EM rocks!

A rare occurrence of a statistics paper in Nature!, well Nature Scientific Reports, where the authors, Jaya Prakesh, Umang Agarwal and Phaneendra K. Yalavarthy, describe using a parallel implementation of the EM algorithm, for an image reconstruction in rock tomography. Due to a 1,887,436,800 x 1,887,436,800 matrix involved in the original 3D model.

Venezuelan COVID figures [and Nature infographics of the week]

Nature has started a  “week in infographics” column with a rather poor panel of figures the first time I checked (single numbers represented by surfaces and so on), but this week shows an unquestionable rendering of the discrepancy between the official COVID-19 death rates in Venezuela and in neighbouring countries. Obviously, there is no reason for all countries to share the same patterns (as seen for instance with the huge difference between France and Germany during the first wave in 2020), especially when considering the economic state of Venezuela, with very few entries from abroad and a general lack of movement throughout the country, but the attached Nature article argues that there is indeed a severe undercounting/reporting. The statistics that exist are actually gathered by a clandestine network of epidemiologists and doctors, with an impossibility to cross-check these as the baseline excess death statistics are not available either. Since 2016. The network responsible for the data collection is clandestine as the scientists involved risk detention or being fired from their position, despite the terrible lack of doctors and medical facilities in the country (which I observed in 2007 when getting an anti-rabies shot proved impossible!)… Another “shoot the messenger” in another dictatorship…  

quantum computing reproducibility crisis?

While standing in a train to my mother’s house in Brittany, I was catching up on earlier Nature issues and came upon this April issue where, following the retraction of a Nature paper on the topic, Sergey Frolov casts doubt on the possible detection of a new type of quantum particle, the Majorana fermion, whose existence still remains inconclusive. The criticism concentrates on the data analysis of signals where the appearance of a narrow peak should support the hypothesised existence. The article is interesting (to me) as a reflection of someone having published positive, then negative articles on the topic, upon the tendency for authors in the field to cherry-pick experiments where some peaks occur. Among dozens or hundred of experiments where they did not. And calling for open data and more stringent review(er)s on the matter (and others). The arguments in the opinion tribune sound most reasonable but I wonder whether or not other particle physicists share the same concern.

reproducibility check [Nature]

While reading the Nature article Swarm Learning, by Warnat-Herresthal et [many] al., which goes beyond federated learning by removing the need for a central coordinator, [if resorting to naïve averaging of the neural network parameters] I came across this reporting summary on the statistics checks made by the authors. With a specific box on Bayesian analysis and MCMC implementation!

moralizing gods drive Nature rejection