We analyze the computational complexity of type inference for untyped -terms in the second-order polymorphic typed -calculus (F 2 ) invented by Girard and Reynolds, as well as higher-order extensions F 3 ; F 4 ; : : : ; F ! proposed by Girard. We prove that recognizing the F 2 - typable terms requires exponential time, and for F ! the problem is nonelementary. We show as well a sequence of lower bounds on recognizing the F k -typable terms, where the bound for F k+1 is exponentially larger than that for F k . The lower bounds are based on generic simulation of Turing Machines, where computation is simulated at the expression and type level simultaneously. Non-accepting computations are mapped to non-normalizing reduction sequences, and hence non-typable terms. The accepting computations are mapped to typable terms, where higher-order types encode reduction sequences, and first-order types encode the entire computation as a circuit, based on a unification simulation of Boolean logic. ...

We investigate finite-rank intersection type systems, analyzing the complexity of their type inference problems and their relation to the problem of recognizing semantically equivalent terms. Intersection types allow something of type T1 /\ T2 to be used in some places at type T1 and in other places at type T2 . A finite-rank intersection type system bounds how deeply the /\ can appear in type expressions. Such type systems enjoy strong normalization, subject reduction, and computable type inference, and they support a pragmatics for implementing parametric polymorphism. As a consequence, they provide a conceptually simple and tractable alternative to the impredicative polymorphism of System F and its extensions, while typing many more programs than the Hindley-Milner type system found in ML and Haskell. While type inference is computable at every rank, we show that its complexity grows exponentially as rank increases. Let K(0, n) = n and K(t + 1, n) = 2^K(t,n); we prove that recognizing the pure lambda-terms of size n that are typable at rank k is complete for dtime[K(k-1, n)]. We then consider the problem of deciding whether two lambda-terms typable at rank k have the same normal form, Generalizing a well-known result of Statman from simple types to finite-rank intersection types. ...

This paper surveys a part of the theory of fi-reduction in -calculus which might aptly be called perpetual reductions. The theory is concerned with perpetual reduction strategies, i.e., reduction strategies that compute infinite reduction paths from -terms (when possible), and with perpetual redexes, i.e., redexes whose contraction in -terms preserves the possibility (when present) of infinite reduction paths. The survey not only recasts classical theorems in a unified setting, but also offers new results, proofs, and techniques, as well as a number of applications to problems in -calculus and type theory. 1. Introduction Considerable attention has been devoted to classification of reduction strategies in type-free -calculus [4, 6, 7, 15, 38, 44, 81]---see also [2, Ch. 13]. We are concerned with strategies differing in the length of reduction paths. This paper draws on several sources. In late 1994, van Raamsdonk and Severi [59] and Srensen [66, 67] independently developed ...

We first present a new proof for the standardization theorem, a fundamental theorem in -calculus. Since our proof is largely built upon structural induction on lambda terms, we can extract some bounds for the number of fi-reduction steps in the standard fi-reduction sequences obtained from transforming any given fi-reduction sequences. This result sharpens the standardization theorem and establishes a link between lazy and eager evaluation orders in the context of computational complexity. As an application, we establish a superexponential bound for the number of fi-reduction steps in fi-reduction sequences from any given simply typed -terms. 1 Introduction The standardization theorem of Curry and Feys [CF58] is a very useful result, stating that if u reduces to v for -terms u and v, then there is a standard fi-reduction from u to v. Using this theorem, we can readily prove the normalization theorem, i.e., a -term has a normal form if and only if the leftmost fi-reduction sequence f...

This paper surveys a part of the theory of fi-reduction in λ-calculus which might aptly be called perpetual reductions. The theory is concerned with perpetual reduction strategies, i.e., reduction strategies that compute infinite reduction paths from λ-terms (when possible), and with perpetual redexes, i.e., redexes whose contraction in λ-terms preserves the possibility (when present) of infinite reduction paths. The survey not only recasts classical theorems in a unified setting, but also offers new results, proofs, and techniques, as well as a number of applications to problems in λ-calculus and type theory.

Abstract. Carrying out a suggestion by Kreisel, we adapt Gödel’s functional interpretation to ordinary first-order predicate logic(PL) and thus devise an algorithm to extract Herbrand terms from PL-proofs. The extraction is carried out in an extension of PL to higher types. The algorithm consists of two main steps: first we extract a functional realizer, next we compute the β-normal-form of the realizer from which the Herbrand terms can be read off. Even though the extraction is carried out in the extended language, the terms are ordinary PL-terms. In contrast to approaches to Herbrand’s theorem based on cut elimination or ε-elimination this extraction technique is, except for the normalization step, of low polynomial complexity, fully modular and furthermore allows an analysis of the structure of the Herbrand terms, in the spirit of Kreisel ([13]), already prior to the normalization step. It is expected that the implementation of functional interpretation in Schwichtenberg’s MINLOG system can be adapted to yield an efficient Herbrand-term extraction tool. 1.

We prove that we may compute the normal form of each term of the simply typed - calculus in a polynomial number of sharable reductions (where the notion of sharing is L'evy's "optimal" one). As a simple corollary, we get that P = NP "up to sharing", i.e. up to the computational overhead due to sharing. 1 Introduction We prove that we may compute the normal form of each term of the simply typed -calculus (up to j-equivalence) in a polynomial number of sharable reductions. The notion of sharing is L'evy's one [Le78], commonly known as "optimal" sharing. The general idea (see Section 1.1 for the formal definition) is to formalize duplication of redexes as residuals modulo permutations. In particular, a redex u with history oe (notation oeu) is a copy of a redex v with history ae iff aev oeu (i.e., there exists ø such that oe = aeø up to permutation equivalence, and u is a residual of v after ø ). The family relation ' is then the symmetric and transitive closure of the copy-relation. ...

This paper revisits models of typed λ-calculus based on filters of intersection types: By using non-idempotent intersections, we simplify a methodology that produces modular proofs of strong normalisation based on filter models. Non-idempotent intersections provide a decreasing measure proving a key termination property, simpler than the reducibility techniques used with idempotent intersections. Such filter models are shown to be captured by orthogonality techniques: we formalise an abstract notion of orthogonality model inspired by classical realisability, and express a filter model as one of its instances, along with two term-models (one of which captures a now common technique for strong normalisation). Applying the above range of model constructions to Curry-style System F describes at different levels of detail how the infinite polymorphism of System F can systematically be reduced to the finite polymorphism of intersection types.