Gaussian process enhanced semi-automatic approximate Bayesian computation: parameter inference in a stochastic differential equation system for chemotaxis

Borowska, Agnieszka; Giurghita, Diana; Husmeier, Dirk

doi:10.1016/j.jcp.2020.109999

Cited by 7 publications

(16 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In previous studies, the regression approach from Section 2.1.4 was used together with uniform, or on a previous run pre-calibrated, distance weights [Borowska et al, 2021, Fearnhead and Prangle, 2012, Jiang et al, 2017]. However, to the regression model outputs, approximating underlying parameters, the same problems apply that motivated the adaptive approach in Prangle [2017]: Parameters varying on larger scales dominate the analysis without scale adjustment, with potentially changing levels of variability over ABC-SMC generations.…”

Section: Methodsmentioning

confidence: 99%

“…via subset selection or auxiliary likelihoods [Drovandi et al, 2011, Nunes and Balding, 2010]. A popular line of approaches uses as statistics the outputs of inverse regression models of parameters on simulated data [Borowska et al, 2021, Fearnhead and Prangle, 2012, Jiang et al, 2017]. Such regression models can be heuristically motivated as summarizing the information in the data in a single value per parameter.…”

Section: Introductionmentioning

confidence: 99%

“…In step 3, Fearnhead and Prangle [2012] employ a linear regression (LR) model on potentially augmented data. Jiang et al [2017] and Borowska et al [2021] respectively use neural networks (NN) and Gaussian processes (GP) instead, aiming at a more accurate description of non-linear relationships, and further process automation. The sufficient performance of LR observed in Fearnhead and Prangle [2012] may be due to the substantial time spent in the pilot run, identifying a high-density region where a linear approximation suffices, while e.g.…”

Section: Introductionmentioning

confidence: 99%

“…Frequently, simply unit weights r = 1 are used (e.g. Borowska et al [2021], Fearnhead and Prangle [2012], Jiang et al [2017], Toni and Stumpf [2010]). However, model outputs can be and vary on different scales, in which case highly variable ones dominate the acceptance decision.…”

Section: Introductionmentioning

confidence: 99%

“…The sufficient performance of LR observed in Fearnhead and Prangle [2012] may be due to the substantial time spent in the pilot run, identifying a high-density region where a linear approximation suffices, while e.g. Jiang et al [2017] observe a clearly better posterior approximation with NN, and Borowska et al [2021] better model predictions with GP.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Informative and adaptive distances and summary statistics in sequential approximate Bayesian computation

Schälte

Hasenauer

2022

Preprint

View full text Add to dashboard Cite

Calibrating model parameters on heterogeneous data can be challenging and inefficient. This holds especially for likelihood-free methods such as approximate Bayesian computation (ABC), which rely on the comparison of relevant features in simulated and observed data and are popular for otherwise intractable problems. To address this problem, methods have been developed to scale-normalize data, and to derive informative low-dimensional summary statistics using inverse regression models of parameters on data. However, while approaches only correcting for scale can be inefficient on partly uninformative data, the use of summary statistics can lead to information loss and relies on the accuracy of employed methods. In this work, we first show that the combination of adaptive scale normalization with regression-based summary statistics is advantageous on heterogeneous parameter scales. Second, we present an approach employing regression models not to transform data, but to inform sensitivity weights quantifying data informativeness. Third, we discuss problems for regression models under non-identifiability, and present a solution using target augmentation. We demonstrate improved accuracy and efficiency of the presented approach on various problems, in particular robustness and wide applicability of the sensitivity weights. Our finding demonstrate the potential of the adaptive approach. The developed algorithms have been made available in the open-source Python toolbox pyABC.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Informative and adaptive distances and summary statistics in sequential approximate Bayesian computation

Schälte

Hasenauer

2022

Preprint

View full text Add to dashboard Cite

show abstract

Multifidelity multilevel Monte Carlo to accelerate approximate Bayesian parameter inference for partially observed stochastic processes

Warne

Prescott

Baker

et al. 2022

Journal of Computational Physics

View full text Add to dashboard Cite

Inferring the interaction rules of complex systems with graph neural networks and approximate Bayesian computation

Gaskell

Campioni

Morales

et al. 2023

J. R. Soc. Interface.

Self Cite

View full text Add to dashboard Cite

Inferring the underlying processes that drive collective behaviour in biological and social systems is a significant statistical and computational challenge. While simulation models have been successful in qualitatively capturing many of the phenomena observed in these systems in a variety of domains, formally fitting these models to data remains intractable. Recently, approximate Bayesian computation (ABC) has been shown to be an effective approach to inference if the likelihood function for a model is unavailable. However, a key difficulty in successfully implementing ABC lies with the design, selection and weighting of appropriate summary statistics, a challenge that is especially acute when modelling high dimensional complex systems. In this work, we combine a Gaussian process accelerated ABC method with the automatic learning of summary statistics via graph neural networks. Our approach bypasses the need to design a model-specific set of summary statistics for inference. Instead, we encode relational inductive biases into a neural network using a graph embedding and then extract summary statistics automatically from simulation data. To evaluate our framework, we use a model of collective animal movement as a test bed and compare our method to a standard summary statistics approach and a linear regression-based algorithm.

show abstract

Gaussian process enhanced semi-automatic approximate Bayesian computation: parameter inference in a stochastic differential equation system for chemotaxis

Cited by 7 publications

References 30 publications

Informative and adaptive distances and summary statistics in sequential approximate Bayesian computation

Informative and adaptive distances and summary statistics in sequential approximate Bayesian computation

Multifidelity multilevel Monte Carlo to accelerate approximate Bayesian parameter inference for partially observed stochastic processes

Inferring the interaction rules of complex systems with graph neural networks and approximate Bayesian computation

Contact Info

Product

Resources

About