a journal of the plague, sword, and famine year [no end on sight]

Read the second volume of The Craft Sequence, Two Serpents Rise, by Max Gladstone,  with great difficulties as I found the story (again) poorly constructed, despite some characters being mostly well-designed (no connection with volume 1, except for taking place in the same universe, if at another time period). Mixing steampunk and hard fantasy involving gods does not work well in general and particularly there…. Following a New York Tĩmes review of the sequel, I also went very quickly through the Unwanted Dead, a first volume by Chris Lloyd, HWA Gold Crown for Best Historical Fiction winner for 2021, following a (s)hell-shocked PTSD-ed Paris police detective during World War II, when German troops arrive in the city. Not very realistic imho, as the nosy inspector happens to cross paths with Hitler during his very brief and unique visit to Paris as well as in Compiègne, and with a disappointing resolution of the wagon murders, but well-documented and with no obvious anachronism (except the unlikely presence of bathrooms in all apartments!, and the detective drinking whisky). (A wee nitpicking: Neuilly-sur-Seine (west of Paris) seemed to be confused with Neuilly-Plaisance (east of Paris), but the author acknowledged to me a general tendency to confuse east and west, just like I usually confuse right and left…) Overall, I found the Berlin Noir (Philip Kerr’s) novels more impressive and engaging!

Had a matcha flan in Paris, following a tip from Le Monde!, but was somewhat disappointed by its mild flavour, if comforted by the hojicha kokicha (made solely of tea stems) they served. And an excellent Filipino dinner in Kenilworth. And a yummy lamb Turkish Gözleme next to the ATI in London. While snacking the rest of week on Mysore dosas made on the street next to the Statistics Department at Warwick.

Watched (via a neighbour screen, on the flight to Martinique!) La Nuit du 12, a French thriller that got elected as Film of the Year (2022) by the Le Masque & La Plume (France Inter) audience, following a police investigation in the Maurienne valley after a particularly grisly murder of a young girl, one of the most fascinating aspects being that the crime remains unsolved despite the police efforts. In an impromptu home-made (!) Michelle Yeoh cycle, rewatched Crouching Tiger, Hidden Dragon after reading a particularly positive article in The Guardian. While the fighting scenes are definitely worth watching, esp. the trio fight on ice, the story remains rather lame. And Everything Everywhere All at Once, which I had also partly watched in the plane, but found highly unsatisfactory overall as lacking purpose, despite some great scenes between Michelle Yeoh and Jamie Lee Curtis ! Concurring with the strongly critical analyses in The New Yorker and the Guardian at the failure of the Daniels to find a purpose and a pace. (To quote from the latter, “these often impressively nutso formal backflips land in a position of pedestrian sentimentality, and then upbraid anyone resisting the viscous flood of sap for their cynicism.”) The scenes around the Everything Bagel are interminable…

séminaire parisien de statistique [09/01/23]

I had missed the séminaire parisien de statistique for most of the Fall semester, hence was determined to attend the first session of the year 2023, the more because the talks were close to my interest. To wit, Chiara Amorino spoke about particle systems for McKean-Vlasov SDEs, when those are parameterised by several parameters, when observing repeatedly discretised versions, hereby establishing the consistence of a contrast estimator of these estimators. I was initially confused by the mention of interacting particles, since the work is not at all about related with simulation. Just wondering whether this contrast could prove useful for a likelihood-free approach in building a Gibbs distribution?

Valentin de Bortoli then spoke on diffusion Schrödinger bridges for generative models, which allowed me to better my understanding of this idea presented by Arnaud at the Flatiron workshop last November. The presentation here was quite different, using a forward versus backward explanation via a sequence of transforms that end up approximately Gaussian, once more reminiscent of sequential Monte Carlo. The transforms are themselves approximate Gaussian versions relying on adiscretised Ornstein-Ulhenbeck process, with a missing score term since said score involves a marginal density at each step of the sequence. It can be represented [as below] as an expectation conditional on the (observed) variate at time zero (with a connection with Hyvärinen’s NCE / score matching!) Practical implementation is done via neural networks.

Last but not least!, my friend Randal talked about his Kick-Kac formula, which connects with the one we considered in our 2004 paper with Jim Hobert. While I had heard earlier version, this talk was mostly on probability aspects and highly enjoyable as he included some short proofs. The formula is expressing the stationary probability measure π of the original Markov chain in terms of explorations between two visits to an accessible set C, more general than a small set. With at first an annoying remaining term due to the set not being Harris recurrent but which eventually cancels out. Memoryless transportation can be implemented because C is free for the picking, for instance the set where the target is bounded by a manageable density, allowing for an accept-reject step. The resulting chain is non-reversible. However, due to the difficulty to simulate from the target restricted to C, a second and parallel Markov chain is instead created. Performances, unsurprisingly, depend on the choice of C, but it can be adapted to the target on the go.

New York City trip

While the Sampling, Transport, Diffusion workshop at the Flatiron Institute kept me happily busy, and while I did not stay any longer, spending a few days in New York City was a treat and I took advantage of my early hours to go running along the river sides, first south of the Flatiron building, then north to the Queensboro Bridge and over it, and last north along the Hudson River. The East River side is much less convenient for running as the path is repeatedly blocked by construction / storage sites and Xing the Queensboro Bridge gave a great view of Manhattan, albeit at the risk of being hit by a bike / scooter / moppet, as the path was shared with [an endless flow of] speeding electric bicycles. As I had never been to this part of the city, I was unaware of the cable car / gondola to Roosevelt Island (surprisingly called tram), which I would have taken given an extra day. Came by uponchance over a Trump Tower, which I ignored was so inappropriately close to the UN Headquarters! Running on the uninterrupted Hudson River trail was much nicer (and busier) despite the freezing wind that day.

For once (!) I stayed in an hotel, reserved by the Flatiron, and for the three nights I was there it was most tolerable, except for the usual background noise found in hotels, both from heating fans and patrons discussing in the corridors after hours. But the staff was helpful to the point of purchasing a kettle for my early morning tea. As the workshop provided an enormous amount of food at all times (and there was a true matcha tea provider around the corner!), it did not matter in the least.