Stochastic Graph Compression for Constant-Memory Differentiable Light Tracing

2026-10-07

AI Summary

This study addresses the memory explosion problem in differentiable ray tracing, where automatic differentiation computation graphs grow with the number of sensor connections. To overcome this limitation, we propose the ResLRB algorithm, which introduces streaming weighted reservoir sampling to stochastically compress the adjoint graph, thereby transcending the constraints of non-branching structures. By integrating reverse-mode automatic differentiation with pseudorandom sequence replay, our method achieves constant-memory backpropagation independent of path length and connection count. The primary contribution of this work is an unbiased gradient estimation framework that operates under a peak memory constraint equivalent to a single scattering event. This advancement renders large-scale differentiable rendering practically feasible under constant memory conditions.

Abstract

We describe Reservoir Light Replay Backpropagation (ResLRB), a method for reverse-mode differentiable light tracing whose memory is constant in path length and in the number of sensor connections per light path. Naive automatic differentiation of a light tracer records a computation graph that grows with the number of valid sensor connections along each path, since flux can be splatted from every scattering vertex; adjoint path replay removes the analogous dependence on depth for viewpoint path tracing, but its non-branching structure does not extend to the splatting case. We compress the adjoint graph during the primal pass by stochastically retaining a single representative sensor connection per light path via a streaming weighted reservoir. The adjoint pass reconstructs the path deterministically by replaying its pseudorandom number sequence and backpropagates only through the retained connection, yielding an unbiased gradient estimator whose peak memory is that of differentiating a single scattering event.

Problem

• differentiable light tracing
• memory efficiency
• automatic differentiation
• computation graph
• adjoint path replay

Innovation

• Differentiable light tracing
• Stochastic graph compression
• Reservoir sampling
• Constant memory
• Unbiased gradient estimator