conditional noise contrastive estimation

At ICML last year, Ciwan Ceylan and Michael Gutmann presented a new version of noise constrative estimation to deal with intractable constants. While noise contrastive estimation relies upon a second independent sample to contrast with the observed sample, this approach uses instead a perturbed or noisy version of the original sample, for instance a Normal generation centred at the original datapoint. And eliminates the annoying constant by breaking the (original and noisy) samples into two groups. The probability to belong to one group or the other then does not depend on the constant, which is a very effective trick. And can be optimised with respect to the parameters of the model of interest. Recovering the score matching function of Hyvärinen (2005). While this is in line with earlier papers by Gutmann and Hyvärinen, this line of reasoning (starting with Charlie Geyer’s logistic regression) never ceases to amaze me!

thermodynamic integration plus temperings

Biljana Stojkova and David Campbel recently arXived a paper on the used of parallel simulated tempering for thermodynamic integration towards producing estimates of marginal likelihoods. Resulting into a rather unwieldy acronym of PT-STWNC for “Parallel Tempering – Simulated Tempering Without Normalizing Constants”. Remember that parallel tempering runs T chains in parallel for T different powers of the likelihood (from 0 to 1), potentially swapping chain values at each iteration. Simulated tempering monitors a single chain that explores both the parameter space and the temperature range. Requiring a prior on the temperature. Whose optimal if unrealistic choice was found by Geyer and Thomson (1995) to be proportional to the inverse (and unknown) normalising constant (albeit over a finite set of temperatures). Proposing the new temperature instead via a random walk, the Metropolis within Gibbs update of the temperature τ then involves normalising constants.

“This approach is explored as proof of concept and not in a general sense because the precision of the approximation depends on the quality of the interpolator which in turn will be impacted by smoothness and continuity of the manifold, properties which are difficult to characterize or guarantee given the multi-modal nature of the likelihoods.”

To bypass this issue, the authors pick for their (formal) prior on the temperature τ, a prior such that the profile posterior distribution on τ is constant, i.e. the joint distribution at τ and at the mode [of the conditional posterior distribution of the parameter] is constant. This choice makes for a closed form prior, provided this mode of the tempered posterior can de facto be computed for each value of τ. (However it is unclear to me why the exact mode would need to be used.) The resulting Metropolis ratio becomes independent of the normalising constants. The final version of the algorithm runs an extra exchange step on both this simulated tempering version and the untempered version, i.e., the original unnormalised posterior. For the marginal likelihood, thermodynamic integration is invoked, following Friel and Pettitt (2008), using simulated tempering samples of (θ,τ) pairs (associated instead with the above constant profile posterior) and simple Riemann integration of the expected log posterior. The paper stresses the gain due to a continuous temperature scale, as it “removes the need for optimal temperature discretization schedule.” The method is applied to the Glaxy (mixture) dataset in order to compare it with the earlier approach of Friel and Pettitt (2008), resulting in (a) a selection of the mixture with five components and (b) much more variability between the estimated marginal  likelihoods for different numbers of components than in the earlier approach (where the estimates hardly move with k). And (c) a trimodal distribution on the means [and unimodal on the variances]. This example is however hard to interpret, since there are many contradicting interpretations for the various numbers of components in the model. (I recall Radford Neal giving an impromptu talks at an ICMS workshop in Edinburgh in 2001 to warn us we should not use the dataset without a clear(er) understanding of the astrophysics behind. If I remember well he was excluded all low values for the number of components as being inappropriate…. I also remember taking two days off with Peter Green to go climbing Craigh Meagaidh, as the only authorised climbing place around during the foot-and-mouth epidemics.) In conclusion, after presumably too light a read (I did not referee the paper!), it remains unclear to me why the combination of the various tempering schemes is bringing a noticeable improvement over the existing. At a given computational cost. As the temperature distribution does not seem to favour spending time in the regions where the target is most quickly changing. As such the algorithm rather appears as a special form of exchange algorithm.

Bernoulli race particle filters

Sebastian Schmon, Arnaud Doucet and George Deligiannidis have recently arXived an AISTATS paper with the above nice title. The motivation for the extension is facing intractable particle weights for state space models, as for instance in discretised diffusions.  In most cases, actually, the weight associated with the optimal forward proposal involves an intractable integral which is the predictive of the current observed variate given the past hidden states. And in some cases, there exist unbiased and non-negative estimators of the targets,  which can thus be substituted, volens nolens,  to the original filter. As in many pseudo-marginal derivations, this new algorithm can be interpreted as targeting an augmented distribution that involves the auxiliary random variates behind the unbiased estimators of the particle weights. A worthwhile remark since it allows for the preservation of the original target as in (8) provided the auxiliary random variates are simulated from the right conditionals. (At least ideally as I have no clue when this is feasible.)

“if Bernoulli resampling is per-formed, the variance for any Monte Carlo estimate will be the same as if the true weights were known and one applies standard multinomial resampling.”

The Bernoulli race in the title stands for a version of the Bernoulli factory problem, where an intractable and bounded component of the weight can be turned into a probability, for which a Bernoulli draw is available, hence providing a Multinomial sampling with the intractable weights since replacing the exact probability with an estimate does not modify the Bernoulli distribution, amazingly so! Even with intractable normalising constants in particle filters. The practicality of the approach may however be restricted by the possibility of some intractable terms being very small and requiring many rejections for one acceptance, as the number of attempts is a compound geometric. The intractability may add to the time request the drawback of keeping this feature hidden as well. Or force some premature interruption in the settings of a parallel implementation.

nested sampling when prior and likelihood clash

A recent arXival by Chen, Hobson, Das, and Gelderblom makes the proposal of a new nested sampling implementation when prior and likelihood disagree, making simulations from the prior inefficient. The paper holds the position that a single given prior is used over and over all datasets that come along:

“…in applications where one wishes to perform analyses on many thousands (or even millions) of different datasets, since those (typically few) datasets for which the prior is unrepresentative can absorb a large fraction of the computational resources.” Chen et al., 2018

My reaction to this situation, provided (a) I want to implement nested sampling and (b) I realise there is a discrepancy, would be to resort to an importance sampling resolution, as we proposed in our Biometrika paper with Nicolas. Since one objection [from the authors] is that identifying outlier datasets is complicated (it should not be when the likelihood function can be computed) and time-consuming, sequential importance sampling could be implemented.

“The posterior repartitioning (PR) method takes advantage of the fact that nested sampling makes use of the likelihood L(θ) and prior π(θ) separately in its exploration of the parameter space, in contrast to Markov chain Monte Carlo (MCMC) sampling methods or genetic algorithms which typically deal solely in terms of the product.” Chen et al., 2018

The above salesman line does not ring a particularly convincing chime in that nested sampling is about as myopic as MCMC since based on the similar notion of a local proposal move, starting from the lowest likelihood argument (the minimum likelihood estimator!) in the nested sample.

“The advantage of this extension is that one can choose (π’,L’) so that simulating from π’ under the constraint L'(θ) > l is easier than simulating from π under the constraint L(θ) > l. For instance, one may choose an instrumental prior π’ such that Markov chain Monte Carlo steps adapted to the instrumental constrained prior are easier to implement than with respect to the actual constrained prior. In a similar vein, nested importance sampling facilitates contemplating several priors at once, as one may compute the evidence for each prior by producing the same nested sequence, based on the same pair (π’,L’), and by simply modifying the weight function.” Chopin & Robert, 2010

Since the authors propose to switch to a product (π’,L’) such that π’.L’=π.L, the solution appears like a special case of importance sampling, with the added drwaback that when π’ is not normalised, its normalised constant must be estimated as well. (With an extra nested sampling implementation?) Furthermore, the advocated solution is to use tempering, which is not so obvious as it seems in small dimensions. As the mass does not always diffuse to relevant parts of the space. A more “natural” tempering would be to use a subsample in the (sub)likelihood for nested sampling and keep the remainder of the sample for weighting the evaluation of the evidence.

same simulation, different acceptance

In doubly intractable settings, where the likelihood involves an intractable constant Z(θ), an auxiliary or pseudo- observation x is generated to incorporate strategically located densities in the acceptance probability towards cancelling out the Z(θ)’s. The funny thing is that Møller et al.  (2005) and Murray et al. (2006) both use the same simulations in their auxiliary algorithms, namely θ’~q(θ|θ,y) and x’~f(x|θ’), but return different acceptance probabilities. The former use an artificial target on the pair (θ’,x’) [with a free conditional on x’] while the later uses a pseudo-marginal argument to estimate the missing constant Z(θ) by importance sampling as noticed by Everitt (2012). This apparent paradox is rather common to simulation in that several importance weights can often be constructed for the same importance function. But in the case of doubly intractable distributions, the first approach offers a surprisingly wide variability in the selection of the conditional on x’, which can be absolutely any density g(x|θ,y). And hence could be optimised for maximal acceptance rate. Or maximal effective sample size. In the original paper of Møller et al.  (2005) a plug-in version f(x|θ) was suggested, with θ replaced with a crude estimate. This morning, when discussing both versions with Julien Stoehr, I realised that a geometric average of f(x|θ)’s could be used as well, since the intractable normalising constants would not be an issue [as opposed to an arithmetic or harmonic average]. I [idly] wonder if anything has been done in this direction…