Likelihood-free inference with emulator networks

By Jan-Matthis Lueckmann, Giacomo Bassetto, Theofanis Karaletsos, Jakob H. Macke. 2018

155 citations recorded by the source corpus.

Approximate Bayesian Computation (ABC) provides methods for Bayesian inference in simulation-based stochastic 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 high-dimensional problems which are challenging for conventional ABC approaches.

Subfields: Artificial Intelligence, Statistics and Probability, Probabilistic & Bayesian Methods

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Statistics and Probability

In Probabilistic & Bayesian Methods

Citations in the corpus

Likelihood-free MCMC with Approximate Likelihood Ratios (2019, methodology): “Other approaches attempt to learn approximate versions of the simulator [55] which is used to perform efﬁcient inference, adapting a similar strategy as in world models for reinforcement learning [56].”
Mining gold from implicit models to improve likelihood-free inference (2018, methodology): “They can be used to train a surrogate for the likelihood p(x|θ) [17, 19, 34] or, in a Bayesian setting, the posterior p(θ|x) [29, 33, 37].”
Fast likelihood-free cosmology with neural density estimators and active learning (2019, methodology): “The methodology described in this section is based on Papamakarios & Murray (2016), Papamakarios et al. (2018), Lueckmann et al. (2018) and Alsing et al. (2018b).”
Mining for Dark Matter Substructure: Inferring Subhalo Population Properties from Strong Lenses with Machine Learning (2019, background): “…et al. 2017; Louppe & Cranmer 2017; Lueckmann et al. 2017; Papamakarios et al. 2017; Tran et al. 2017; Chen et al. 2018; Dinev & Gutmann 2018; Grathwohl et al. 2018; Huang et al. 2018; Kingma & Dhariwal 2018; Lueckmann et al. 2018; Papamakarios et al. 2018; Alsing et al. 2019; Hermans et al. 2019).”
Sequential Neural Likelihood: Fast Likelihood-free Inference with Autoregressive Flows (2018, background): “Bayesian (cid:107) (cid:107) An active-learning-style approach is to use the uncertainty in the posterior estimate to choose what simulation to run next [34, 44]; in this scheme, the next simulation is chosen to maximally reduce the Bayesian uncertainty.”
Likelihood-free MCMC with Amortized Approximate Ratio Estimators (2019, background): “Emulator networks (Lueckmann et al., 2018) exploit the uncertainty within an ensemble to guide simulations.”
Automatic Posterior Transformation for Likelihood-Free Inference (2019, methodology): “A powerful recent approach, sequential neural likelihood (SNL), trains a neural conditional density estimator (e.g. masked ﬂow, Papamakarios et al., 2017) to estimate p ( x | θ ) for all x and θ (Papamakarios et al., 2018; Lueckmann et al., 2018).”
Sequential Neural Methods for Likelihood-free Inference (2018, methodology): “[14] proposed likelihood-free inference with emulator networks, a method which corresponds to SNL with an active learning component.”
Nuisance hardened data compression for fast likelihood-free inference (2019, methodology): “1 The parameter proposal may be adaptively modified to optimize acquisition of relevant simulated parameter-data pairs (Papamakarios et al. 2018), or Bayesian optimization-style acquisition rules may be adopted for even more optimal simulation acquisition (Lueckmann et al. 2018).”