A Note on Simulation-Based Inference by Matching Random Features
arXiv:2111.09220
Abstract
We can, and should, do statistical inference on simulation models by adjusting the parameters in the simulation so that the values of {\em randomly chosen} functions of the simulation output match the values of those same functions calculated on the data. Results from the "state-space reconstruction" or "geometry from a time series'' literature in nonlinear dynamics indicate that just such functions will typically suffice to identify a model with a -dimensional parameter space. Results from the "random features" literature in machine learning suggest that using random functions of the data can be an efficient replacement for using optimal functions. In this preliminary, proof-of-concept note, I sketch some of the key results, and present numerical evidence about the new method's properties. A separate, forthcoming manuscript will elaborate on theoretical and numerical details.
arXiv comment41 pages, 14 figures
References in corpus (7)
- The frontier of simulation-based inference
- The Randomized Dependence Coefficient
- Large-Scale Kernel Methods for Independence Testing
- On the Error of Random Fourier Features
- Random Feature Stein Discrepancies
- A framework for the calibration of social simulation models
- Minimax density estimation for growing dimension