Further reading

Sheaves and sites organize localization and descent [ Mac Lane and Moerdijk , 1992 ] . Standard category-theory texts supply the universal-property view of quotients and presentation invariance; see Mac Lane [ 1998 ] and Riehl [ 2017 ] . Mirror descent and Mirror–Prox show how geometry can select a numerical repair without redefining its structural obligation [ Beck and Teboulle , 2003 , Juditsky et al. , 2011 ] .

Admission has close analogues in double/debiased machine learning and targeted learning, which seek valid inference beyond in-sample fit [ Chernozhukov et al. , 2018 , van der Laan and Rose , 2011 ] .