Elves to the ABC rescue!

Marko Järvenpää, Michael Gutmann, Arijus Pleska, Aki Vehtari, and Pekka Marttinen have written a paper on Efficient Acquisition Rules for Model-Based Approximate Bayesian Computation soon to appear in Bayesian Analysis that gives me the right nudge to mention the ELFI software they have been contributing to for a while. Where the acronym stands for engine for likelihood-free inference. Written in Python, DAG based, and covering methods like the

• ABC rejection sampler
• Sequential Monte Carlo ABC sampler
• Bayesian Optimization for Likelihood-Free Inference (BOLFI) framework
• Bayesian Optimization (not likelihood-free)
• No-U-Turn-Sampler (not likelihood-free)

[Warning: I did not experiment with the software! Feel free to share.]

“…little work has focused on trying to quantify the amount of uncertainty in the estimator of the ABC posterior density under the chosen modelling assumptions. This uncertainty is due to a finite computational budget to perform the inference and could be thus also called as computational uncertainty.”

The paper is about looking at the “real” ABC distribution, that is, the one resulting from a realistic perspective of a finite number of simulations and acceptances. By acquisition, the authors mean an efficient way to propose the next value of the parameter θ, towards minimising the uncertainty in the ABC density estimate. Note that this involves a loss function that must be chosen by the analyst and then available for the minimisation program. If this sounds complicated…

“…our interest is to design the evaluations to minimise the uncertainty in a quantity that itself describes the uncertainty of the parameters of a costly simulation model.”

it indeed is and it requires modelling choices. As in Guttman and Corander (2016), which was also concerned by designing the location of the learning parameters, the modelling is based here on a Gaussian process for the discrepancy between the observed and the simulated data. Which provides an estimate of the likelihood, later used for selecting the next sampling value of θ. The final ABC sample is however produced by a GP estimation of the ABC distribution.As noted by the authors, the method may prove quite time consuming: for instance, one involved model required one minute of computation time for selecting the next evaluation location. (I had a bit of a difficulty when reading the paper as I kept hitting notions that are local to the paper but not immediately or precisely defined. As “adequation function” [p.11] or “discrepancy”. Maybe correlated with short nights while staying at CIRM for the Masterclass, always waking up around 4am for unknown reasons!)

a jump back in time

As the Department of Statistics in Warwick is slowly emptying its shelves and offices for the big migration to the new building that is almost completed, books and documents are abandoned in the corridors and the work spaces. On this occasion, I thus happened to spot a vintage edition of the Valencia 3 proceedings. I had missed this meeting and hence the volume for, during the last year of my PhD, I was drafted in the French Navy and as a result prohibited to travel abroad. (Although on reflection I could have safely done it with no one in the military the wiser!) Reading through the papers thirty years later is a weird experience, as I do not remember most of the papers, the exception being the mixture modelling paper by José Bernardo and Javier Giròn which I studied a few years later when writing the mixture estimation and simulation paper with Jean Diebolt. And then again in our much more recent non-informative paper with Clara Grazian.  And Prem Goel’s survey of Bayesian software. That is, 1987 state of the art software. Covering an amazing eighteen list. Including versions by Zellner, Tierney, Schervish, Smith [but no MCMC], Jaynes, Goldstein, Geweke, van Dijk, Bauwens, which apparently did not survive the ages till now. Most were in Fortran but S was also mentioned. And another version of Tierney, Kass and Kadane on Laplace approximations. And the reference paper of Dennis Lindley [who was already retired from UCL at that time!] on the Hardy-Weinberg equilibrium. And another paper by Don Rubin on using SIR (Rubin, 1983) for simulating from posterior distributions with missing data. Ten years before the particle filter paper, and apparently missing the possibility of weights with infinite variance.

There already were some illustrations of Bayesian analysis in action, including one by Jay Kadane reproduced in his book. And several papers by Jim Berger, Tony O’Hagan, Luis Pericchi and others on imprecise Bayesian modelling, which was in tune with the era, the imprecise probability book by Peter Walley about to appear. And a paper by Shaw on numerical integration that mentioned quasi-random methods. Applied to a 12 component Normal mixture.Overall, a much less theoretical content than I would have expected. And nothing about shrinkage estimators, although a fraction of the speakers had worked on this topic most recently.

At a less fundamental level, this was a time when LaTeX was becoming a standard, as shown by a few papers in the volume (and as I was to find when visiting Purdue the year after), even though most were still typed on a typewriter, including a manuscript addition by Dennis Lindley. And Warwick appeared as a Bayesian hotpot!, with at least five papers written by people there permanently or on a long term visit. (In case a local is interested in it, I have kept the volume, to be found in my new office!)

Bayesian workers, unite!

This afternoon, Alexander Ly is defending his PhD thesis at the University of Amsterdam. While I cannot attend the event, I want to celebrate the event and a remarkable thesis around the Bayes factor [even though we disagree on its role!] and the Jeffreys’s Amazing Statistics Program (!), otherwise known as JASP. Plus commend the coolest thesis cover I ever saw, made by the JASP graphical designer Viktor Beekman and representing Harold Jeffreys leading empirical science workers in the best tradition of socialist realism! Alexander wrote a post on the JASP blog to describe the thesis, the cover, and his plans for the future. Congratulations!

To predict and serve?

Kristian Lum and William Isaac published a paper in Significance last week [with the above title] about predictive policing systems used in the USA and presumably in other countries to predict future crimes [and therefore prevent them]. This sounds like a good idea for a science fiction plot, à la Philip K Dick [in his short story, The Minority Report], but that it is used in real life definitely sounds frightening, especially when the civil rights of the targeted individuals are impacted. (Although some politicians in different democratic countries increasingly show increasing contempt for keeping everyone’ rights equal…) I also feel terrified by the social determinism behind the very concept of predicting crime from socio-economic data (and possibly genetic characteristics in a near future, bringing us back to the dark days of physiognomy!)

“…crimes that occur in locations frequented by police are more likely to appear in the database simply because that is where the police are patrolling.”

Kristian and William examine in this paper one statistical aspect of the police forces relying on crime prediction software, namely the bias in the data exploited by the software and in the resulting policing. (While the accountability of the police actions when induced by such software is not explored, this is obviously related to the Nature editorial of last week, “Algorithm and blues“, which [in short] calls for watchdogs on AIs and decision algorithms.) When the data is gathered from police and justice records, any bias in checks, arrests, and condemnations will be reproduced in the data and hence will repeat the bias in targeting potential criminals. As aptly put by the authors, the resulting machine learning algorithm will be “predicting future policing, not future crime.” Worse, by having no reservation about over-fitting [the more predicted crimes the better], it will increase the bias in the same direction. In the Oakland drug-user example analysed in the article, the police concentrates almost uniquely on a few grid squares of the city, resulting into the above self-predicting fallacy. However, I do not see much hope in using other surveys and datasets towards eliminating this bias, as they also carry their own shortcomings. Even without biases, predicting crimes at the individual level just seems a bad idea, for statistical and ethical reasons.

Journal of Open Source Software

A week ago, I received a request for refereeing a paper for the Journal of Open Source Software, which I have never seen (or heard of) before. The concept is quite interesting with a scope much broader than statistical computing (as I do not know anyone in the board and no-one there seems affiliated with a Statistics department). Papers are very terse, describing the associated code in one page or two, and the purpose of refereeing is to check the code. (I was asked to evaluate an MCMC R package but declined for lack of time.) Which is a pretty light task if the code is friendly enough to operate right away and provide demos. Best of luck to this endeavour!

new reproducibility initiative in TOMACS

[A quite significant announcement last October from TOMACS that I had missed:]

To improve the reproducibility of modeling and simulation research, TOMACS  is pursuing two strategies.

Number one: authors are encouraged to include sufficient information about the core steps of the scientific process leading to the presented research results and to make as many of these steps as transparent as possible, e.g., data, model, experiment settings, incl. methods and configurations, and/or software. Associate editors and reviewers will be asked to assess the paper also with respect to this information. Thus, although not required, submitted manuscripts which provide clear information on how to generate reproducible results, whenever possible, will be considered favorably in the decision process by reviewers and the editors.

Number two: we will form a new replicating computational results activity in modeling and simulation as part of the peer reviewing process (adopting the procedure RCR of ACM TOMS). Authors who are interested in taking part in the RCR activity should announce this in the cover letter. The associate editor and editor in chief will assign a RCR reviewer for this submission. This reviewer will contact the authors and will work together with the authors to replicate the research results presented. Accepted papers that successfully undergo this procedure will be advertised at the TOMACS web page and will be marked with an ACM reproducibility brand. The RCR activity will take place in parallel to the usual reviewing process. The reviewer will write a short report which will be published alongside the original publication. TOMACS also plans to publish short reports about lessons learned from non-successful RCR activities.

[And now the first paper reviewed according to this protocol has been accepted:]

The paper Automatic Moment-Closure Approximation of Spatially Distributed Collective Adaptive Systems is the first paper that took part in the new replicating computational results (RCR) activity of TOMACS. The paper completed successfully the additional reviewing as documented in its RCR report. This reviewing is aimed at ensuring that computational results presented in the paper are replicable. Digital artifacts like software, mechanized proofs, data sets, test suites, or models, are evaluated referring to ease of use, consistency, completeness, and being well documented.

MCMSki IV [day 3]

Already on the final day..! And still this frustration in being unable to attend three sessions at once… Andrew Gelman started the day with a non-computational talk that broached on themes that are familiar to readers of his blog, on the misuse of significance tests and on recommendations for better practice. I then picked the Scaling and optimisation of MCMC algorithms session organised by Gareth Roberts, with optimal scaling talks by Tony Lelièvre, Alex Théry and Chris Sherlock, while Jochen Voss spoke about the convergence rate of ABC, a paper I already discussed on the blog. A fairly exciting session showing that MCMC’ory (name of a workshop I ran in Paris in the late 90’s!) is still well and alive!

After the break (sadly without the ski race!), the software round-table session was something I was looking for. The four softwares covered by this round-table were BUGS, JAGS, STAN, and BiiPS, each presented according to the same pattern. I would have like to see a “battle of the bands”, illustrating pros & cons for each language on a couple of models & datasets. STAN got the officious prize for cool tee-shirts (we should have asked the STAN team for poster prize tee-shirts). And I had to skip the final session for a flu-related doctor appointment…

I called for a BayesComp meeting at 7:30, hoping for current and future members to show up and discuss the format of the future MCMski meetings, maybe even proposing new locations on other “sides of the Italian Alps”! But (workshop fatigue syndrome?!), no-one showed up. So anyone interested in discussing this issue is welcome to contact me or David van Dyk, the new BayesComp program chair.