**W**hile attending my last session at MCqMC 2018, in Rennes, before taking a train back to Paris, I was confronted by this radical opinion upon our previous work with Matt Moores (Warwick) and other coauthors from QUT, where the speaker, Maksym Byshkin from Lugano, defended a new approach for maximum likelihood estimation using novel MCMC methods. Based on the point fixe equation characterising maximum likelihood estimators for exponential families, when theoretical and empirical moments of the natural statistic are equal. Using a Markov chain with stationary distribution the said exponential family, the fixed point equation can be turned into a zero divergence equation, requiring simulation of pseudo-data from the model, which depends on the unknown parameter. Breaking this circular argument, the authors note that simulating pseudo-data that reproduce the observed value of the sufficient statistic is enough. Which is related with Geyer and Thomson (1992) famous paper about Monte Carlo maximum likelihood estimation. From there I was and remain lost as I cannot see why a derivative of the expected divergence with respect to the parameter θ can be computed when this divergence is found by Monte Carlo rather than exhaustive enumeration. And later used in a stochastic gradient move on the parameter θ… Especially when the null divergence is imposed on the parameter. In any case, the final slide shows an application to a large image and an Ising model, solving the problem (?) in 140 seconds and suggesting indecency, when our much slower approach is intended to produce a complete posterior simulation in this context.

## Archive for Brittany

## indecent exposure

Posted in Statistics with tags ABC, Bayesian optimisation, Bretagne, Brittany, exponential families, image analysis, image processing, inference, Lugano, maximum likelihood estimation, MCqMC 2018, pre-processing, Rennes on July 27, 2018 by xi'an## Rennes street art [#3]

Posted in pictures, Running, Travel with tags Brittany, graffitis, MCqMC 2018, Rennes, street art on July 24, 2018 by xi'an## unusual clouds [jatp]

Posted in pictures, Travel, Wines with tags ÌPA, beer, Brittany, Cité d'Ys, clouds, jatp, MCqMC 2018, Rennes, summer light, sunset, thunderstorm on July 19, 2018 by xi'an## a thread to bin them all [puzzle]

Posted in Books, Kids, R, Travel with tags Brittany, clouds, FiveThirtyEight, mathematical puzzle, meerkats, R, Rennes, sunset, The Riddler on July 9, 2018 by xi'an

**T**he most recent riddle on the Riddler consists in finding the shorter sequence of digits (in 0,1,..,9) such that all 10⁴ numbers between 0 (or 0000) and 9,999 can be found as a group of consecutive four digits. This sequence is obviously longer than 10⁴+3, but how long? On my trip to Brittany last weekend, I wrote an R code first constructing the sequence at random by picking with high preference the next digit among those producing a new four-digit number

tenz=10^(0:3) wn2dg=function(dz) 1+sum(dz*tenz) seqz=rep(0,10^4) snak=wndz=sample(0:9,4,rep=TRUE) seqz[wn2dg(wndz)]=1 while (min(seqz)==0){ wndz[1:3]=wndz[-1];wndz[4]=0 wndz[4]=sample(0:9,1,prob=.01+.99*(seqz[wn2dg(wndz)+0:9]==0)) snak=c(snak,wndz[4]) sek=wn2dg(wndz) seqz[sek]=seqz[sek]+1}

which usually returns a value above 75,000. I then looked through the sequence to eliminate useless replicas

for (i in sample(4:(length(snak)-5))){ if ((seqz[wn2dg(snak[(i-3):i])]>1) &(seqz[wn2dg(snak[(i-2):(i+1)])]>1) &(seqz[wn2dg(snak[(i-1):(i+2)])]>1) &(seqz[wn2dg(snak[i:(i+3)])]>1)){ seqz[wn2dg(snak[(i-3):i])]=seqz[wn2dg(snak[(i-3):i])]-1 seqz[wn2dg(snak[(i-2):(i+1)])]=seqz[wn2dg(snak[(i-2):(i+1)])]-1 seqz[wn2dg(snak[(i-1):(i+2)])]=seqz[wn2dg(snak[(i-1):(i+2)])]-1 seqz[wn2dg(snak[i:(i+3)])]=seqz[wn2dg(snak[i:(i+3)])]-1 snak=snak[-i] seqz[wn2dg(snak[(i-3):i])]=seqz[wn2dg(snak[(i-3):i])]+1 seqz[wn2dg(snak[(i-2):(i+1)])]=seqz[wn2dg(snak[(i-2):(i+1)])]+1 seqz[wn2dg(snak[(i-1):(i+2)])]=seqz[wn2dg(snak[(i-1):(i+2)])]+1}}

until none is found. A first attempt produced 12,911 terms in the sequence. A second one 12,913. A third one 12,871. Rather consistent figures but not concentrated enough to believe in achieving a true minimum. An overnight run produced 12,779 as the lowest value. Checking the answer the week after, it appears that 10⁴+3 *is* the correct answer!