# Adversarial Variational Optimization of Non-Differentiable Simulators

arXiv: https://arxiv.org/abs/1707.07113 · HTML: https://mlsanity.com/paper/1707.07113/

## Key facts

- Cited by 22 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 22 Jul 2017.

## Built on by (most cited first, 22 of 22)

- [DeepZero: Scaling up Zeroth-Order Optimization for Deep Model Training](https://mlsanity.com/paper/2310.02025/) (2023)
- [Learning Diffusion Priors from Observations by Expectation Maximization](https://mlsanity.com/paper/2405.13712/) (2024)
- [Compositional Score Modeling for Simulation-based Inference](https://mlsanity.com/paper/2209.14249/) (2022)
- [GATSBI: Generative Adversarial Training for Simulation-Based Inference](https://mlsanity.com/paper/2203.06481/) (2022)
- [Neural-Sim: Learning to Generate Training Data with NeRF](https://mlsanity.com/paper/2207.11368/) (2022)
- [Task2Sim : Towards Effective Pre-training and Transfer from Synthetic Data](https://mlsanity.com/paper/2112.00054/) (2021)
- [Mind Mappings: Enabling Efficient Algorithm-Accelerator Mapping Space Search](https://mlsanity.com/paper/2103.01489/) (2021)
- [Simulated Adversarial Testing of Face Recognition Models](https://mlsanity.com/paper/2106.04569/) (2021)
- [Sourcerer: Sample-based Maximum Entropy Source Distribution Estimation](https://mlsanity.com/paper/2402.07808/) (2024)
- [Finding Differences Between Transformers and ConvNets Using Counterfactual Simulation Testing](https://mlsanity.com/paper/2211.16499/) (2022)
- [Human Body Measurement Estimation with Adversarial Augmentation](https://mlsanity.com/paper/2210.05667/) (2022)
- [Multi-fidelity Constrained Optimization for Stochastic Black Box Simulators](https://mlsanity.com/paper/2311.15137/) (2023)
- [Predictive variational inference: Learn the predictively optimal posterior distribution](https://mlsanity.com/paper/2410.14843/) (2024)
- [Contrastive Normalizing Flows for Uncertainty-Aware Parameter Estimation](https://mlsanity.com/paper/2505.08709/) (2025)
- [Subspace-based Approximate Hessian Method for Zeroth-Order Optimization](https://mlsanity.com/paper/2507.06125/) (2025)
- [Goldilocks-curriculum Domain Randomization and Fractal Perlin Noise with Application to Sim2Real Pneumonia Lesion Detection](https://mlsanity.com/paper/2204.13849/) (2022)
- [Meta-simulation for the Automated Design of Synthetic Overhead Imagery](https://mlsanity.com/paper/2209.08685/) (2022)
- [End-to-End Mesh Optimization of a Hybrid Deep Learning Black-Box PDE Solver](https://mlsanity.com/paper/2404.11766/) (2024)
- [From concrete mixture to structural design -- a holistic optimization procedure in the presence of uncertainties](https://mlsanity.com/paper/2312.03607/) (2023)
- [Simulating, Fast and Slow: Learning Policies for Black-Box Optimization](https://mlsanity.com/paper/2406.04261/) (2024)
- [Distributional Sensitivity Analysis: Enabling Differentiability in Sample-Based Inference](https://mlsanity.com/paper/2508.09347/) (2025)
- [Correcting Sensor-Induced Distribution Drift with Wasserstein Adversarial Learning](https://mlsanity.com/paper/2606.18561/) (2026)

## Abstract (arXiv)

Complex computer simulators are increasingly used across fields of science as generative models tying parameters of an underlying theory to experimental observations. Inference in this setup is often difficult, as simulators rarely admit a tractable density or likelihood function. We introduce Adversarial Variational Optimization (AVO), a likelihood-free inference algorithm for fitting a non-differentiable generative model incorporating ideas from generative adversarial networks, variational optimization and empirical Bayes. We adapt the training procedure of generative adversarial networks by replacing the differentiable generative network with a domain-specific simulator. We solve the resulting non-differentiable minimax problem by minimizing variational upper bounds of the two adversarial objectives. Effectively, the procedure results in learning a proposal distribution over simulator parameters, such that the JS divergence between the marginal distribution of the synthetic data and the empirical distribution of observed data is minimized. We evaluate and compare the method with simulators producing both discrete and continuous data.