• Corpus ID: 53037254

Efficient Bayesian Experimental Design for Implicit Models

@inproceedings{Kleinegesse2018EfficientBE,
  title={Efficient Bayesian Experimental Design for Implicit Models},
  author={Steven Kleinegesse and Michael U Gutmann},
  booktitle={International Conference on Artificial Intelligence and Statistics},
  year={2018}
}
Bayesian experimental design involves the optimal allocation of resources in an experiment, with the aim of optimising cost and performance. For implicit models, where the likelihood is intractable but sampling from the model is possible, this task is particularly difficult and therefore largely unexplored. This is mainly due to technical difficulties associated with approximating posterior distributions and utility functions. We devise a novel experimental design framework for implicit models… 
[PDF] Semantic Reader
35 Citations
Highly Influential Citations
4
Background Citations
17
Methods Citations
20

Figures from this paper

35 Citations

Bayesian Experimental Design for Implicit Models by Mutual Information Neural Estimation

It is shown that training a neural network to maximise a lower bound on MI allows us to jointly determine the optimal design and the posterior and gracefully extends Bayesian experimental design for implicit models to higher design dimensions.

A Scalable Gradient-Free Method for Bayesian Experimental Design with Implicit Models

This paper proposes a novel approach that leverages recent advances in stochastic approximate gradient ascent incorporated with a smoothed variational MI estimator for efficient and robust BED.

A Hybrid Gradient Method to Designing Bayesian Experiments for Implicit Models

This work proposes a hybrid gradient approach that leverages recent advances in variational MI estimator and evolution strategies combined with black-box stochastic gradient ascent (SGA) to maximize the MI lower bound.

Sequential Bayesian Experimental Design for Implicit Models via Mutual Information

This work devises a novel sequential design framework for parameter estimation that uses the Mutual Information between model parameters and simulated data as a utility function to find optimal experimental designs, which has not been done before for implicit models.

Gradient-based Bayesian Experimental Design for Implicit Models using Mutual Information Lower Bounds

A framework for Bayesian experimental design with implicit models, where the data-generating distribution is intractable but sampling from it is still possible, is introduced and a comprehensive empirical comparison of prominent lower bounds applied to the aforementioned tasks is provided.

Optimal Bayesian experimental design for subsurface flow

A novel approach for development of polynomial chaos expansion surrogate model for the design utility function by utilizing the orthogonality of PCE basis polynomials in order to replace the expensive integration over the space of possible observations by direct construction ofPCE approximation for the expected information gain.

Variational Estimators for Bayesian Optimal Experimental Design

This work introduces several classes of fast EIG estimators suited to the experiment design context by building on ideas from variational inference and mutual information estimation, showing theoretically and empirically that these estimators can provide significant gains in speed and accuracy over previous approaches.

Optimal Bayesian experimental design for subsurface flow problems

Approximate Bayesian Optimisation for Neural Networks

The use of approximate inference with Bayesian Neural Networks as an alternative to GPs to model distributions over functions is explored and a link between density-ratio estimation and class probability estimation based on approximate inference is provided.

A General Recipe for Likelihood-free Bayesian Optimization

FBO directly models the acquisition function without having to separately perform inference with a probabilistic surrogate model, and can be reduced to optimizing a weighted classification problem, where the weights corre-spond to the utility being chosen.

References

SHOWING 1-10 OF 32 REFERENCES

Bayesian Design of Experiments for Intractable Likelihood Models Using Coupled Auxiliary Models and Multivariate Emulation

A strategy is proposed that relies on the development of new methodology involving auxiliary modelling to approximate the expected utility, under an intractable likelihood model, applied with the latest approaches to maximising approximated expected utilities.

Bayesian Design of Experiments Using Approximate Coordinate Exchange

This work provides the most general solution to date for decision-theoretical Bayesian design through a novel approximate coordinate exchange algorithm that uses a Gaussian process emulator to approximate the expected utility as a function of a single design coordinate in a series of conditional optimization steps.

An induced natural selection heuristic for finding optimal Bayesian experimental designs

Simulation-Based Optimal Design

We review simulation based methods in optimal design. Expected utility maximization, i.e., optimal design, is concerned with maximizing an mtegra! expression representing expected utility with…

Likelihood-Free Inference by Ratio Estimation

An alternative inference approach that is as easy to use as synthetic likelihood but not as restricted in its assumptions, and that, in a natural way, enables automatic selection of relevant summary statistic from a large set of candidates is presented.

Likelihood-Free Extensions for Bayesian Sequentially Designed Experiments

In this work, likelihood-free extensions of the standard SMC algorithm are proposed and a specific simulation-based approximation of the likelihood known as the synthetic likelihood is investigated.

Optimal Bayesian design for discriminating between models with intractable likelihoods in epidemiology

Massive optimal data compression and density estimation for scalable, likelihood-free inference in cosmology

The first cosmological application of Density Estimation Likelihood-Free Inference (delfi), which learns a parametrized model for joint distribution of data and parameters, yielding both the parameter posterior and the model evidence, is presented.

Approximate Bayesian computation in population genetics.

A key advantage of the method is that the nuisance parameters are automatically integrated out in the simulation step, so that the large numbers of nuisance parameters that arise in population genetics problems can be handled without difficulty.

Optimal Observation Times in Experimental Epidemic Processes

It is shown that optimal designs derived by the Bayesian approach are similar for observational studies of a single epidemic and for studies involving replicated epidemics in independent subpopulations.