# Neural Path Guiding with Distribution Factorization

arXiv: https://arxiv.org/abs/2506.00839 · HTML: https://mlsanity.com/paper/2506.00839/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 14 of its 52 references that are in the map.
- First posted on arXiv 1 Jun 2025.

## Builds on (most cited first, 14 of 14)

- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [Density estimation using Real NVP](https://mlsanity.com/paper/1605.08803/) (2016)
- [Neural Importance Sampling](https://mlsanity.com/paper/1808.03856/) (2018)
- [Real-time Neural Radiance Caching for Path Tracing](https://mlsanity.com/paper/2106.12372/) (2021)
- [Discretizing Continuous Action Space for On-Policy Optimization](https://mlsanity.com/paper/1901.10500/) (2019)
- [Learning to Importance Sample in Primary Sample Space](https://mlsanity.com/paper/1808.07840/) (2018)
- [Neural Parametric Mixtures for Path Guiding](https://mlsanity.com/paper/2504.04315/) (2025)
- [Path Guiding Using Spatio-Directional Mixture Models](https://mlsanity.com/paper/2111.13094/) (2021)
- [Online Neural Path Guiding with Normalized Anisotropic Spherical Gaussians](https://mlsanity.com/paper/2303.08064/) (2023)
- [Learning Light Transport the Reinforced Way](https://mlsanity.com/paper/1701.07403/) (2017)
- [Online path sampling control with progressive spatio-temporal filtering](https://mlsanity.com/paper/2005.07547/) (2020)
- [Neural Product Importance Sampling via Warp Composition](https://mlsanity.com/paper/2409.18974/) (2024)
- [Second-Order Occlusion-Aware Volumetric Radiance Caching](https://mlsanity.com/paper/1805.09258/) (2018)
- [Manifold Path Guiding for Importance Sampling Specular Chains](https://mlsanity.com/paper/2311.12818/) (2023)

## Built on by (most cited first, 1 of 1)

- [PureLight: Learning Complex Luminaires with Light Tracing](https://mlsanity.com/paper/2606.04319/) (2026)

## Abstract (arXiv)

In this paper, we present a neural path guiding method to aid with Monte Carlo (MC) integration in rendering. Existing neural methods utilize distribution representations that are either fast or expressive, but not both. We propose a simple, but effective, representation that is sufficiently expressive and reasonably fast. Specifically, we break down the 2D distribution over the directional domain into two 1D probability distribution functions (PDF). We propose to model each 1D PDF using a neural network that estimates the distribution at a set of discrete coordinates. The PDF at an arbitrary location can then be evaluated and sampled through interpolation. To train the network, we maximize the similarity of the learned and target distributions. To reduce the variance of the gradient during optimizations and estimate the normalization factor, we propose to cache the incoming radiance using an additional network. Through extensive experiments, we demonstrate that our approach is better than the existing methods, particularly in challenging scenes with complex light transport.