multilevel Monte Carlo Convergence using the Milstein Scheme. In Monte Carlo and Quasi-Monte Carlo Methods 2006, Alexander Keller, Stefan Heinrich, and Harald Niederreiter (Eds.). Springer, Berlin, Heidelberg, 343--358. Google ScholarCross Ref
- Michael B. Giles. 2013. Multilevel Monte Carlo Methods. In Monte Carlo and Quasi-Monte Carlo Methods 2012, Josef Dick, Y. Frances Kuo, W. Gareth Peters, and H. Ian Sloan (Eds.). Springer, Berlin, Heidelberg, 83--103. Google ScholarCross Ref
- Xavier Glorot and Yoshua Bengio. 2010. Understanding the Difficulty of Training Deep Feedforward Neural Networks. In Proc. 13th International Conference on Artificial Intelligence and Statistics (May 13--15). JMLR.org, 249--256.Google Scholar
- Peter W. Glynn and Roberto Szechtman. 2002. Some New Perspectives on the Method of Control Variates. In Monte Carlo and Quasi-Monte Carlo Methods 2000, Kai-Tai Fang, Harald Niederreiter, and Fred J. Hickernell (Eds.). Springer, Berlin, Heidelberg, 27--49. Google ScholarCross Ref
- Will Grathwohl, Dami Choi, Yuhuai Wu, Geoff Roeder, and David Duvenaud. 2018. Back-propagation through the Void: Optimizing control variates for black-box gradient estimation. International Conference on Learning Representations.Google Scholar
- Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep Residual Learning for Image Recognition. In IEEE Conference on Computer Vision and Pattern Recognition (CVPR).Google Scholar
- Stefan Heinrich. 1998. Monte Carlo Complexity of Global Solution of Integral Equations. Journal of Complexity 14, 2 (1998), 151 -- 175. Google ScholarDigital Library
- Stefan Heinrich. 2000. The Multilevel Method of Dependent Tests. In Advances in Stochastic Simulation Methods. Birkhäuser Boston, Boston, MA, 47--61. Google ScholarCross Ref
- Sebastian Herholz, Yangyang Zhao, Oskar Elek, Derek Nowrouzezahrai, Hendrik P. A. Lensch, and Jaroslav Křivánek. 2019. Volume Path Guiding Based on Zero-Variance Random Walk Theory. ACM Trans. Graph. 38, 3, Article 25 (June 2019), 19 pages. Google ScholarDigital Library
- Pedro Hermosilla, Sebastian Maisch, Tobias Ritschel, and Timo Ropinski. 2019. Deep-learning the Latent Space of Light Transport. Computer Graphics Forum 38, 4 (2019).Google Scholar
- Timothy C. Hesterberg and Barry L. Nelson. 1998. Control Variates for Probability and Quantile Estimation. Management Science 44, 9 (Sept. 1998), 1295--1312. Google ScholarDigital Library
- Chin-Wei Huang, David Krueger, Alexandre Lacoste, and Aaron C. Courville. 2018. Neural Autoregressive Flows. arXiv:1804.00779 (April 2018).Google Scholar
- Sergey Ioffe and Christian Szegedy. 2015. Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift. arXiv:1502.03167 (2015).Google Scholar
- Wenzel Jakob. 2010. Mitsuba Renderer. http://www.mitsuba-renderer.org.Google Scholar
- Simon Kallweit, Thomas Müller, Brian McWilliams, Markus Gross, and Jan Novák. 2017. Deep Scattering: Rendering Atmospheric Clouds with Radiance-Predicting Neural Networks. ACM Trans. Graph. 36, 6, Article 231 (Nov. 2017), 11 pages. Google ScholarDigital Library
- Alexander Keller. 2001. Hierarchical Monte Carlo Image Synthesis. Mathematics and Computers in Simulation 55, 1--3 (2001), 79 -- 92. The Second {IMACS} Seminar on Monte Carlo Methods. Google ScholarDigital Library
- Angelien Kemna and Ton Vorst. 1990. A Pricing Method for Options based on Average Asset Values. Journal of Banking & Finance 14, 1 (1990), 113--129. Google ScholarCross Ref
- Diederik P. Kingma and Jimmy Ba. 2014. Adam: A Method for Stochastic Optimization. arXiv:1412.6980 (June 2014).Google Scholar
- Diederik P. Kingma and Prafulla Dhariwal. 2018. Glow: Generative Flow with Invertible 1x1 Convolutions. arXiv:1807.03039 (July 2018).Google Scholar
- Diederik P. Kingma, Tim Salimans, Rafal Jozefowicz, Xi Chen, Ilya Sutskever, and Max Welling. 2016. Improved Variational Inference with inverse Autoregressive Flow. In Advances in Neural Information Processing Systems. 4743--4751.Google Scholar
- Ivan Kobyzev, Simon Prince, and Marcus A. Brubaker. 2019. Normalizing Flows: An Introduction and Review of Current Methods. arXiv:stat.ML/1908.09257Google Scholar
- Ivo Kondapaneni, Petr Vevoda, Pascal Grittmann, Tomáš Skřivan, Philipp Slusallek, and Jaroslav Křivánek. 2019. Optimal Multiple Importance Sampling. ACM Trans. Graph. 38, 4, Article 37 (July 2019), 14 pages. Google ScholarDigital Library
- Eric P. Lafortune and Yves D. Willems. 1994. The Ambient Term as a Variance Reducing Technique for Monte Carlo Ray Tracing. In Proc. EGWR. 163--171.Google Scholar
- Eric P. Lafortune and Yves D. Willems. 1995. A 5D Tree to Reduce the Variance of Monte Carlo Ray Tracing. In Proc. EGWR. 11--20.Google Scholar
- Stephen S. Lavenberg, Thomas L. Moeller, and Peter D. Welch. 1982. Statistical Results on Control Variables with Application to Queueing Network Simulation. Operations Research 30, 1 (1982), 182--202. Google ScholarDigital Library
- Jaakko Lehtinen, Jacob Munkberg, Jon Hasselgren, Samuli Laine, Tero Karras, Miika Aittala, and Timo Aila. 2018. Noise2Noise: Learning Image Restoration without Clean Data. arXiv:cs.CV/1803.04189Google Scholar
- Stephen Lombardi, Tomas Simon, Jason Saragih, Gabriel Schwartz, Andreas Lehrmann, and Yaser Sheikh. 2019. Neural Volumes: Learning Dynamic Renderable Volumes from Images. ACM Trans. Graph. 38, 4, Article 65 (July 2019), 14 pages. Google ScholarDigital Library
- Maxim Maximov, Laura Leal-Taixe, Mario Fritz, and Tobias Ritschel. 2019. Deep Appearance Maps. In The IEEE International Conference on Computer Vision (ICCV).Google Scholar
- Abhimitra Meka, Christian Häne, Rohit Pandey, Michael Zollhöfer, Sean Fanello, Graham Fyffe, Adarsh Kowdle, Xueming Yu, Jay Busch, Jason Dourgarian, Peter Denny, Sofien Bouaziz, Peter Lincoln, Matt Whalen, Geoff Harvey, Jonathan Taylor, Shahram Izadi, Andrea Tagliasacchi, Paul Debevec, Christian Theobalt, Julien Valentin, and Christoph Rhemann. 2019. Deep Reflectance Fields: High-quality Facial Reflectance Field Inference from Color Gradient Illumination. ACM Trans. Graph. 38, 4, Article 77 (July 2019), 12 pages. Google ScholarDigital Library
- Antonietta Mira, Reza Solgi, and Daniele Imparato. 2013. Zero variance Markov chain Monte Carlo for Bayesian estimators. Statistics and Computing 23 (2013), 653--662.Google ScholarDigital Library
- Thomas Müller. 2019. "Practical Path Guiding" in Production. In ACM SIGGRAPH Courses: Path Guiding in Production, Chapter 10. ACM, New York, NY, USA, 18:1--18:77. Google ScholarDigital Library
- Thomas Müller, Markus Gross, and Jan Novák. 2017. Practical Path Guiding for Efficient Light-Transport Simulation. Computer Graphics Forum 36, 4 (June 2017), 91--100. Google ScholarDigital Library
- Thomas Müller, Brian Mcwilliams, Fabrice Rousselle, Markus Gross, and Jan Novák. 2019. Neural Importance Sampling. ACM Trans. Graph. 38, 5, Article 145 (Oct. 2019), 19 pages. Google ScholarDigital Library
- Oliver Nalbach, Elena Arabadzhiyska, Dushyant Mehta, Hans-Peter Seidel, and Tobias Ritschel. 2017. Deep Shading: Convolutional Neural Networks for Screen-Space Shading. 36, 4 (2017).Google Scholar
- Barry L. Nelson. 1990. Control Variate Remedies. Operations Research 38, 6 (1990), 974--992. Google ScholarDigital Library
- Jan Novák, Andrew Selle, and Wojciech Jarosz. 2014. Residual Ratio Tracking for Estimating Attenuation in Participating Media. ACM Trans. Graph. 33, 6 (Nov. 2014). Google ScholarDigital Library
- C. Oates, M. Girolami, and N. Chopin. 2014. Control functionals for Monte Carlo integration. Journal of The Royal Statistical Society Series B-statistical Methodology 79 (2014), 695--718.Google ScholarCross Ref
- Art Owen and Yi Zhou. 2000. Safe and Effective Importance Sampling. J. Amer. Statist. Assoc. 95, 449 (2000), 135--143. http://www.jstor.org/stable/2669533Google ScholarCross Ref
- George Papamakarios, Iain Murray, and Theo Pavlakou. 2017. Masked Autoregressive Flow for Density Estimation. In Advances in Neural Information Processing Systems. 2338--2347.Google Scholar
- George Papamakarios, Eric Nalisnick, Danilo Jimenez Rezende, Shakir Mohamed, and Balaji Lakshminarayanan. 2019. Normalizing Flows for Probabilistic Modeling and Inference. arXiv:stat.ML/1912.02762Google Scholar
- Vincent Pegoraro, Carson Brownlee, Peter S. Shirley, and Steven G. Parker. 2008a. Towards Interactive Global Illumination Effects via Sequential Monte Carlo Adaptation. In Proceedings of the 3rd IEEE Symposium on Interactive Ray Tracing. 107--114.Google Scholar
- Vincent Pegoraro, Ingo Wald, and Steven G. Parker. 2008b. Sequential Monte Carlo Adaptation in Low-Anisotropy Participating Media. Computer Graphics Forum 27, 4 (2008), 1097--1104.Google ScholarDigital Library
- Matt Pharr, Wenzel Jacob, and Greg Humphreys. 2016. Physically Based Rendering - From Theory to Implementation. Morgan Kaufmann, Third Edition.Google Scholar
- Peiran Ren, Jiaping Wang, Minmin Gong, Stephen Lin, Xin Tong, and Baining Guo. 2013. Global Illumination with Radiance Regression Functions. ACM Trans. Graph. 32, 4, Article 130 (July 2013), 12 pages. Google ScholarDigital Library
- Danilo Rezende and Shakir Mohamed. 2015. Variational Inference with Normalizing Flows. In International Conference on Machine Learning. 1530--1538.Google Scholar
- Fabrice Rousselle, Wojciech Jarosz, and Jan Novák. 2016. Image-space Control Variates for Rendering. ACM Trans. Graph. 35, 6, Article 169 (Nov. 2016), 12 pages. Google ScholarDigital Library
- Fabrice Rousselle, Claude Knaus, and Matthias Zwicker. 2011. Adaptive Sampling and Reconstruction Using Greedy Error Minimization. ACM Trans. Graph. 30, 6, Article 159 (Dec. 2011), 12 pages. Google ScholarDigital Library
- Vincent Sitzmann, Justus Thies, Felix Heide, Matthias Nießner, Gordon Wetzstein, and Michael Zollhöfer. 2018. DeepVoxels: Learning Persistent 3D Feature Embeddings. In CVPR.Google Scholar
- Charles Stein. 1972. A bound for the error in the normal approximation to the distribution of a sum of dependent random variables. In Proceedings of the Sixth Berkeley Symposium on Mathematical Statistics and Probability, Volume 2: Probability Theory. University of California Press, Berkeley, Calif., 583--602. https://projecteuclid.org/euclid.bsmsp/1200514239Google Scholar
- László Szécsi, Mateu Sbert, and László Szirmay-Kalos. 2004. Combined Correlated and Importance Sampling in Direct Light Source Computation and Environment Mapping. Computer Graphics Forum 23 (2004), 585--594.Google ScholarCross Ref
- László Szirmay-Kalos, Balázs Tóth, and Milán Magdics. 2011. Free Path Sampling in High Resolution Inhomogeneous Participating Media. Computer Graphics Forum 30, 1 (2011), 85--97.Google ScholarCross Ref
- Esteban Tabak and Cristina V. Turner. 2013. A Family of Nonparametric Density Estimation Algorithms. Communications on Pure and Applied Mathematics 66, 2 (2013), 145--164. Google ScholarCross Ref
- Esteban Tabak and Eric Vanden Eijnden. 2010. Density Estimation by dual Ascent of the Log-Likelihood. Communications in Mathematical Sciences 8, 1 (2010), 217--233.Google ScholarCross Ref
- Ayush Tewari, Ohad Fried, Justus Thies, Vincent Sitzmann, Stephen Lombardi, Kalyan Sunkavalli, Ricardo Martin-Brualla, Tomas Simon, Jason Saragih, Matthias Nießner, Rohit Pandey, Sean Fanello, Gordon Wetzstein, Jun-Yan Zhu, Christian Theobalt, Maneesh Agrawala, Eli Shechtman, Dan B Goldman, and Michael Zollhöfer. 2020. State of the Art on Neural Rendering. arXiv:cs.CV/2004.03805Google Scholar
- Justus Thies, Michael Zollhöfer, and Matthias Nießner. 2019. Deferred Neural Rendering: Image Synthesis Using Neural Textures. ACM Trans. Graph. 38, 4, Article 66 (July 2019), 12 pages. Google ScholarDigital Library
- Eric Veach. 1997. Robust Monte Carlo methods for light transport simulation. Ph.D. Dissertation. Stanford, CA, USA.Google ScholarDigital Library
- Eric Veach and Leonidas J. Guibas. 1995. Optimally Combining Sampling Techniques for Monte Carlo Rendering. In Proc. SIGGRAPH. 419--428. Google ScholarDigital Library
- Delio Vicini, Vladlen Koltun, and Wenzel Jakob. 2019. A Learned Shape-Adaptive Subsurface Scattering Model. ACM Trans. Graph. 38, 4, Article 127 (July 2019), 15 pages. Google ScholarDigital Library
- Marc Sabate Vidales, David Siska, and Lukasz Szpruch. 2018. Unbiased deep solvers for parametric PDEs. arXiv:1810.05094 (Oct. 2018).Google Scholar
- Jiří Vorba, Ondřej Karlík, Martin Šik, Tobias Ritschel, and Jaroslav Křivánek. 2014. On-line Learning of Parametric Mixture Models for Light Transport Simulation. ACM Trans. Graph. 33, 4 (Aug. 2014).Google ScholarDigital Library
- Jiří Vorba and Jaroslav Křivánek. 2016. Adjoint-Driven Russian Roulette and Splitting in Light Transport Simulation. ACM Trans. Graph. 35, 4 (jul 2016).Google ScholarDigital Library
- Ruosi Wan, Mingjun Zhong, Haoyi Xiong, and Zhanxing Zhu. 2019. Neural Control Variates for Variance Reduction. arXiv:1806.00159 (Oct. 2019).Google Scholar
- Tomoya Yamaguchi, Tatsuya Yatagawa, and Shigeo Morishima. 2018. Efficient Metropolis Path Sampling for Material Editing and Re-rendering. In Pacific Graphics Short Papers, Hongbo Fu, Abhijeet Ghosh, and Johannes Kopf (Eds.). The Eurographics Association. Google ScholarDigital Library
- Quan Zheng and Matthias Zwicker. 2019. Learning to Importance Sample in Primary Sample Space. Computer Graphics Forum 38, 2 (2019), 169--179. Google ScholarCross Ref
Index Terms
- Neural control variates
Recommendations
Neural Control Variates for Monte Carlo Variance Reduction
Machine Learning and Knowledge Discovery in DatabasesAbstractIn statistics and machine learning, approximation of an intractable integration is often achieved by using the unbiased Monte Carlo estimator, but the variances of the estimation are generally high in many applications. Control variates approaches ...
Combining antithetic variates and control variates in simulation experiments
Antithetic variates and control variates are two well-known variance reduction techniques. We consider combining antithetic variates and control variates to estimate the mean response in a stochastic simulation experiment. When applying antithetic ...
Control Variates for Probability and Quantile Estimation
In stochastic systems, quantiles indicate the level of system performance that can be delivered with a specified probability, while probabilities indicate the likelihood that a specified level of system performance can be achieved. We present new ...
Comments