===================== Fun ===================== * Interesting pieces in the chemical literature `collected by Prof. Hintermann `_ * `Chocolate Consumption, Cognitive Function, and Nobel Laureates `_ * `Growth of Diamond Films from Tequila `_ * `Learnability is unprovable. `_ Similar to the logical paradoxes Gödel discovered and the continuum hypothesis there might be a universe in wish something is learnable from a finite sample whereas there also might be one where this is not possible * `Two layers are enough to memorise random labels `_ (especially the influence of regularization is interesting). A wider context is in this nice blogpost: https://www.inference.vc/everything-that-works-works-because-its-bayesian-2/ * :code:`import this` * `statistical thermodynamics as the maximally noncommital choice (link to information theory via maximum-entropy distributions and Shannon entropy) `_ * I do not like describing entropy as disorder as I think that it can be misleading: * `Disorder - A Cracked Crutch for Supporting Entropy Discussions `_ * `Entropy producing complex, ordered structures `_ (cp. Onsager rods or freezing hard spheres, also interesting is the influence of the entropy tax on self-assembly) * `Order through entropy `_ * Living in Switzerland, `this `_ is essential knowledge * Pickard's `hyperspatial geometry optimization `_ seems quite interesting for me to expedite geometry searches and it reminded me of the Nosé-Hoover thermostat ... * `Our AI skills might be worth less than we think `_ * I didn't think about the `enviornmental impact of streaming music `_ before reading this article. * Short reads: * `Counterexample Euler's conjecture `_ * `Writer's block treatment `_ * Bad reads: * Found on `Twitter `_: `The theoretical calculations indicated that with increasing Si/Ge ratio or a decrease in the amount of Ge, the framework energy of the D4R units rapidly increase` in `a paper without a computational chemist amongst the reviewers `_