Showing changes from revision #3 to #4:
Added | Removed | Changed
J. M. E. Hyland, 27.11.2012, Classical lambda calculus in modern dress, arXiv:1211.5762
Let denote a standard skeleton of finite sets.
An algebraic theory is defined to be a functor equipped with the structure of a cartesian operad:
We have sets of -ary multimaps (map with in-coming edges) with variable renamings (symmetric operad). (Think as the hom space of the “underlying“ multicategory.)
There are projections and in particular the identity . (One could also call the -th projection “-projection” or “-identity”.)
There are compositions which are associative, unital, compatible with projections, and natural in and .
The definition of algebraic theory encodes the rule rules of type formation:
where we interpret as to be declared in and as the result of substituting the strings of terms in .
(Paul Taylor writes
This is a beta-rule. )
(some comments on two classical approaches which are not used in the paper:
monads on whose functor part is given by the coend formula
Lawvere theories)
Term-formation-in-context in Martin-Löf’s approach:
computation rule:
This means: an interpretation of lambda calculus is a cartesian multicategory with semiclosed structure.
Let be an algebraic theory.
(1) A semiclosed structure on is defined to be collection of retractions
which is natural in and compatible with the actions
and
(2) A -theory is defined to be an algebraic theory equipped with a semi-closed structure.