EM degeneracy

At the MHC 2021 conference today (to which I biked to attend for real!, first time since BayesComp!) I listened to Christophe Biernacki exposing the dangers of EM applied to mixtures in the presence of missing data, namely that the algorithm has a rising probability to reach a degenerate solution, namely a single observation component. Rising in the proportion of missing data. This is not hugely surprising as there is a real (global) mode at this solution. If one observation components are prohibited, they should not be accepted in the EM update. Just as in Bayesian analyses with improper priors, the likelihood should bar single or double  observations components… Which of course makes EM harder to implement. Or not?! MCEM, SEM and Gibbs are obviously straightforward to modify in this case.

Judith Rousseau also gave a fascinating talk on the properties of non-parametric mixtures, from a surprisingly light set of conditions for identifiability to posterior consistency . With an interesting use of several priors simultaneously that is a particular case of the cut models. Namely a correct joint distribution that cannot be a posterior, although this does not impact simulation issues. And a nice trick turning a hidden Markov chain into a fully finite hidden Markov chain as it is sufficient to recover a Bernstein von Mises asymptotic. If inefficient. Sylvain LeCorff presented a pseudo-marginal sequential sampler for smoothing, when the transition densities are replaced by unbiased estimators. With connection with approximate Bayesian computation smoothing. This proves harder than I first imagined because of the backward-sampling operations…

Institut de Mathématique d’Orsay [jatp]

humanitarian project in Madagascar

As the budget of the humanitarian trip to Madagascar our daughter organises with other students of the Paris-Sud Medical School next summer is still short of several thousand euros, I repost the call for support I made a few months ago.

Their project is called Mada Tsatsaka, mada for Madagascar and tsatsaka for a local lizard. The team plans to bring basic drugs and educational material and to work in a dispensary, an orphanage, as well as a shelter for women victims of violence. (More below!)

I thus bring this project to the ‘Og’s readers’ attention in case they wish to support. The best approach is use this web site for donations (in English) to Evadeh Mada Tsatsaka. (Evadeh is the mother association for all humanitarian projects in the medical school.) A free-of-charge (!) alternative is to shop on amazon.com following this associate link as I vouch to transfer all my associate gains in the next four months to the project.

Upon request, more details on the project:

• 2 weeks in Maventibao working in a clinic : Mada Clinics, helping two nurses with free medical examinations and providing extra medical equipment and drugs. And also helping with drinking water improvement. The team further hopes to help with the purchase of a car associated with the clinic and linking with the hospital in  Diego (4h away) and with hiring a doctor in the nearby clinic of Amboangamamy.
• 2 weeks in an orphanage in Antananarivo, Ankanifitahiana, in collaboration with BLOC Léo Madagascar, helping in financing and installing a library and a music room, and participating in classes and games with the children. Depending on the funding, the team would also like to help with installing a solar oven.

Mada Tsatsaka

Along with other students of the Paris-Sud Medical School, our daughter is planning a humanitarian trip to Madagascar next summer, following a long-term tradition in her school with groups sent to Madagascar as well as Vietnam, Peru and Nepal. The project is called Mada Tsatsaka, mada for Madagascar and tsatsaka for a local lizard. They plan to bring basic drugs and educational material and to work in a dispensary, an orphanage, as well as a shelter for women victims of violence. They have been now working for several months at meeting the budget for this project, from selling pies and quiches on local markets (I helped with the quiches!) to wrapping gifts at supermakets all around, to engaging local authorities and charities, but are still far from reaching this goal.

I thus bring this project to the ‘Og’s readers’ attention in case they wish to support. The best approach is use this donation form (in English) for Evadeh Mada Tsatsaka. A free of charge (!) alternative is to shop on amazon.com following this associate link as I vouch to transfer all my associate gains in the next six months to the project.

talk in Orsay (message in a beetle)

Yesterday (March 27), I gave a seminar at Paris-Sud University, Orsay, in the stats department, on ABC model choice. It was an opportunity to talk about recent advances we have made with Jean-Michel Marin and Pierre Pudlo on using machine-learning devices to improve ABC. (More to come soon!) And to chat with Gilles Celeux about machine learning and classification. Actually, given that one of my examples was about the Asian lady beetle invasion and that the buildings of the Paris-Sud University have suffered from this invasion, I should have advertised the talk with the more catchy title of “message in a beetle”…

This seminar was also an opportunity to experiment with mixed transportation. Indeed, since I had some errands to run in Paris in morning I decided to bike there (in Paris), work at CREST, and then take my bike in the RER train down to Orsay as I did not have the time and leisure to bike all the 20k there. Since it was the middle of the day, the carriage was mostly empty and I managed to type a blog entry without having to worry about the bike being a nuisance…. The only drag was to enter the platform in Paris (Cité Universitaire) as there was no clear access for bike. Fortunately, a student kindly helped me to get over the gate with my bike, as I could not manage on my own… Nonetheless, I will certainly repeat the experience on my next trip to Orsay (but would not dare take the bike inside/under Paris per se because of the (over-)crowded carriages there).