Data-Driven Discovery and Design of Superionic Lithium Conductors for High Performance Lithium Ion Batteries

We also implement locally weighted least squares linear regression with the following cost function: We search for the optimal model by performing leaveone-out cross validation over all models • Solid state lithium ion batteries show great promise as high performance energy storage devices; however, designing high ionic conductivity solid electrolytes remains a challenge. • There has yet to be a model that accurately generalizes trends in ionic conductivity among crystal structures to rapidly identify promising electrolyte materials in silico.