[more than] everything you always wanted to know about marginal likelihood

Earlier this year, F. Llorente, L. Martino, D. Delgado, and J. Lopez-Santiago have arXived an updated version of their massive survey on marginal likelihood computation. Which I can only warmly recommend to anyone interested in the matter! Or looking for a base camp to initiate a graduate project. They break the methods into four families

1. Deterministic approximations (e.g., Laplace approximations)
2. Methods based on density estimation (e.g., Chib’s method, aka the candidate’s formula)
3. Importance sampling, including sequential Monte Carlo, with a subsection connecting with MCMC
4. Vertical representations (mostly, nested sampling)

Besides sheer computation, the survey also broaches upon issues like improper priors and alternatives to Bayes factors. The parts I would have done in more details are reversible jump MCMC and the long-lasting impact of Geyer’s reverse logistic regression (with the noise contrasting extension), even though the link with bridge sampling is briefly mentioned there. There is even a table reporting on the coverage of earlier surveys. Of course, the following postnote of the manuscript

The Christian Robert’s blog deserves a special mention , since Professor C. Robert has devoted several entries of his blog with very interesting comments regarding the marginal likelihood estimation and related topics.

does not in the least make me less objective! Some of the final recommendations

• use of Naive Monte Carlo [simulate from the prior] should be always considered [assuming a proper prior!]
  
• a multiple-try method is a good choice within the MCMC schemes
• optimal umbrella sampling estimator is difficult and costly to implement , so its best performance may not be achieved in practice
• adaptive importance sampling uses the posterior samples to build a suitable normalized proposal, so it benefits from localizing samples in regions of high posterior probability while preserving the properties of standard importance sampling
  
• Chib’s method is a good alternative, that provide very good performances [but is not always available]
• the success [of nested sampling] in the literature is surprising.

transformation MCMC

For reasons too long to describe here, I recently came across a 2013 paper by Dutta and Bhattacharya (from ISI Kolkata) entitled MCMC based on deterministic transforms, which sounded a bit dubious until I realised the deterministic label apply to the choice of the transformation and not to the Metropolis-Hastings proposal… The core of the proposed method is to make a proposal that simultaneously considers a move and its inverse, namely from x to either x’=T(x,ε) or x”=T⁻¹(x,ε) , where ε is an independent random noise, possibly degenerated to a manifold of lesser dimension. Due to the symmetry the acceptance probability is then a ratio of the target, multiplied by the x-Jacobian of T (as in reversible jump). I tried the method on a mixture of Gamma distributions target (in red) with an Exponential scale change and the resulting sample indeed fitted said target.

The authors even make an argument in favour of a unidimensional noise, although this amounts to running an implicit Gibbs sampler. Argument based on a reduced simulation cost for ε, albeit the full dimensional transform x’=T(x,ε) still requires to be computed. And as noted in the paper this also requires checking for irreducibility. The claim for higher efficiency found therein is thus mostly unsubstantiated…

“The detailed balance requirement also demands that, given x, the regions covered by the forward and the backward transformations are disjoint.”

The above statement is also surprising in that the generic detailed balance condition does not impose such a restriction.

 

distributed evidence

Alexander Buchholz (who did his PhD at CREST with Nicolas Chopin), Daniel Ahfock, and my friend Sylvia Richardson published a great paper on the distributed computation of Bayesian evidence in Bayesian Analysis. The setting is one of distributed data from several sources with no communication between them, which relates to consensus Monte Carlo even though model choice has not been particularly studied from that perspective. The authors operate under the assumption of conditionally conjugate models, i.e., the existence of a data augmentation scheme into an exponential family so that conjugate priors can be used. For a division of the data into S blocks, the fundamental identity in the paper is

where α is the normalising constant of the sub-prior exp{log[p(θ)]/S} and the other terms are associated with this prior. Under the conditionally conjugate assumption, the integral can be approximated based on the latent variables. Most interestingly, the associated variance is directly connected with the variance of

under the joint:

“The variance of the ratio measures the quality of the product of the conditional sub-posterior as an importance sample proposal distribution.”

Assuming this variance is finite (which is likely). An approximate alternative is proposed, namely to replace the exact sub-posterior with a Normal distribution, as in consensus Monte Carlo, which should obviously require some consideration as to which parameterisation of the model produces the “most normal” (or the least abnormal!) posterior. And ensures a finite variance in the importance sampling approximation (as ensured by the strong bounds in Proposition 5). A problem shared by the bridgesampling package.

“…if the error that comes from MCMC sampling is relatively small and that the shard sizes are large enough so that the quality of the subposterior normal approximation is reasonable, our suggested approach will result in good approximations of the full data set marginal likelihood.”

The resulting approximation can also be handy in conjunction with reversible jump MCMC, in the sense that RJMCMC algorithms can be run in parallel on different chunks or shards of the entire dataset. Although the computing gain may be reduced by the need for separate approximations.

averaged acceptance ratios

In another recent arXival, Christophe Andrieu, Sinan Yıldırım, Arnaud Doucet, and Nicolas Chopin study the impact of averaging estimators of acceptance ratios in Metropolis-Hastings algorithms. (It is connected with the earlier arXival rephrasing Metropolis-Hastings in terms of involutions discussed here.)

“… it is possible to improve performance of this algorithm by using a modification where the acceptance ratio r(ξ) is integrated with respect to a subset of the proposed variables.”

This interpretation of the current proposal makes it a form of Rao-Blackwellisation, explicitly mentioned on p.18, where, using a mixture proposal, with an adapted acceptance probability, it depends on the integrated acceptance ratio only. Somewhat magically using this ratio and its inverse with probability ½. And it increases the average Metropolis-Hastings acceptance probability (albeit with a larger number of simulations). Since the ideal averaging is rarely available, the authors implement a Monte Carlo averaging version. With applications to the exchange algorithm and to reversible jump MCMC. The major application is to pseudo-marginal settings with a high complexity (in the number T of terms) and where the authors’ approach does scale efficiently with T. There is even an ABC side to the story as one illustration is made of the ABC approximation to the posterior of an α-stable sample. As an encompassing proposal for handling Metropolis-Hastings environments with latent variables and several versions of the acceptance ratios, this is quite an interesting paper that I think we will study in further detail with our students.

deterministic moves in Metropolis-Hastings

A curio on X validated where an hybrid Metropolis-Hastings scheme involves a deterministic transform, once in a while. The idea is to flip the sample from one mode, ν, towards the other mode, μ, with a symmetry of the kind

μ-α(x+μ) and ν-α(x+ν)

with α a positive coefficient. Or the reciprocal,

-μ+(μ-x)/α and -ν+(ν-x)/α

for… reversibility reasons. In that case, the acceptance probability is simply the Jacobian of the transform to the proposal, just as in reversible jump MCMC.

Why the (annoying) Jacobian? As explained in the above slides (and other references), the Jacobian is there to account for the change of measure induced by the transform.

Returning to the curio, the originator of the question had spotted some discrepancy between the target and the MCMC sample, as the moments did not fit well enough. For a similar toy model, a balanced Normal mixture, and an artificial flip consisting of

x’=±1-x/2 or x’=±2-2x

implemented by

  u=runif(5)
  if(u[1]<.5){
    mhp=mh[t-1]+2*u[2]-1
    mh[t]=ifelse(u[3]<gnorm(mhp)/gnorm(mh[t-1]),mhp,mh[t-1])
  }else{
    dx=1+(u[4]<.5)
    mhp=ifelse(dx==1,
               ifelse(mh[t-1]<0,1,-1)-mh[t-1]/2,
               2*ifelse(mh[t-1]<0,-1,1)-2*mh[t-1])
    mh[t]=ifelse(u[5]<dx*gnorm(mhp)/gnorm(mh[t-1])/(3-dx),mhp,mh[t-1])

I could not spot said discrepancy beyond Monte Carlo variability.