Archive for Università degli studi di Padova
Bayesian privacies [slides]
Posted in Books, Statistics, Travel, University life with tags adversarial privacy, Bayesian privacy, differential privacy, inference, loss function, multi-agent decision theory, optimal decision, persuasive privacy, seminar, slides, slideshare, summary statistics, Università degli studi di Padova, utility on April 4, 2026 by xi'an
seminari di scienza statistiche a Padova
Posted in Books, Statistics, Travel, University life with tags Bayesian Adversarial Privacy, Italy, Padova, seminar, Università Ca' Foscari Venezia, Università degli studi di Padova, Venezia, visiting position on April 2, 2026 by xi'an
Di nuovo a Venezia per un’altra settimana
Posted in pictures, Travel, University life with tags Dorsoduro, Italia, Italy, La Serenessima, Padova, Università Ca' Foscari Venezia, Università degli studi di Padova, Venezia, Venice on March 30, 2026 by xi'an
a journal of the [experienced] plague and pestilence year
Posted in Books, Kids, Mountains, pictures, Running, Travel, University life with tags Argentan half-marathon, Belgium, book review, books, Britain, COVID-19, Daniel Defoe, fantasy, film review, Game of Thrones, House of Dragons, Journal of the Plague Year, pandemics, Russian invasion, short stories, Ukraine, Università degli studi di Padova, young adult books on November 4, 2022 by xi'an
Read The Cybernetic Tea Shop, by Meredith Katz, which is a short and rather clever (if YA) novel about the hazy boundary between humans and humanoids. Plus involving tea addicts! (Which is presumably why Amazon suggested it to me following my reading A Psalm for the Wide Built). And further read over a few sleepless nights the terrible Isandor series starting with City of Strife, by Claudie Arseneault, which had an interesting built of characters and fantasy universe, only to collapse into the usual cracks of super-evil villeins, a massive imbalance of power and a focus on the mundane (like foods and romantic attractions) when their society is under attack. The writing style is also heavily handed, to the point that I found myself skipping more and more paragraphs as the story unfolded. And will definitely not consider the incoming volume.
Went smoothly through my first (?) COVID positivity, which only caused a mild fever over one single day, amidst common cold symptom. Luckily did not pass it to anyone in my immediate vicinity, and resumed running if not swimming almost immediately (if not hard enough to train for the Argentan 1/2 marathon!). But sadly missed the 800th anniversary conference in Padova, as I was still testing positive the day before. I may have gotten infected in Britain or Belgium, despite my constant use of a mask (except in restaurants!).
Watched three more episodes of House of the Dragon, with great characters but a definitive lack of scope (when compared with Game of Thrones). The story remains at a highly local level of power fights and bickering, with existential threats inexistent. Still relatively enjoyable.
Finite mixture models do not reliably learn the number of components
Posted in Books, Statistics, University life with tags Bayes factor, David McKay, Edinburgh, empirical Bayes methods, finite mixtures, ICML 2021, infinite mixture, Mike Titterington, Padova, posterior concentration, Radford Neal, Scotland, Università degli studi di Padova, University of Glasgow on October 15, 2022 by xi'an
When preparing my talk for Padova, I found that Diana Cai, Trevor Campbell, and Tamara Broderick wrote this ICML / PLMR paper last year on the impossible estimation of the number of components in a mixture.
“A natural check on a Bayesian mixture analysis is to establish that the Bayesian posterior on the number of components increasingly concentrates near the truth as the number of data points becomes arbitrarily large.” Cai, Campbell & Broderick (2021)
Which seems to contradict [my formerly-Glaswegian friend] Agostino Nobile who showed in his thesis that the posterior on the number of components does concentrate at the true number of components, provided the prior contains that number in its support. As well as numerous papers on the consistency of the Bayes factor, including the one against an infinite mixture alternative, as we discussed in our recent paper with Adrien and Judith. And reminded me of the rebuke I got in 2001 from the late David McKay when mentioning that I did not believe in estimating the number of components, both because of the impact of the prior modelling and of the tendency of the data to push for more clusters as the sample size increased. (This was a most lively workshop Mike Titterington and I organised at ICMS in Edinburgh, where Radford Neal also delivered an impromptu talk to argue against using the Galaxy dataset as a benchmark!)
“In principle, the Bayes factor for the MFM versus the DPM could be used as an empirical criterion for choosing between the two models, and in fact, it is quite easy to compute an approximation to the Bayes factor using importance sampling” Miller & Harrison (2018)
This is however a point made in Miller & Harrison (2018) that the estimation of k logically goes south if the data is not from the assumed mixture model. In this paper, Cai et al. demonstrate that the posterior diverges, even when it depends on the sample size. Or even the sample as in empirical Bayes solutions.
Xi'an's Og 