Archive for noninformative priors

a case for Bayesian deep learnin

Posted in Books, pictures, Statistics, Travel, University life with tags , , , , , , , , , , on September 30, 2020 by xi'an

Andrew Wilson wrote a piece about Bayesian deep learning last winter. Which I just read. It starts with the (posterior) predictive distribution being the core of Bayesian model evaluation or of model (epistemic) uncertainty.

“On the other hand, a flat prior may have a major effect on marginalization.”

Interesting sentence, as, from my viewpoint, using a flat prior is a no-no when running model evaluation since the marginal likelihood (or evidence) is no longer a probability density. (Check Lindley-Jeffreys’ paradox in this tribune.) The author then goes for an argument in favour of a Bayesian approach to deep neural networks for the reason that data cannot be informative on every parameter in the network, which should then be integrated out wrt a prior. He also draws a parallel between deep ensemble learning, where random initialisations produce different fits, with posterior distributions, although the equivalent to the prior distribution in an optimisation exercise is somewhat vague.

“…we do not need samples from a posterior, or even a faithful approximation to the posterior. We need to evaluate the posterior in places that will make the greatest contributions to the [posterior predictive].”

The paper also contains an interesting point distinguishing between priors over parameters and priors over functions, ony the later mattering for prediction. Which must be structured enough to compensate for the lack of data information about most aspects of the functions. The paper further discusses uninformative priors (over the parameters) in the O’Bayes sense as a default way to select priors. It is however unclear to me how this discussion accounts for the problems met in high dimensions by standard uninformative solutions. More aggressively penalising priors may be needed, as those found in high dimension variable selection. As in e.g. the 10⁷ dimensional space mentioned in the paper. Interesting read all in all!

how can a posterior be uniform?

Posted in Books, Statistics with tags , , , , , , on September 1, 2020 by xi'an

A bemusing question from X validated:

How can we have a posterior distribution that is a uniform distribution?

With the underlying message that a uniform distribution does not depend on the data, since it is uniform! While it is always possible to pick the parameterisation a posteriori so that the posterior is uniform, by simply using the inverse cdf transform, or to pick the prior a posteriori so that the prior cancels the likelihood function, there exist more authentic discrete examples of a data realisation leading to a uniform distribution, as eg in the Multinomial model. I deem the confusion to stem from the impression either that uniform means non-informative (what we could dub Laplace’s daemon!) or that it could remain uniform for all realisations of the sampled rv.

noninformative Bayesian prior with a finite support

Posted in Statistics, University life with tags , , , , , , on December 4, 2018 by xi'an

A few days ago, Pierre Jacob pointed me to a PNAS paper published earlier this year on a form of noninformative Bayesian analysis by Henri Mattingly and coauthors. They consider a prior that “maximizes the mutual information between parameters and predictions”, which sounds very much like José Bernardo’s notion of reference priors. With the rather strange twist of having the prior depending on the data size m even they work under an iid assumption. Here information is defined as the difference between the entropy of the prior and the conditional entropy which is not precisely defined in the paper but looks like the expected [in the data x] Kullback-Leibler divergence between prior and posterior. (I have general issues with the paper in that I often find it hard to read for a lack of precision and of definition of the main notions.)

One highly specific (and puzzling to me) feature of the proposed priors is that they are supported by a finite number of atoms, which reminds me very much of the (minimax) least favourable priors over compact parameter spaces, as for instance in the iconic paper by Casella and Strawderman (1984). For the same mathematical reason that non-constant analytic functions must have separated maxima. This is conducted under the assumption and restriction of a compact parameter space, which must be chosen in most cases. somewhat arbitrarily and not without consequences. I can somehow relate to the notion that a finite support prior translates the limited precision in the estimation brought by a finite sample. In other words, given a sample size of m, there is a maximal precision one can hope for, producing further decimals being silly. Still, the fact that the support of the prior is fixed a priori, completely independently of the data, is both unavoidable (for the prior to be prior!) and very dependent on the choice of the compact set. I would certainly prefer to see a maximal degree of precision expressed a posteriori, meaning that the support would then depend on the data. And handling finite support posteriors is rather awkward in that many notions like confidence intervals do not make much sense in that setup. (Similarly, one could argue that Bayesian non-parametric procedures lead to estimates with a finite number of support points but these are determined based on the data, not a priori.)

Interestingly, the derivation of the “optimal” prior is operated by iterations where the next prior is the renormalised version of the current prior times the exponentiated Kullback-Leibler divergence, which is “guaranteed to converge to the global maximum” for a discretised parameter space. The authors acknowledge that the resolution is poorly suited to multidimensional settings and hence to complex models, and indeed the paper only covers a few toy examples of moderate and even humble dimensions.

Another difficulty with the paper is the absence of temporal consistency: since the prior depends on the sample size, the posterior for n i.i.d. observations is no longer the prior for the (n+1)th observation.

“Because it weights the irrelevant parameter volume, the Jeffreys prior has strong dependence on microscopic effects invisible to experiment”

I simply do not understand the above sentence that apparently counts as a criticism of Jeffreys (1939). And would appreciate anyone enlightening me! The paper goes into comparing priors through Bayes factors, which ignores the main difficulty of an automated solution such as Jeffreys priors in its inability to handle infinite parameter spaces by being almost invariably improper.

visual effects

Posted in Books, pictures, Statistics with tags , , , , , , , , , , , on November 2, 2018 by xi'an

As advertised and re-discussed by Dan Simpson on the Statistical Modeling, &tc. blog he shares with Andrew and a few others, the paper Visualization in Bayesian workflow he wrote with Jonah Gabry, Aki Vehtari, Michael Betancourt and Andrew Gelman was one of three discussed at the RSS conference in Cardiff, last week month, as a Read Paper for Series A. I had stored the paper when it came out towards reading and discussing it, but as often this good intention led to no concrete ending. [Except concrete as in concrete shoes…] Hence a few notes rather than a discussion in Series B A.

Exploratory data analysis goes beyond just plotting the data, which should sound reasonable to all modeling readers.

Fake data [not fake news!] can be almost [more!] as valuable as real data for building your model, oh yes!, this is the message I am always trying to convey to my first year students, when arguing about the connection between models and simulation, as well as a defense of ABC methods. And more globally of the very idea of statistical modelling. While indeed “Bayesian models with proper priors are generative models”, I am not particularly fan of using the prior predictive [or the evidence] to assess the prior as it may end up in a classification of more or less all but terrible priors, meaning that all give very little weight to neighbourhoods of high likelihood values. Still, in a discussion of a TAS paper by Seaman et al. on the role of prior, Kaniav Kamary and I produced prior assessments that were similar to the comparison illustrated in Figure 4. (And this makes me wondering which point we missed in this discussion, according to Dan.)  Unhappy am I with the weakly informative prior illustration (and concept) as the amount of fudging and calibrating to move from the immensely vague choice of N(0,100) to the fairly tight choice of N(0,1) or N(1,1) is not provided. The paper reads like these priors were the obvious and first choice of the authors. I completely agree with the warning that “the utility of the the prior predictive distribution to evaluate the model does not extend to utility in selecting between models”.

MCMC diagnostics, beyond trace plots, yes again, but this recommendation sounds a wee bit outdated. (As our 1998 reviewww!) Figure 5(b) links different parameters of the model with lines, which does not clearly relate to a better understanding of convergence. Figure 5(a) does not tell much either since the green (divergent) dots stand within the black dots, at least in the projected 2D plot (and how can one reach beyond 2D?) Feels like I need to rtfm..!

“Posterior predictive checks are vital for model evaluation”, to wit that I find Figure 6 much more to my liking and closer to my practice. There could have been a reference to Ratmann et al. for ABC where graphical measures of discrepancy were used in conjunction with ABC output as direct tools for model assessment and comparison. Essentially predicting a zero error with the ABC posterior predictive. And of course “posterior predictive checking makes use of the data twice, once for the fitting and once for the checking.” Which means one should either resort to loo solutions (as mentioned in the paper) or call for calibration of the double-use by re-simulating pseudo-datasets from the posterior predictive. I find the suggestion that “it is a good idea to choose statistics that are orthogonal to the model parameters” somewhat antiquated, in that this sounds like rephrasing the primeval call to ancillary statistics for model assessment (Kiefer, 1975), while pretty hard to implement in modern complex models.

a jump back in time

Posted in Books, Kids, Statistics, Travel, University life with tags , , , , , , , , , , , on October 1, 2018 by xi'an

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!)