the incredible shrinking pools

I was reading an opinion piece in The Guardian about the sorry state of public pools in England. With more and more closing for lack of proper funding, this being aggravated by the explosion in heating costs, as pools are excluded from governmental help. And the resulting impact on public health (and the NHS), since providing one less opportunity for exercising. And on general safety, since nowadays less children can swim… Which reminded me of the difficulty to find a pool in Oxford and Oaxaca. And of the cost of entering one in Roma and Milano. And of the relative accessibility of French pools, at least in cities, as shown by the estimation that 95% of the French high school students can swim to some extent.

a desolation called peace [book review]

This book by Arkady Martine is a sequel to the Hugo Awarded Memory called Empire, which I appreciated, pverall. I also did enjoy this one, no matter how different the settings are.

“The statistical chance of imago-integration failure leading to irreversible psychological and/or neurological damage is 0.3%”

Indeed, this second (and last?) volume is much more space-opera-esque in that most of the action takes place on a spatial fleet trying to fight an incomprehensible and invading alien force (whose mindset is rendered through an initially obscure chapter!). And subject to internal tensions, despite its military hierarchical structure. While the attempts at communicating with this unknown enemy are central to the story, they echo the main theme of the Teixcalaanli series (duology?), which is on how to reach the delicate balance between complete assimilation into a rich and fascinating culture and isolationism in order to preserve one’s original culture and way of life, doubled by the dilemmas caused by falling in love with someone from this other culture. (This may be the strongest aspect of the novel.) The related theme is the opposition between collective and individualistic societies, even though power competition is described in both the Teixcalaanli and the space station societies. (All three groups have achieved a way to operate as a collective. I actually wondered whether the “desolation” in the title was itself an intended collective, as in a murder of crows, esp. since a major character is nicknamed Swarm, but I could not find this collective ever being used.) It brought back some memories of Ender’s Game, in the sense of facing a radically different but still sapient species and backing away from complete annihilation. (The futuristic component of the book is as sketchy as in the previous one, with USB sticks being carried by spaceships as the only way to communicate, for no clear reason… But this is far from being of importance.) As an aside, the author linked her Teixcalaanli construct with the Mixtec civilisation, from Oaxaca.

Oaxaca murals [#4]

Oaxaca murals [#3]

ABC by classification

As a(nother) coincidence, yesterday, we had a reading group discussion at Paris Dauphine a few days after Veronika Rockova presented the paper in person in Oaxaca. The idea in ABC by classification that she co-authored with Yuexi Wang and Tetsuya Kaj is to use the empirical Kullback-Leibler divergence as a substitute to the intractable likelihood at the parameter value θ. In the generalised Bayes setting of Bissiri et al. Since this quantity is not available it is estimated as well. By a classification method that somehow relates to Geyer’s 1994 inverse logistic proposal, using the (ABC) pseudo-data generated from the model associated with θ. The convergence of the algorithm obviously depends on the choice of the discriminator used in practice. The paper also makes a connection with GANs as a potential alternative for the generalised Bayes representation. It mostly focus on the frequentist validation of the ABC posterior, in the sense of exhibiting a posterior concentration rate in n, the sample size, while requiring performances of the discriminators that may prove hard to check in practice. Expanding our 2018 result to this setting, with the tolerance decreasing more slowly than the Kullback-Leibler estimation error.

Besides the shared appreciation that working with the Kullback-Leibler divergence was a nice and under-appreciated direction, one point that came out of our discussion is that using the (estimated) Kullback-Leibler divergence as a form of distance (attached with a tolerance) is less prone to variability (or more robust) than using directly (and without tolerance) the estimate as a substitute to the intractable likelihood, if we interpreted the discrepancy in Figure 3 properly. Another item was about the discriminator function itself: while a machine learning methodology such as neural networks could be used, albeit with unclear theoretical guarantees, it was unclear to us whether or not a new discriminator needed be constructed for each value of the parameter θ. Even when the simulations are run by a deterministic transform.