Using Gaussian Processes to Design Dynamic Experiments for Black-Box Model Discrimination under Uncertainty

Diverse domains of science and engineering use parameterised mechanistic models. Engineers and scientists can often hypothesise several rival models to explain a specific process or phenomenon. Consider a model discrimination setting where we wish to find the best mechanistic, dynamic model candidate and the best model parameter estimates. Typically, several rival mechanistic models can explain the available data, so design of dynamic experiments for model discrimination helps optimally collect additional data by finding experimental settings that maximise model prediction divergence. We argue there are two main approaches in the literature for solving the optimal design problem: (i) the analytical approach, using linear and Gaussian approximations to find closed-form expressions for the design objective, and (ii) the data-driven approach, which often relies on computationally intensive Monte Carlo techniques. Olofsson et al. (ICML 35, 2018) introduced Gaussian process (GP) surrogate models to hybridise the analytical and data-driven approaches, which allowed for computationally efficient design of experiments for discriminating between black-box models. In this study, we demonstrate that we can extend existing methods for optimal design of dynamic experiments to incorporate a wider range of problem uncertainty. We also extend the Olofsson et al. (2018) method of using GP surrogate models for discriminating between dynamic black-box models. We evaluate our approach on a well-known case study from literature, and explore the consequences of using GP surrogates to approximate gradient-based methods.

[1]  G. Altarelli The Higgs and the Excessive Success of the Standard Model , 2014, 1407.2122.

[2]  Giuseppe Carlo Calafiore,et al.  The scenario approach to robust control design , 2006, IEEE Transactions on Automatic Control.

[3]  Guido Buzzi-Ferraris,et al.  An improved version of a sequential design criterion for discriminating among rival multiresponse models , 1990 .

[4]  Dieter Fox,et al.  Gaussian Processes and Reinforcement Learning for Identification and Control of an Autonomous Blimp , 2007, Proceedings 2007 IEEE International Conference on Robotics and Automation.

[5]  Thomas B. Schön,et al.  Learning deep dynamical models from image pixels , 2014, ArXiv.

[6]  R. Wets,et al.  Stochastic programming , 1989 .

[7]  Ruth Misener,et al.  GPdoemd: a python package for design of experiments for model discrimination , 2018, Comput. Chem. Eng..

[8]  József Bokor,et al.  Analysis and Control of Nonlinear Process Systems , 2004 .

[9]  Carl E. Rasmussen,et al.  Gaussian processes for machine learning , 2005, Adaptive computation and machine learning.

[10]  Anthony N. Pettitt,et al.  A Review of Modern Computational Algorithms for Bayesian Optimal Design , 2016 .

[11]  Uwe D. Hanebeck,et al.  Analytic moment-based Gaussian process filtering , 2009, ICML '09.

[12]  William J. Hill,et al.  Discrimination Among Mechanistic Models , 1967 .

[13]  Alexander Mitsos,et al.  Optimal experimental design for optimal process design: A trilevel optimization formulation , 2019, AIChE Journal.

[14]  Dirk Lebiedz,et al.  A robust optimization approach to experimental design for model discrimination of dynamical systems , 2011, Math. Program..

[15]  W. Collins,et al.  Evaluation of climate models , 2013 .

[16]  Massimiliano Barolo,et al.  Online Model-Based Redesign of Experiments for Parameter Estimation in Dynamic Systems , 2009 .

[17]  Agathe Girard,et al.  Propagation of uncertainty in Bayesian kernel models - application to multiple-step ahead forecasting , 2003, 2003 IEEE International Conference on Acoustics, Speech, and Signal Processing, 2003. Proceedings. (ICASSP '03)..

[18]  Richard D. Braatz,et al.  Stochastic nonlinear model predictive control with probabilistic constraints , 2014, 2014 American Control Conference.

[19]  Uwe Naumann,et al.  The Art of Differentiating Computer Programs - An Introduction to Algorithmic Differentiation , 2012, Software, environments, tools.

[20]  David J. C. MacKay,et al.  Information-Based Objective Functions for Active Data Selection , 1992, Neural Computation.

[21]  Fabrizio Bezzo,et al.  A methodology for direct exploitation of available information in the online model-based redesign of experiments , 2016, Comput. Chem. Eng..

[22]  Alexander Mitsos,et al.  Optimization-based global structural identifiability , 2019, Comput. Chem. Eng..

[23]  Wolfgang Marquardt,et al.  Optimal Experimental Design for Discriminating Numerous Model Candidates: The AWDC Criterion , 2010 .

[24]  Dirk Lebiedz,et al.  An optimal experimental design approach to model discrimination in dynamic biochemical systems , 2010, Bioinform..

[25]  Jun Fu,et al.  Local optimization of dynamic programs with guaranteed satisfaction of path constraints , 2015, Autom..

[26]  L. Biegler,et al.  A Unified Framework for Kinetic Parameter Estimation Based on Spectroscopic Data with or without Unwanted Contributions , 2019, Industrial & Engineering Chemistry Research.

[27]  Jan Steensels,et al.  Improving industrial yeast strains: exploiting natural and artificial diversity , 2014, FEMS microbiology reviews.

[28]  Eric Walter,et al.  Optimal input design for model discrimination using Pontryagin's maximum principle: Application to kinetic model structures , 2014, Autom..

[29]  Carl E. Rasmussen,et al.  PILCO: A Model-Based and Data-Efficient Approach to Policy Search , 2011, ICML.

[30]  Marc Peter Deisenroth,et al.  Doubly Stochastic Variational Inference for Deep Gaussian Processes , 2017, NIPS.

[31]  Piotr Bania,et al.  Bayesian Input Design for Linear Dynamical Model Discrimination , 2019, Entropy.

[32]  Sandro Macchietto,et al.  The optimal design of dynamic experiments , 1989 .

[33]  Marc Peter Deisenroth,et al.  Design of Experiments for Model Discrimination Hybridising Analytical and Data-Driven Approaches , 2018, ICML.

[34]  Alexander Mitsos,et al.  Guaranteed satisfaction of inequality state constraints in PDE-constrained optimization , 2020, Autom..

[35]  E. Süli,et al.  Numerical Solution of Ordinary Differential Equations , 2021, Foundations of Space Dynamics.

[36]  Alexander Shapiro,et al.  Sample Average Approximation Method for Chance Constrained Programming: Theory and Applications , 2009, J. Optimization Theory and Applications.

[37]  Ian R. Manchester,et al.  Input design for model discrimination and fault detection via convex relaxation , 2013, 2014 American Control Conference.

[38]  Richard D. Braatz,et al.  Optimal Experimental Design for Probabilistic Model Discrimination Using Polynomial Chaos , 2014 .

[39]  D. Petranovic,et al.  Pharmaceutical protein production by yeast: towards production of human blood proteins by microbial fermentation. , 2012, Current opinion in biotechnology.

[40]  Marc Peter Deisenroth,et al.  Variational Integrator Networks for Physically Meaningful Embeddings , 2019, ArXiv.

[41]  William G. Hunter,et al.  Designs for Discriminating Between Two Rival Models , 1965 .

[42]  David J. C. MacKay,et al.  Information Theory, Inference, and Learning Algorithms , 2004, IEEE Transactions on Information Theory.

[43]  Yee Whye Teh,et al.  Variational Bayesian Optimal Experimental Design , 2019, NeurIPS.

[44]  Steven P. Asprey,et al.  On the design of optimally informative dynamic experiments for model discrimination in multiresponse nonlinear situations , 2003 .

[45]  Joel A. Paulson,et al.  Optimal Bayesian experiment design for nonlinear dynamic systems with chance constraints , 2019, Journal of Process Control.