AI Paper Atlas

Research map / Optimization / Optimizers

Stochastic Gradient Estimation for Higher-Order Differentiable Rendering

Zican Wang, Michael Fischer, Tobias Ritschel

ICCV 2025 · Optimization › Optimizers · approach: Stochastic gradient MCMC · 3 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use aggregate importance sampling for the Hessian-vector product (ablation)OurHVP · Texture (1024 parameters)relative convergence time: 4.00 -> 1.00Improved
use aggregate importance sampling for the Hessian-vector product (ablation)OurHVP · CNN5relative time to .9 image error: 3.64 -> 1.00Improved
use aggregate importance sampling for the Hessian-vector product (ablation)OurHVP · Mug (single parameter)relative time to .99 image error: 1.00 vs 1.14 (OurHVPA not faster)Hurt
use aggregate importance sampling (ablation)OurHVP · Box2 (2 dimensions)relative time to .99 image error: 0.64 vs 1.00Hurt
use HVP instead of full Hessian (ablation)OurH · Texturerelative time to .9 image error: 40.12 -> 4.01Improved
use HVP instead of full Hessian (ablation)OurH · Box10relative time to .99 image error: 5.33 -> 1.44Improved
use second-order Hessian estimate instead of first-order smoothed gradientOurG · Mugrelative time to .99 image error: 3.91 -> 1.00Improved
use our importance-sampled smoothed gradient instead of FR22FR22 · all tasksconvergence timeNo clear effect
replace gradient descent by higher-order optimization with OurHVPAgradient descent baselines · all taskswall-clock convergence speedup: 2.71x averageImproved
use the Mitsuba analytic gradientMitsuba · Mug, Shadow, Quad, Box, Texture, CNNconvergence: does not converge on plateausHurt
increase rendering samples 0.25x to 3xOurHVPA · Mugoptimization quality and convergenceNo clear effect
higher-order method on higher-dimensional problemOurHVPA · Texture and CNNconvergence: performance delta lessens with dimensionalityMixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy