Search arXivSearch

arXiv · 1706.04700

Normalizing the Taylor expansion of non-deterministic λ-terms, via parallel reduction of resource vectors

Abstract

It has been known since Ehrhard and Regnier's seminal work on the Taylor expansion of $λ$-terms that this operation commutes with normalization: the expansion of a $λ$-term is always normalizable and its normal form is the expansion of the Böhm tree of the term. We generalize this result to the non-uniform setting of the algebraic $λ$-calculus, i.e. $λ$-calculus extended with linear combinations of terms. This requires us to tackle two difficulties: foremost is the fact that Ehrhard and Regnier's techniques rely heavily on the uniform, deterministic nature of the ordinary $λ$-calculus, and thus cannot be adapted; second is the absence of any satisfactory generic extension of the notion of Böhm tree in presence of quantitative non-determinism, which is reflected by the fact that the Taylor expansion of an algebraic $λ$-term is not always normalizable. Our solution is to provide a fine grained study of the dynamics of $β$-reduction under Taylor expansion, by introducing a notion of reduction on resource vectors, i.e. infinite linear combinations of resource $λ$-terms. The latter form the multilinear fragment of the differential $λ$-calculus, and resource vectors are the target of the Taylor expansion of $λ$-terms. We show the reduction of resource vectors contains the image of any $β$-reduction step, from which we deduce that Taylor expansion and normalization commute on the nose. We moreover identify a class of algebraic $λ$-terms, encompassing both normalizable algebraic $λ$-terms and arbitrary ordinary $λ$-terms: the expansion of these is always normalizable, which guides the definition of a generalization of Böhm trees to this setting.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Lionel Vaux. 2019-07-30. Normalizing the Taylor expansion of non-deterministic λ-terms, via parallel reduction of resource vectors. https://doi.org/10.23638/lmcs-15(3%3A9)2019

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

Deciding Predicate Logical Theories of Real-Valued Functions

The notion of a real-valued function is central to mathematics, computer science, and many other scientific fields. Despite this importance, there are hardly any positive results on decision procedures for predicate logical theories that reason about real-valued functions. This paper defines a first-order predicate language for reasoning about multi-dimensional smooth real-valued functions and their derivatives, and demonstrates that - despite the obvious undecidability barriers - certain positive decidability results for such a language are indeed possible.

cs.LO

Structural Liveness of Conservative Petri Nets

We show that the EXPSPACE-hardness result for structural liveness of Petri nets [Jancar and Purser, 2019] holds even for a simple subclass of conservative nets. As our main result, we prove that for structurally live conservative nets, the values of the minimal live markings are at most doubly exponential in the size of the net. This implies the EXPSPACE-completeness of structural liveness for conservative Petri nets. The result also applies to structurally bounded Petri nets, whereas the complexity of the general case remains open. As a proof ingredient of independent interest, we present an extension of known results on the bounds of minimal integer solutions to Boolean combinations of linear equalities, inequalities, and divisibility constraints.

cs.LO

Verifying Numerical Methods with Isabelle/HOL

Modern machine learning pipelines and ODE solvers are built on numerical algorithms. Reliable numerical methods are thus a prerequisite for trustworthy machine learning and cyber-physical systems. We evaluate a framework designed for verifying imperative programs and the Isabelle proof assistant as tools for proving the total correctness of four numerical algorithms: the bisection method, the fixed-point method, the perceptron, and the gradient descent algorithm. Our verifications required subtle extensions and generalisations to Isabelle's version of Taylor's theorem and higher-order derivatives. Finally, we reflect on the framework's automation, friendly syntax, and on further requirements to turn it into a verification tool for numerical methods.

cs.LO