random walk on a torus [riddle]

The Riddler of this week(-end) has a simple riddle to propose, namely given a random walk on the {1,2,…,N} torus with a ⅓ probability of death, what is the probability of death occurring at the starting point?

The question is close to William Feller’s famous Chapter III on random walks. With his equally famous reflection principle. Conditioning on the time n of death, which as we all know is definitely absorbing (!), the event of interest is a passage at zero, or any multiple of N (omitting the torus cancellation), at time n-1 (since death occurs the next time). For a passage in zero, this does not happen if n is even (since n-1 is odd) and else it is a Binomial event with probability

For a passage in kN, with k different from zero, kN+n must be odd and the probability is then

which leads to a global probability of

i.e.

Since this formula is rather unwieldy I looked for another approach in a métro ride [to downtown Paris to enjoy a drink with Stephen Stiegler]. An easier one is to allocate to each point on the torus a probability p[i] to die at position 1 and to solve the system of equations that is associated with it. For instance, when N=3, the system of equations is reduced to

which leads to a probability of ½ to die at position 0 when leaving from 0. When letting N grows to infinity, the torus structure no longer matters and the probability of dying at position 0 implies returning in position 0, which is a special case of the above combinatoric formula, namely

which happens to be equal to

as can be [unnecessarily] checked by a direct R simulation. This √5 is actually the most surprising part of the exercise!

Sunday morning reading

A very interesting issue of Nature I read this morning while having breakfast. A post-brexit read of a pre-brexit issue. Apart from the several articles arguing against Brexit and its dire consequences on British science [but preaching to the converted for which percentage of the Brexit voters does read Nature?!], a short vignette on the differences between fields for the average time spent for refereeing a paper (maths takes twice as long as social sciences and academics older than 65 half the time of researchers under 36!). A letter calling for action against predatory publishers. And the first maths paper published since I started reading Nature on an almost-regular basis: it studies mean first-passage time for non-Markov random walks. Which are specified as time-homogeneous increments. It is sort of a weird maths paper in that I do not see where the maths novelty stands and why the paper only contains half a dozen formulas… Maybe not a maths paper after all.

optimal simulation on a convex set

This morning, we had a jam session at the maths department of Paris-Dauphine where a few researchers & colleagues of mine presented their field of research to the whole department. Very interesting despite or thanks to the variety of topics, with forays into the three-body problem(s) [and Poincaré‘s mistake], mean fields for Nash equilibrium (or how to exit a movie theatre), approximate losses in machine learning and so on. Somehow, there was some unity as well through randomness, convexity and optimal transport. One talk close to my own interests was obviously the study of simulation within convex sets by Joseph Lehec from Paris-Dauphine [and Sébastien Bubeck & Ronen Eldan] as they had established a total variation convergence result at a speed only increasing polynomially with the dimension.  The underlying simulation algorithm is rather theoretical in that it involves random walk (or Langevin corrected) moves where any excursion outside the convex support is replaced with its projection on the set. Projection that may prove pretty expensive to compute if the convex set is defined for instance as the intersection of many hyperplanes. So I do not readily see how the scheme can be recycled into a competitor to a Metropolis-Hastings solution in that the resulting chain hits the boundary from time to time. With the same frequency over iterations. A solution is to instead use Metropolis-Hastings of course, while another one is to bounce on the boundary and then correct by Metropolis-Hastings… The optimal scales in the three different cases are quite different, from √d in the Metropolis-Hastings cases to d√d in the projection case. (I did not follow the bouncing option to the end, as it lacks a normalising constant.) Here is a quick and not particularly helpful comparison of the exploration patterns of both approaches in dimension 50 for the unit sphere and respective scales of 10/d√d [blue] and 1/√d [gold].

R typos

At MCMskv, Alexander Ly (from Amsterdam) pointed out to me some R programming mistakes I made in the introduction to Metropolis-Hastings algorithms I wrote a few months ago for the Wiley on-line encyclopedia! While the outcome (Monte Carlo posterior) of the corrected version is moderately changed this is nonetheless embarrassing! The example (if not the R code) was a mixture of a Poisson and a Geometric distributions borrowed from our testing as mixture paper. Among other things, I used a flat prior on the mixture weights instead of a Beta(1/2,1/2) prior and a simple log-normal random walk on the mean parameter instead of a more elaborate second order expansion discussed in the text. And I also inverted the probabilities of success and failure for the Geometric density. The new version is now available on arXiv, and hopefully soon on the Wiley site, but one (the?) fact worth mentioning here is that the (right) corrections in the R code first led to overflows, because I was using the Beta random walk Be(εp,ε(1-p)) which major drawback I discussed here a few months ago. With the drag that nearly zero or one values of the weight parameter produced infinite values of the density… Adding 1 (or 1/2) to each parameter of the Beta proposal solved the problem. And led to a posterior on the weight still concentrating on the correct corner of the unit interval. In any case, a big thank you to Alexander for testing the R code and spotting out the several mistakes…

independent Metropolis-Hastings

“In this paper we have demonstrated the potential benefits, both theoretical and practical, of the independence sampler over the random walk Metropolis algorithm.”

Peter Neal and Tsun Man Clement Lee arXived a paper on optimising the independent Metropolis-Hastings algorithm. I was a bit surprised at this “return” of the independent sampler, which I hardly mention in my lectures, so I had a look at the paper. The goal is to produce an equivalent to what Gelman, Gilks and Wild (1996) obtained for random walk samplers.  In the formal setting when the target is a product of n identical densities f, the optimal number k of components to update in one Metropolis-Hastings (within Gibbs) round is approximately 2.835/I, where I is the symmetrised Kullback-Leibler distance between the (univariate) target f and the independent proposal q. When I is finite. The most surprising part is that the optimal acceptance rate is again 0.234, as in the random walk case. This is surprising in that I usually associate the independent Metropolis-Hastings algorithm with high acceptance rates. But this is of course when calibrating the proposal q, not the block size k of the Gibbs part. Hence, while this calibration of the independent Metropolis-within-Gibbs sampler is worth the study and almost automatically applicable, it remains that it only applies to a certain category of problems where blocking can take place. As in the disease models illustrating the paper. And requires an adequate choice of proposal distribution for, otherwise, the above quote becomes inappropriate.

Non-reversible Markov Chains for Monte Carlo sampling

This “week in Warwick” was not chosen at random as I was aware there is a workshop on non-reversible MCMC going on. (Even though CRiSM sponsored so many workshops in September that almost any week would have worked for the above sentence!) It has always been kind of a mystery to me that non-reversibility could make a massive difference in practice, even though I am quite aware that it does. And I can grasp some of the theoretical arguments why it does. So it was quite rewarding to sit in this Warwick amphitheatre and learn about overdamped Langevin algorithms and other non-reversible diffusions, to see results where convergence times moved from n to √n, and to grasp some of the appeal of lifting albeit in finite state spaces. Plus, the cartoon presentation of Hamiltonian Monte Carlo by Michael Betancourt was a great moment, not only because of the satellite bursting into flames on the screen but also because it gave a very welcome intuition about why reversibility was inefficient and HMC appealing. So I am grateful to my two colleagues, Joris Bierkens and Gareth Roberts, for organising this exciting workshop, with a most profitable scheduling favouring long and few talks. My next visit to Warwick will also coincide with a workshop on intractable likelihood, next November. This time part of the new Alan Turing Institute programme.

the Flatland paradox [#2]

Another trip in the métro today (to work with Pierre Jacob and Lawrence Murray in a Paris Anticafé!, as the University was closed) led me to infer—warning!, this is not the exact distribution!—the distribution of x, namely

since a path x of length l(x) will corresponds to N draws if N-l(x) is an even integer 2p and p undistinguishable annihilations in 4 possible directions have to be distributed over l(x)+1 possible locations, with Feller’s number of distinguishable distributions as a result. With a prior π(N)=1/N on N, hence on p, the posterior on p is given by

Now, given N and  x, the probability of no annihilation on the last round is 1 when l(x)=N and in general

which can be integrated against the posterior. The numerical expectation is represented for a range of values of l(x) in the above graph. Interestingly, the posterior probability is constant for l(x) large  and equal to 0.8125 under a flat prior over N.

Getting back to Pierre Druilhet’s approach, he sets a flat prior on the length of the path θ and from there derives that the probability of annihilation is about 3/4. However, “the uniform prior on the paths of lengths lower or equal to M” used for this derivation which gives a probability of length l proportional to 3^l is quite different from the distribution of l(θ) given a number of draws N. Which as shown above looks much more like a Binomial B(N,1/2).

However, being not quite certain about the reasoning involving Fieller’s trick, I ran an ABC experiment under a flat prior restricted to (l(x),4l(x)) and got the above, where the histogram is for a posterior sample associated with l(x)=195 and the gold curve is the potential posterior. Since ABC is exact in this case (i.e., I only picked N’s for which l(x)=195), ABC is not to blame for the discrepancy! I asked about the distribution on Stack Exchange maths forum (and a few colleagues here as well) but got no reply so far… Here is the R code that goes with the ABC implementation:

#observation:
elo=195
#ABC version
T=1e6
el=rep(NA,T)
N=sample(elo:(4*elo),T,rep=TRUE)
for (t in 1:T){
#generate a path
  paz=sample(c(-(1:2),1:2),N[t],rep=TRUE)
#eliminate U-turns
  uturn=paz[-N[t]]==-paz[-1]
  while (sum(uturn>0)){
    uturn[-1]=uturn[-1]*(1-
              uturn[-(length(paz)-1)])
    uturn=c((1:(length(paz)-1))[uturn==1],
            (2:length(paz))[uturn==1])
    paz=paz[-uturn]
    uturn=paz[-length(paz)]==-paz[-1]
    }
  el[t]=length(paz)}
#subsample to get exact posterior
poster=N[abs(el-elo)==0]