Approximate injectivity (Q1656718): Difference between revisions

The category of metric spaces and nonexpansive maps can be made complete, cocomplete, and symmetric monoidal closed, by enlarging it with metrics that allow distance infinity between two points but otherwise obey the same laws. Consider a category enriched in this category of generalized metric spaces. It now makes sense to talk about when two morphisms \(A \to B\) are \(\varepsilon\) close to each other, rather than simply equal. This allows one to ask approximative questions. For example, the authors develop a theory of colimits that are (strictly) universal among cocones that commute up to \(\varepsilon\). Similarly, one can speak about \(\varepsilon\)-monomorphisms, and \(\varepsilon\)-pure morphisms. But their main goal is to develop an approximative version of the theory of injectivity. An object \(K\) is \(\varepsilon\)-injective for a morphism \(f \colon A \to B\) when every \(g \colon A \to K\) allows \(h \colon B \to K\) such that \(h \circ f\) is \(\varepsilon\)-close to \(g\). For a class of morphisms \(F\), consider the class of objects \(\varepsilon\)-injective to every \(f \in F\). Such classes of objects are called \(\varepsilon\)-injectivity classes. The main result characterizes \(\varepsilon\)-injectivity classes as those classes of morphisms that are closed under products, directed colimits, and \(\varepsilon\)-pure morphisms. This is later upgraded to a version that quantifies over all \(\varepsilon>0\); these approximate injectivity classes turn out to coincide with ordinary injectivity classes. (All of this requires appropriate background assumptions, namely that the category is enriched in generalized metric spaces, its underlying ordinary category is locally \(\lambda\)-presentable for an uncountable regular cardinal \(\lambda\), and any \(\lambda\)-presentable object is also \(\lambda\)-presentable in the underlying ordinary category.) The main example is the category of Banach spaces and nonexpansive linear maps. The last section applies the general theory here to show that there exists a separable \(\aleph_0\)-approximately-saturated Banach space, which is also known as the Gurarii space.