Archive for Bayesian learning

What is luck? [book review]

Posted in Books, Statistics, Travel, University life with tags , , , , , , , , , , , , , , , , , , , , , , , , on December 10, 2021 by xi'an

I was sent—by Columbia University Press—this book for a potential review in CHANCE: What are the chances? (Why we believe in luck?) was written by Barbara Blatchley, professor of Psychology and Neuroscience at Agnes Scott College in Decatur, Georgia. I have read rather quickly its 193 pages over the recent trips I made to Marseille and Warwick. The topic is truly about luck and the psychology of the feeling of being luck or unlucky. There is thus rather little to relate to as a statistician, as this is not a book about chance! (I always need to pay attention when using both words, since, in French chance primarily means luck, while malchance means bad luck. And the French term for chance and randomness is hasard…) The book is pleasant to read, even though the accumulation of reports about psychological studies may prove tiresome in the long run and, for a statistician, worrisome as to which percentage of such studies were properly validated by statistical arguments…

“…the famous quote by Louis Pasteur: “Dans les champs de l’observation, le hasard ne favorise que les esprits préparés”s (…) Pasteur never saw a challenge he couldn’t overcome with patience and preparation.” (p.19)

Even the part about randomness is a-statistical and mostly a-probabilist, rather focusing on our subjective and biased (un)ability to judge randomness. The author introduces us to the concepts of apophenia, which is “the unmotivated seeing of connections accompanied with a specific feeling of abnormal meaningfulness”, and of patternicity for the “tendency to find meaningful patterns in meaningless noise”. She also states that (Neyman-Pearson) Type I error is about seeing a pattern in random noise while Type II errors are for conclusion of meaningless when the data is meaningful (p.15). Which is reductive to say the least, but lead her to recall the four types of luck proposed by James Austin (which I first misread as Jane Austin).

“There is a long-standing and deeply intimate connection between luck, religion, and belief in the supernatural.” (p.28)

I enjoyed very much the sections on these connections between a belief in luck and religions, even though the anthropological references to ancient religions are not strongly connected to luck, but rather to the belief that gods and goddesses could modify one’s fate (and avoiding the most established religions). Still, I appreciate her stressing the fact that if one believes in luck (as opposed to sheer randomness), this expresses at the very least a form of irrational belief in higher powers that can bend randomness in one’s favour (or disfavour). Which is the seed for more elaborate if irrational beliefs. (For illustrations, Borgès’ stories come to mind.)

“B.F. Skinner believed that superstitious behaviour was a consequence of learning and reinforcement.” (p.85)

There are also parts where (a belief in) luck and (human) learning are connected, but, unfortunately, no mention is made of the (vaguely) Bayesian nature of the (plastic, p. 188) brain modus operandi. The large section on the brain found in the book is instead physiological, since concerned with finding regions where the belief in luck could be located. In relation with attention-deficit disorders. (Revealing the interesting existence (for me) of mirror neurons, dedicated to predicting what could happen! Described as “predictive coding”, p.153). The last chapter “How to get lucky” contains a rather lengthy account of “Clever Hans”, the 1990 German counting horse (!). Who, as well-known, reacted to subtle and possibly unconscious signals from his trainer rather than to an equine feeling for arithmetic…

One of the clearest conclusions of the book is (imho) that a belief in luck may improve the life of the believers, while a belief in being unlucky may deteriorate it. The Taoist tale finishing the book is a pure gem. But I am still in the dark as to whether or not my exceptional number of bike punctures in the past year qualifies as bad luck!

“Luck is the way you face the randomness of the world.” (p.191)

As an irrelevant aside, one anecdote at the beginning of the book brought back memories of the Wabash River flowing through Lafayette, IN, as it tells of the luck of two Purdue female rowers who attempted a transatlantic race and survived capsizing in the middle of the Atlantic. It also made me regret I had not realised at the time there was a rowing opportunity there!

Sugiton at dawn [jatp]

Posted in Mountains, pictures, Running, Travel with tags , , , , , , , , , , on October 28, 2021 by xi'an

off to Luminy!!!

Posted in Mountains, pictures, Running, Statistics, Travel, University life with tags , , , , , , , , , , , , on October 24, 2021 by xi'an

end-to-end Bayesian learning [CIRM]

Posted in Books, Kids, Mountains, pictures, Running, Statistics, University life with tags , , , , , , , , , , , , , , , , , on February 1, 2021 by xi'an

Next Fall, there will be a workshop at CIRM, Luminy, Marseilles, on Bayesian learning. It takes place 22-29 October 2021 on this wonderful campus at the border with the beautiful Parc National des Calanques, in a wonderfully renovated CIRM building and involves friends and colleagues of mine as organisers and plenary speakers. (I am not involved!, but plan to organise a scalable MCMC workshop there the year after!) The conference is well-supported and the housing fees will be minimal since the centre is also subsidized by CNRS. The deadline for contributed talks and posters is 22 March, while it is 15 June for registration. Hopefully by this time the horizon will have cleared up enough to consider traveling and meeting again. Hopefully. (In which case I will miss this wonderful conference due to other meeting and teaching commitments in the Fall.)

online approximate Bayesian learning

Posted in Statistics with tags , , , , , , , on September 25, 2020 by xi'an

My friends and coauthors Matthieu Gerber and Randal Douc have just arXived a massive paper on online approximate Bayesian learning, namely the handling of the posterior distribution on the parameters of a state-space model, which remains a challenge to this day… Starting from the iterated batch importance sampling (IBIS) algorithm of Nicolas (Chopin, 2002) which he introduced in his PhD thesis. The online (“by online we mean that the memory and computational requirement to process each observation is finite and bounded uniformly in t”) method they construct is guaranteed for the approximate posterior to converge to the (pseudo-)true value of the parameter as the sample size grows to infinity, where the sequence of approximations is a Cesaro mixture of initial approximations with Gaussian or t priors, AMIS like. (I am somewhat uncertain about the notion of a sequence of priors used in this setup. Another funny feature is the necessity to consider a fat tail t prior from time to time in this sequence!) The sequence is in turn approximated by a particle filter. The computational cost of this IBIS is roughly in O(NT), depending on the regeneration rate.

%d bloggers like this: