1: \begin{abstract}

2: We derive methods to compute higher order differentials (\hahvp) of the rendering operator.

3: Our approach is based on importance sampling of a convolution that represents the differentials of rendering parameters and shows to be applicable to both rasterization and path tracing.

4: We further suggest an aggregate sampling strategy to importance-sample multiple dimensions of one convolution kernel simultaneously.

5: We demonstrate that this information improves convergence when used in higher-order optimizers such as Newton or Conjugate Gradient relative to a gradient descent baseline in several inverse rendering tasks. 

6: \end{abstract}

7: