Novel and flexible parameter estimation methods for data-consistent inversion in mechanistic modeling
This work addresses parameter estimation bias in mechanistic modeling for fields like biological sciences, offering incremental improvements to existing methodologies.
The paper tackles bias from uninformative priors in Bayesian parameter estimation for mechanistic models by proposing methods within the stochastic inverse problem framework, introducing rejection sampling, MCMC, and GAN-based approaches, and reformulating SIP with constrained optimization and a novel GAN to address limitations.
Predictions for physical systems often rely upon knowledge acquired from ensembles of entities, e.g., ensembles of cells in biological sciences. For qualitative and quantitative analysis, these ensembles are simulated with parametric families of mechanistic models (MM). Two classes of methodologies, based on Bayesian inference and Population of Models, currently prevail in parameter estimation for physical systems. However, in Bayesian analysis, uninformative priors for MM parameters introduce undesirable bias. Here, we propose how to infer parameters within the framework of stochastic inverse problems (SIP), also termed data-consistent inversion, wherein the prior targets only uncertainties that arise due to MM non-invertibility. To demonstrate, we introduce new methods to solve SIP based on rejection sampling, Markov chain Monte Carlo, and generative adversarial networks (GANs). In addition, to overcome limitations of SIP, we reformulate SIP based on constrained optimization and present a novel GAN to solve the constrained optimization problem.