1: \begin{abstract}

2: Programs involving discontinuities introduced by control flow constructs such as conditional branches pose challenges to mathematical optimization methods that assume a degree of smoothness in the objective function's response surface.

3: Smooth interpretation (SI) is a form of abstract interpretation that approximates the convolution of a program's output with a Gaussian kernel, thus smoothing its output in a principled manner.

4: Here, we combine SI with automatic differentiation (AD) to efficiently compute gradients of smoothed programs.

5: In contrast to AD across a regular program execution, these gradients also capture the effects of alternative control flow paths.

6: The combination of SI with AD enables the direct gradient-based parameter synthesis for branching programs, allowing for instance the calibration of simulation models or their combination with neural network models in machine learning pipelines.

7: We detail the effects of the approximations made for tractability in SI and propose a novel Monte Carlo estimator that avoids the underlying assumptions by estimating the smoothed programs' gradients through a combination of AD and sampling.

8: Using DiscoGrad, our tool for automatically translating simple \cpp\ programs to a smooth differentiable form, we perform an extensive evaluation.

9: We compare the combination of SI with AD and our Monte Carlo estimator to existing gradient-free and stochastic methods on four non-trivial and originally discontinuous problems ranging from classical simulation-based optimization to neural network-driven control.

10: While the optimization progress with the SI-based estimator depends on the complexity of the programs' control flow, our Monte Carlo estimator is competitive in all problems, exhibiting the fastest convergence by a substantial margin in our highest-dimensional problem.

11: \end{abstract}

12: