improving synthetic likelihood

Chris Drovandi gave an after-dinner [QUT time!] talk for the One World ABC webinar on a recent paper he wrote with Jacob Proddle, Scott Sisson and David Frazier. Using a regular MCMC step on a synthetic likelihood approximation to the posterior. Or a (simulation based) unbiased estimator of it.

By evaluating the variance of the log-likelihood estimator, the authors show that the number of simulations n need scale like n²d² to keep the variance under control. And suggest PCA decorrelation of the summary statistic components as a mean to reduce the variance since it then scales as n²d. Rather idly, I wonder at the final relevance of precisely estimating the (synthetic) likelihood when considering it is not the true likelihood and when the n² part seems more damning. Moving from d² to d seems directly related to the estimation of a full correlation matrix for the Normal synthetic distribution of the summary statistic versus the estimation of a diagonal matrix. The usual complaint that performances highly depend on the choice of the summary statistic also applies here, in particular when its dimension is much larger than the dimension d of the parameter (as in the MA example). Although this does not seem to impact the scale of the variance.

focused Bayesian prediction

In this fourth session of our One World ABC Seminar, my friend and coauthor Gael Martin, gave an after-dinner talk on focused Bayesian prediction, more in the spirit of Bissiri et al. than following a traditional ABC approach.  because along with Ruben Loaiza-Maya and [my friend and coauthor] David Frazier, they consider the possibility of a (mild?) misspecification of the model. Using thus scoring rules à la Gneiting and Raftery. Gael had in fact presented an earlier version at our workshop in Oaxaca, in November 2018. As in other solutions of that kind, difficulty in weighting the score into a distribution. Although asymptotic irrelevance, direct impact on the current predictions, at least for the early dates in the time series… Further calibration of the set of interest A. Or the focus of the prediction. As a side note the talk perfectly fits the One World likelihood-free seminar as it does not use the likelihood function!

“The very premise of this paper is that, in reality, any choice of predictive class is such that the truth is not contained therein, at which point there is no reason to presume that the expectation of any particular scoring rule will be maximized at the truth or, indeed, maximized by the same predictive distribution that maximizes a different (expected) score.”

This approach requires the proxy class to be close enough to the true data generating model. Or in the word of the authors to be plausible predictive models. And to produce the true distribution via the score as it is proper. Or the closest to the true model in the misspecified family. I thus wonder at a possible extension with a non-parametric version, the prior being thus on functionals rather than parameters, if I understand properly the meaning of Π(P_θ). (Could the score function be misspecified itself?!) Since the score is replaced with its empirical version, the implementation is  resorting to off-the-shelf MCMC. (I wonder for a few seconds if the approach could be seen as a pseudo-marginal MCMC but the estimation is always based on the same observed sample, hence does not directly fit the pseudo-marginal MCMC framework.)

[Notice: Next talk in the series is tomorrow, 11:30am GMT+1.]

stratified ABC [One World ABC webinar]

The third episode of the One World ABC seminar (Season 1!) was kindly delivered by Umberto Picchini on Stratified sampling and bootstrapping for ABC which I already if briefly discussed after BayesComp 2020. Which sounds like a million years ago… His introduction on the importance of estimating the likelihood using a kernel, while 600% justified wrt his talk, made the One World ABC seminar sounds almost like groundhog day!  The central argument is in the computational gain brought by simulating a single θ dependent [expensive] dataset followed by [cheaper] bootstrap replicates. Which turns de fact into bootstrapping the summary statistics.

If I understand correctly, the post-stratification approach of Art Owen (2013?, I cannot find the reference) corrects a misrepresentation of mine. Indeed, defining a partition with unknown probability weights seemed to me to annihilate the appeal of stratification, because the Bernoulli variance of the estimated probabilities brought back the same variability as the mother estimator. But with bootstrap, this requires only two simulations, one for the weights and one for the target. And further allows for a larger ABC tolerance in fine. Free lunch?!

The speaker in two weeks (21 May or Ascension Thursday!) is my friend and co-author Gael Martin from Monash University, who will speak on Focused Bayesian prediction, at quite a late time down under..!

rare ABC [webinar impressions]

A second occurrence of the One World ABC seminar by Ivis Kerama, and Richard Everitt (Warwick U), on their on-going pape with and Tom Thorne, Rare Event ABC-SMC², which is not about rare event simulation but truly about ABC improvement. Building upon a previous paper by Prangle et al. (2018). And also connected with Dennis’ talk a fortnight ago in that it exploits an autoencoder representation of the simulated outcome being H(u,θ). It also reminded me of an earlier talk by Nicolas Chopin.

This approach avoids using summary statistics (but relies on a particular distance) and implements a biased sampling of the u’s to produce outcomes more suited to the observation(s). Almost sounds like a fiducial ABC! Their stopping rule for decreasing the tolerance is to spot an increase in the variance of the likelihood estimates. As the method requires many data generations for a single θ, it only applies in certain settings. The ABC approximation is indeed used as an estimation of likelihood ratio (which makes sense for SMC² but is biased because of ABC). I got slightly confused during Richard’s talk by his using the term of unbiased estimator of the likelihood before I realised he was talking of the ABC posterior. Thanks to both speakers, looking forward the talk by Umberto Picchini in a fortnight (on a joint paper with Richard).

One World webinars

Just a notice that our ABC World seminar has joined the “franchise” of the One World seminars

• One World Probability Seminar
• One World PDE Seminar
• Waves in One World
• One World Mathematical Game Theory Seminar
• One World Seminar: Mathematical Methods for Arbitrary Data Sources
• One World Cognitive Psychologie Seminar
• One World IMAGing and INvErse problems (IMAGINE) Seminar
• One World Optimization Seminar
• One World Actuarial Research Seminar
• One World Mathematics of INformation, Data, and Signals (MINDS) Seminar

and that, on Thursday, 23 April, at 12:30 (CEST) IIvis Kerama and Richard Everitt will talk on  Rare event ABC-SMC², while, also on Thursday, at 16:00 (CEST) Michela Ottobre will talk Fast non mean-field network: uniform in time averaging in the One World Probability Seminar.