Likelihood-free inference with emulator networks

Jan-Matthis Lueckmann, Giacomo Bassetto, Theofanis Karaletsos, Jakob H. Macke
; Proceedings of The 1st Symposium on Advances in Approximate Bayesian Inference, PMLR 96:32-53, 2019.

Abstract

Approximate Bayesian Computation (ABC) provides methods for Bayesian inference in simulation-based models which do not permit tractable likelihoods. We present a new ABC method which uses probabilistic neural emulator networks to learn synthetic likelihoods on simulated data - both ’local’ emulators which approximate the likelihood for specific observed data, as well as ’global’ ones which are applicable to a range of data. Simulations are chosen adaptively using an acquisition function which takes into account uncertainty about either the posterior distribution of interest, or the parameters of the emulator. Our approach does not rely on user-defined rejection thresholds or distance functions. We illustrate inference with emulator networks on synthetic examples and on a biophysical neuron model, and show that emulators allow accurate and efficient inference even on problems which are challenging for conventional ABC approaches.

Cite this Paper


BibTeX
@InProceedings{pmlr-v96-lueckmann19a, title = {Likelihood-free inference with emulator networks}, author = {Lueckmann, Jan-Matthis and Bassetto, Giacomo and Karaletsos, Theofanis and Macke, Jakob H.}, booktitle = {Proceedings of The 1st Symposium on Advances in Approximate Bayesian Inference}, pages = {32--53}, year = {2019}, editor = {Francisco Ruiz and Cheng Zhang and Dawen Liang and Thang Bui}, volume = {96}, series = {Proceedings of Machine Learning Research}, address = {}, month = {02 Dec}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v96/lueckmann19a/lueckmann19a.pdf}, url = {http://proceedings.mlr.press/v96/lueckmann19a.html}, abstract = {Approximate Bayesian Computation (ABC) provides methods for Bayesian inference in simulation-based models which do not permit tractable likelihoods. We present a new ABC method which uses probabilistic neural emulator networks to learn synthetic likelihoods on simulated data - both ’local’ emulators which approximate the likelihood for specific observed data, as well as ’global’ ones which are applicable to a range of data. Simulations are chosen adaptively using an acquisition function which takes into account uncertainty about either the posterior distribution of interest, or the parameters of the emulator. Our approach does not rely on user-defined rejection thresholds or distance functions. We illustrate inference with emulator networks on synthetic examples and on a biophysical neuron model, and show that emulators allow accurate and efficient inference even on problems which are challenging for conventional ABC approaches.} }
Endnote
%0 Conference Paper %T Likelihood-free inference with emulator networks %A Jan-Matthis Lueckmann %A Giacomo Bassetto %A Theofanis Karaletsos %A Jakob H. Macke %B Proceedings of The 1st Symposium on Advances in Approximate Bayesian Inference %C Proceedings of Machine Learning Research %D 2019 %E Francisco Ruiz %E Cheng Zhang %E Dawen Liang %E Thang Bui %F pmlr-v96-lueckmann19a %I PMLR %J Proceedings of Machine Learning Research %P 32--53 %U http://proceedings.mlr.press %V 96 %W PMLR %X Approximate Bayesian Computation (ABC) provides methods for Bayesian inference in simulation-based models which do not permit tractable likelihoods. We present a new ABC method which uses probabilistic neural emulator networks to learn synthetic likelihoods on simulated data - both ’local’ emulators which approximate the likelihood for specific observed data, as well as ’global’ ones which are applicable to a range of data. Simulations are chosen adaptively using an acquisition function which takes into account uncertainty about either the posterior distribution of interest, or the parameters of the emulator. Our approach does not rely on user-defined rejection thresholds or distance functions. We illustrate inference with emulator networks on synthetic examples and on a biophysical neuron model, and show that emulators allow accurate and efficient inference even on problems which are challenging for conventional ABC approaches.
APA
Lueckmann, J., Bassetto, G., Karaletsos, T. & Macke, J.H.. (2019). Likelihood-free inference with emulator networks. Proceedings of The 1st Symposium on Advances in Approximate Bayesian Inference, in PMLR 96:32-53

Related Material