Budget constrained simulations

Posted in Statistics by Pierre Jacob on 5 February 2019

You’ll have to read the post to get what these lines are about.

Hi all,

This post is about some results from “Bias Properties of Budget Constrained Simulations“, by Glynn & Heidelberger and published in Operations Research in 1990. I have found these results extremely useful, and our latest manuscript on unbiased MCMC recalls them in detail. Below I go through some of the results and describe the simulations that lead to the above figure.



Another update on unbiased smoothing

Posted in General by Pierre Jacob on 19 January 2019


Hi all,

This is a short update on my research on unbiased smoothing with coupled conditional particle filters. In a previous post I naively explained that I was done with the project since the article was accepted for publication in a journal.

However, a bug was found in the code thanks to a very careful reader. So I’ve fixed the code, re-launched all the simulations, and updated the arXiv version straight away; that was on September 5, 2018. The conclusions of the article are unchanged, thankfully; in fact, the text of the article is exactly the same, only the numerical values in the tables and the figures are different. Phew!

Since then the version with the buggy results was retracted from the journal, and hopefully, the updated version will be published soon. In the meantime the arXiv version is up-to-date.

Another take on the Hyvärinen score for model comparison

Posted in Statistics by Stephane Shao on 22 September 2018


Exact log-Bayes factors (log-BF) and H-factors (HF) of M1 against M2, computed for 100 independent samples (thin solid lines) of 1000 observations generated as i.i.d. N(1,1), under three increasingly vague priors for θ1.

In a former post, Pierre wrote about Bayesian model comparison and the limitations of Bayes factors in the presence of vague priors. Here we are, one year later, and I am happy to announce that our joint work with Jie Ding and Vahid Tarokh has been recently accepted for publication. As way of celebrating, allow me to give you another take on the matter.


Final update on unbiased smoothing

Posted in Statistics by Pierre Jacob on 27 August 2018

Two coupled chains, marginally following a conditional particle filter algorithm with ancestor sampling, and meeting in 20 iterations. Script here.


Two years ago I blogged about couplings of conditional particle filters for smoothing.  The paper with Fredrik Lindsten and Thomas Schön has just been accepted for publication at JASA, and the arXiv version and github repository are hopefully in their final forms. Here I’ll mention a few recent developments and follow-up articles by other researchers.


Couplings of Normal variables

Posted in R, Statistics by Pierre Jacob on 24 August 2018



Just to play a bit with the gganimate package, and to celebrate National Coupling Day, the above plot shows different couplings of two univariate Normal distributions, Normal(0,1) and Normal(2,1). That is, each point is a pair (x,y) where x follows a Normal(0,1) and y follows a Normal(2,1). Below I’ll recall briefly how each coupling operates, in the Normal case. The code is available at the end of the post.


Different ways of using MCMC algorithms

Posted in General, Statistics by Pierre Jacob on 19 July 2018



This post is about different ways of using Markov chain Monte Carlo (MCMC) algorithms for numerical integration or sampling. It can be a hard job to design an MCMC algorithm for a given target distribution. Once it’s finally implemented, it gives a way of sampling a new point X’ given an existing point X. From there, the algorithm can be used in various ways to construct estimators of integrals/distribution of interest. Some ways are more amenable to parallel computing than others. I give some examples with references below.


Bayesian workshop in Grenoble, September 6-7

Posted in General, Seminar/Conference by Julyan Arbel on 23 May 2018


We are organising a two-day Bayesian workshop in Grenoble in September 6-7, 2018. It will be the second edition of the Italian-French statistics seminar (link to first edition), titled this year: Bayesian learning theory for complex data modeling. The workshop will give to young statisticians the opportunity to learn from and interact with highly qualified senior researchers in probability, theoretical and applied statistics, with a particular focus on Bayesian methods.

Anyone interested in this field is welcome. There will be two junior sessions and a poster session with a call for abstract open until June 30. A particular focus will be given to researchers in the early stage of their career, or currently studying for a PhD, MSc or BSc. The junior session is supported by ISBA through travel awards.

There will be a social dinner on September 6, and a hike organised in the mountains on September 8.

Confirmed invited speakers

• Simon Barthelmé, Gipsa-lab, Grenoble, France
• Arnoldo Frigessi, University of Oslo, Norway
• Benjamin Guedj, Inria Lille – Nord Europe, France
• Alessandra Guglielmi, Politecnico di Milano, Italy
• Antonio Lijoi, University Bocconi, Milan, Italy
• Bernardo Nipoti, Trinity College Dublin, Ireland
• Sonia Petrone, University Bocconi, Milan, Italy

Important Dates:

• June 30, 2018: Abstract submission closes
• July 20, 2018: Notification on abstract acceptance
• August 25, 2018: Registration closes

More details and how to register:

We look forward to seeing you in Grenoble.



Tagged with: , ,

Scaling of MCMC with dimension (experiments)

Posted in Statistics by Pierre Jacob on 15 May 2018

Taken from Wikipedia’s article on dimension.

Hi all,

In this post, I’ll go through numerical experiments illustrating the scaling of some MCMC algorithms with respect to the dimension. I will focus on a simple setting, to illustrate some theoretical results developed by Gareth Roberts, Andrew Stuart, Alexandros Beskos, Jeff Rosenthal and many of their co-authors over many years, for instance here for random walk Metropolis-Hastings (RWMH, and see here more recently), here for Modified Adjusted Langevin Algorithm (MALA), here for Hamiltonian Monte Carlo (HMC).


A healthy dose of foundational crisis

Posted in General by Rémi Bardenet on 26 March 2018

I finally took the time to read about axiomatic foundations of Bayesian statistics. I like axioms, I like Bayesians stats, so this was definitely going to be a pleasant opportunity to read some books, comfortably seated on the sofa I just added to my office. Moreover, my team in Lille includes supporters of Dempster-Shafer belief functions, another framework for uncertainty modelling and decision-making, so being precise on my own axioms was the best way to discuss more constructively with my colleagues.

Long story short, I took the red pill: there is a significant gap between axiomatic constructions of the Bayesian paradigm and current Bayesian practice. None of this is new, but I had never been told. It’s not keeping me awake at night, but it’s bothering my office mates who cannot stop me blabbering about it over coffee. The good side is that smart people have thought about this in the past. Also, reading about this helped me understand some of the philosophical nuances between the thought processes of different Bayesians, say de Finettians vs. Jeffreysians. I will not attempt a survey in a blog post, nor do I feel to be knowledgeable enough for this, but I thought I could spare my office mates today and annoy Statisfaction’s readers for once.

Take Savage’s axioms, for instance. I’ve always heard that they were the current justification behind the saying “being Bayesian is being a coherent decision-maker”. To be precise, let \Theta be the set of states of the world, that is, everything useful to make your decision. To fix ideas, in a statistical experiment, your decision might be a “credible” interval on some real parameter, so \Theta should at least be the product of \mathbb{R} times whatever space your data live in. Now an action a is defined to be a map from \Theta to some set of outcomes \mathcal{Z}. For the interval problem, an action a_I corresponds to the choice of a particular interval I and the outcomes \mathcal{Z} should contain whatever you need to assess the performance of your action, say, the indicator of the parameter actually belonging to your interval I, and the length of I. Outcomes are judged by utility, that is, we consider functions u:\mathcal{Z}\rightarrow\mathbb{R}_+ that map outcomes to nonnegative rewards. In our example, this could be a weighted sum of the indicator and the interval length. The weights translate your preference for an interval that actually captures the value of the parameter of interest over a short interval. Now, the axioms give the equivalence between the two following bullets:

  • (being Bayesian) There is a unique pair (u,\pi), made of a utility function and a finitely additive probability measure \pi defined on all subsets of the set \Theta of states of the world, such that you choose your actions by maximizing an expected utility criterion:

\displaystyle a^\star \in \arg\max_a \mathbb{E}_\pi u(a(\theta)),

  • (being coherent) Ranking actions according to a preference relation that satisfies a few abstract properties that make intuitive sense for most applications, such as transitivity: if you prefer a_I to a_J and a_J to a_G, then you prefer a_I to a_G. Add to this a few structural axioms that impose constraints on \Theta on \mathcal{Z}.

Furthermore, there is a natural notion of conditional preference among actions that follows from Savage’s axioms. Taken together, these axioms give an operational definition of our “beliefs” that seems to match Bayesian practice. In particular, 1) our beliefs take the form of a probability measure –which depends on our utility–, 2) we should update these beliefs by conditioning probabilities, and 3) make decisions using expected utility with respect to our belief. This is undeniably beautiful. Not only does Savage avoid shaky arguments or interpretations by using your propensity to act to define your beliefs, but he also avoids using “extraneous probabilities”. By the latter I mean any axiom that artificially brings mathematical probability structures into the picture, such as “there exists an ideal Bernoulli coin”.

But the devil is in the details. For instance, some of the less intuitive of Savage’s axioms require the set of states of the world to be uncountable and the utility bounded. Also, the measure \pi is actually only required to be finitely additive, and it has to be defined on all subsets of the set of states of the world. Now-traditional notions like Lebesgue integration, \sigma-additivity, or \sigma-algebras do not appear. In particular, if you want to put a prior on the mean of a Gaussian that lives in \Theta=\mathbb{R}, Savage says your prior should weight all subsets of the real line, so forget about using any probability measure that has a density with respect to the Lebesgue measure! Or, to paraphrase de Finetti, \sigma-additive probability does not exist. Man, before reading about axioms I thought “Haha, let’s see whether someone has actually worked out the technical details to justify Bayesian nonparametrics with expected utility, this must be technically tricky”; now I don’t even know how to fit the mean of a Gaussian anymore. Thank you, Morpheus-Savage.

There are axiomatic ways around these shortcomings. From what I’ve read they all either include extraneous probabilities or rather artificial mathematical constructions. Extraneous probabilities lead to philosophically beautiful axioms and interpretations, see e.g. Chapter 2 of Bernardo and Smith (2000), and they can get you finite and countably finite sets of states of the world, for instance, whereas Savage’s axioms do not. Stronger versions also give you \sigma-additivity, see below. Loosely speaking, I understand extraneous probabilities as measuring uncertainty with respect to an ideal coin, similarly to measuring heat in degrees Celsius by comparing a physical system to freezing or boiling water. However, I find extraneous probability axioms harder to swallow than (most of) Savage’s axioms, and they involve accepting a more general notion of probability than personal propensity to act.

If you want to bypass extraneous probability and still recover \sigma-additivity, you could follow Villegas (1964), and try to complete the state space \Theta so that well-behaved measures \pi extend uniquely to \sigma-additive measures on a \sigma-algebra on this bigger set of states \hat\Theta. Defining the extended \hat\Theta involves sophisticated functional analysis, and requires to add potentially hard-to-intepret states of the world, so losing some of the interpretability of Savage’s construction. Authors of reference books seem reluctant to go in that direction: De Groot (1970), for instance, solves the issue by using a strong extraneous probability axiom that allows working in the original set \Theta with \sigma-additive beliefs. Bernardo & Smith use extraneous probabilities, but keep their measures finitely additive until the end of Chapter 2. Then they admit departing from axioms for practical purposes and define “generalized beliefs” in Chapter 3, defined on a \sigma-algebra of the original \Theta. Others seem to more readily accept the gap between axioms and practice, and look for a more pragmatic justification of the combined use of expected utility and countably additive probabilities. For instance, Robert (2007) introduces posterior expected utility, and then argues that it has desirable properties among decision-making frameworks, such as respecting the likelihood principle. This is unlike Savage’s approach, for whom the (or rather, a finitely additive version of the) likelihood principle is a consequence of the axioms. I think this is an interesting subtlelty.

To conclude, I just wanted to share my excitement for having read some fascinating works on decision-theoretic axioms for Bayesian statistics. There still is some unresolved tension between having both an applicable and axiomatized Bayesian theory of belief. I would love this post to generate discussions, and help me understand the different thought processes behind each Bayesian being Bayesian (and each non-Bayesian being non-Bayesian). For instance, I had not realised how conceptually different the points of view in the reference books of Robert and Bernardo & Smith were. This definitely helped me understand (Xi’an) Robert’s short three answers to this post.

If this has raised your interest, I will mention here a few complementary sources that I have found useful, ping me if you want more. Chapters 2 and 3 of Bernardo and Smith (2000) contain a detailed description of their set of axioms with extraneous probability, and they give great lists of pointers on thorny issues at the end of each chapter. A lighter read is Parmigiani and Inoue (2009), which I think is a great starting point, with emphasis on the main ideas of de Finetti, Ramsey, Savage, and Anscombe and Aumann, how they apply, and how they relate to each other, rather than the technical details. Technical details and exhaustive reviews of sets of axioms for subjective probability can be found in their references to Fishburn’s work, which I have found to be beautifully clear, rigorous and complete, although like many papers involving low-level set manipulations, the proofs sometimes feel like they are written for robots. But after all, a normative theory of rationality is maybe only meant for robots.


AI in Grenoble, 2nd to 6th July 2018

Posted in General, Seminar/Conference by Julyan Arbel on 22 March 2018

visuel PAISS T2

This is an advertisement for on conference on AI organised at Inria Grenoble by Thoth team and Naver labs : This AI summer school comprises lectures and practical sessions conducted by renowned experts in different areas of artificial intelligence.

This event is the revival of a past series of very successful summer schools which took place in Grenoble and Paris. The latest edition of this series was held in 2013. While originally focusing on computer vision, the summer school now targets a broader AI audience, and will also include presentations about machine learning, natural language processing, robotics, and cognitive science.

Note that NAVER LABS is funding a number of students to attend PAISS. Apply before 4th April. (more…)

Tagged with: , , ,
%d bloggers like this: