Materials Informatics Challenges

machine learningsolid state physicsinfo-graphicscetztikz

The three stages of a materials-informatics pipeline and what goes wrong at each. Data challenges cover the four Vs together with curation concerns like standardization and long-term storage. Descriptor challenges are about encoding a structure so the representation is invariant under symmetry, translation, rotation and permutation while staying compact and cheap to compute. Model challenges are the ones that decide whether a result is usable at all: extrapolation, transfer learning, reproducibility, computational cost and whether known physics is built in. Inspired by https://tex.stackexchange.com/a/387466.