First-order Unification Research Articles

Constructive Many-one Reduction from the Halting Problem to Semi-unification (Extended Version)

Semi-unification is the combination of first-order unification and first-order matching. The undecidability of semi-unification has been proven by Kfoury, Tiuryn, and Urzyczyn in the 1990s by Turing reduction from Turing machine immortality (existence of a diverging configuration). The particular Turing reduction is intricate, uses non-computational principles, and involves various intermediate models of computation. The present work gives a constructive many-one reduction from the Turing machine halting problem to semi-unification. This establishes RE-completeness of semi-unification under many-one reductions. Computability of the reduction function, constructivity of the argument, and correctness of the argument is witnessed by an axiom-free mechanization in the Coq proof assistant. Arguably, this serves as comprehensive, precise, and surveyable evidence for the result at hand. The mechanization is incorporated into the existing, well-maintained Coq library of undecidability proofs. Notably, a variant of Hooper's argument for the undecidability of Turing machine immortality is part of the mechanization.

NeuroLISP: High-level symbolic programming with attractor neural networks

Despite significant improvements in contemporary machine learning, symbolic methods currently outperform artificial neural networks on tasks that involve compositional reasoning, such as goal-directed planning and logical inference. This illustrates a computational explanatory gap between cognitive and neurocomputational algorithms that obscures the neurobiological mechanisms underlying cognition and impedes progress toward human-level artificial intelligence. Because of the strong relationship between cognition and working memory control, we suggest that the cognitive abilities of contemporary neural networks are limited by biologically-implausible working memory systems that rely on persistent activity maintenance and/or temporal nonlocality. Here we present NeuroLISP, an attractor neural network that can represent and execute programs written in the LISP programming language. Unlike previous approaches to high-level programming with neural networks, NeuroLISP features a temporally-local working memory based on itinerant attractor dynamics, top-down gating, and fast associative learning, and implements several high-level programming constructs such as compositional data structures, scoped variable binding, and the ability to manipulate and execute programmatic expressions in working memory (i.e., programs can be treated as data). Our computational experiments demonstrate the correctness of the NeuroLISP interpreter, and show that it can learn non-trivial programs that manipulate complex derived data structures (multiway trees), perform compositional string manipulation operations (PCFG SET task), and implement high-level symbolic AI algorithms (first-order unification). We conclude that NeuroLISP is an effective neurocognitive controller that can replace the symbolic components of hybrid models, and serves as a proof of concept for further development of high-level symbolic programming in neural networks.

Functions-as-constructors higher-order unification: extended pattern unification

Unification is a central operation in constructing a range of computational logic systems based on first-order and higher-order logics. First-order unification has several properties that guide its incorporation in such systems. In particular, first-order unification is decidable, unary, and can be performed on untyped term structures. None of these three properties hold for full higher-order unification: unification is undecidable, unifiers can be incomparable, and term-level typing can dominate the search for unifiers. The so-called pattern subset of higher-order unification was designed to be a small extension to first-order unification that respects the laws governing λ-binding (i.e., the equalities for α, β, and η-conversion) but which also satisfied those three properties. While the pattern fragment of higher-order unification has been used in numerous implemented systems and in various theoretical settings, it is too weak for many applications. This paper defines an extension of pattern unification that should make it more generally applicable, especially in proof assistants that allow for higher-order functions. This extension’s main idea is that the arguments to a higher-order, free variable can be more than just distinct bound variables. In particular, such arguments can be terms constructed from (sufficient numbers of) such bound variables using term constructors and where no argument is a subterm of any other argument. We show that this extension to pattern unification satisfies the three properties mentioned above.

Proof-relevant unification: Dependent pattern matching with only the axioms of your type theory

AbstractDependently typed languages such as Agda, Coq, and Idris use a syntactic first-order unification algorithm to check definitions by dependent pattern matching. However, standard unification algorithms implicitly rely on principles such asuniqueness of identity proofsandinjectivity of type constructors. These principles are inadmissible in many type theories, particularly in the new and promising branch known as homotopy type theory. As a result, programs and proofs in these new theories cannot make use of dependent pattern matching or other techniques relying on unification, and are as a result much harder to write, modify, and understand. This paper proposes a proof-relevant framework for reasoning formally about unification in a dependently typed setting. In this framework, unification rules compute not just a unifier but also a corresponding soundness proof in the form of anequivalencebetween two sets of equations. By rephrasing the standard unification rules in a proof-relevant manner, they are guaranteed to preserve soundness of the theory. In addition, it enables us to safely add new rules that can exploit the dependencies between the types of equations, such as rules for eta-equality of record types and higher dimensional unification rules for solving equations between equality proofs. Using our framework, we implemented a complete overhaul of the unification algorithm used by Agda. As a result, we were able to replace previousad-hocrestrictions with formally verified unification rules, fixing a substantial number of bugs in the process. In the future, we may also want to integrate new principles with pattern matching, for example, the higher inductive types introduced by homotopy type theory. Our framework also provides a solid basis for such extensions to be built on.

Type-Based Cost Analysis for Lazy Functional Languages

We present a static analysis for determining the execution costs of lazily evaluated functional languages, such as Haskell. Time- and space-behaviour of lazy functional languages can be hard to predict, creating a significant barrier to their broader acceptance. This paper applies a type-based analysis employing amortisation and cost effects to statically determine upper bounds on evaluation costs. While amortisation performs well with finite recursive data, we significantly improve the precision of our analysis for co-recursive programs (i.e. dealing with potentially infinite data structures) by tracking self-references. Combining these two approaches gives a fully automatic static analysis for both recursive and co-recursive definitions. The analysis is formally proven correct against an operational semantic that features an exchangeable parametric cost-model. An arbitrary measure can be assigned to all syntactic constructs, allowing to bound, for example, evaluation steps, applications, allocations, etc. Moreover, automatic inference only relies on first-order unification and standard linear programming solving. Our publicly available implementation demonstrates the practicability of our technique on editable non-trivial examples.

Unifiers as equivalences: proof-relevant unification of dependently typed data

Dependently typed languages such as Agda, Coq and Idris use a syntactic first-order unification algorithm to check definitions by dependent pattern matching. However, these algorithms don’t adequately consider the types of the terms being unified, leading to various unintended results. As a consequence, they require ad hoc restrictions to preserve soundness, but this makes them very hard to prove correct, modify, or extend. This paper proposes a framework for reasoning formally about unification in a dependently typed setting. In this framework, unification rules compute not just a unifier but also a corresponding correctness proof in the form of an equivalence between two sets of equations. By rephrasing the standard unification rules in a proof-relevant manner, they are guaranteed to preserve soundness of the theory. In addition, it enables us to safely add new rules that can exploit the dependencies between the types of equations. Using our framework, we reimplemented the unification algorithm used by Agda. As a result, we were able to replace previous ad hoc restrictions with formally verified unification rules, fixing a number of bugs in the process. We are convinced this will also enable the addition of new and interesting unification rules in the future, without compromising soundness along the way.

First-order unification in the PVS proof assistant

This work presents formalizations of properties and algorithms for the first-order unification theory in the proof assistant PVS. In previous works a formalization of the theorem of existence of most general unifiers (mgu's) for first-order unifiable problems was presented and then, it was illustrated how the formalizations of correctness and completeness used for the proof of the theorem of existence, can be adapted in order to verify greedy unification algorithms 'a la Robinson. The distinguishing feature of these two formalizations is that they remain close to the textbook proofs which was possible through the use of the higher-order specification language and proof engine of PVS. That work consists of two elaborated PVS theories having been the one related with the theorem of existence of mgu's applied for obtaining a full formalization of the Knuth-Bendix(Huet) critical pair theorem. In addition to these two developments, this work describes a formalization of the correction and completeness of a more elaborated first-order unification algorithm. In contrast with the greedy algorithm, the current one is more efficient. It navigates in a post-order traversal the tree structure of the two terms being unified in such a way that after the difference at the first position of conflict (found in inner-most, left-most order) between the terms is eliminated through a linkage substitution (mapping a variable into a term), the search of the following position of conflict is computed through application of auxiliary functions starting from the previous position. Unlike this more efficient traversal, the greedy unification algorithm always restarts the search of conflicts from the root of the terms being unified. The new formalization gave rise to a more elaborated PVS development in which unlike a relatively straightforward application of correctness and completeness formalizations of the theorem of existence of mgu's, it was necessary the formalization of a series of structural auxiliary lemmas to deal with formalization of properties of the tree data structure used to represent terms.

A Formalization of the Theorem of Existence of First-Order Most General Unifiers

This work presents a formalization of the theorem of existence of most general unifiers in first-order signatures in the higher-order proof assistant PVS. The distinguishing feature of this formalization is that it remains close to the textbook proofs that are based on proving the correctness of the well-known Robinson's first-order unification algorithm. The formalization was applied inside a PVS development for term rewriting systems that provides a complete formalization of the Knuth-Bendix Critical Pair theorem, among other relevant theorems of the theory of rewriting. In addition, the formalization methodology has been proved of practical use in order to verify the correctness of unification algorithms in the style of the original Robinson's unification algorithm.

Unification and matching on compressed terms

Term unification plays an important role in many areas of computer science, especially in those related to logic. The universal mechanism of grammar-based compression for terms, in particular the so-called singleton tree grammars (STGAs) , have recently drawn considerable attention. Using STGs, terms of exponential size and height can be represented in linear space. Furthermore, the term representation by directed acyclic graphs (dags) can be efficiently simulated. The present article is the result of an investigation on term unification and matching when the terms given as input are represented using different compression mechanisms for terms such as dags and singleton tree grammars. We describe a polynomial time algorithm for context matching with dags, when the number of different context variables is fixed for the problem. For the same problem, NP-completeness is obtained when the terms are represented using the more general formalism of singleton tree grammars. For first-order unification and matching polynomial time algorithms are presented, each of them improving previous results for those problems.

A Machine Checked Model of Idempotent MGU Axioms For Lists of Equational Constraints

We present formalized proofs verifying that the first-order unification algorithm defined over lists of satisfiable constraints generates a most general unifier (MGU), which also happens to be idempotent. All of our proofs have been formalized in the Coq theorem prover. Our proofs show that finite maps produced by the unification algorithm provide a model of the axioms characterizing idempotent MGUs of lists of constraints. The axioms that serve as the basis for our verification are derived from a standard set by extending them to lists of constraints. For us, constraints are equalities between terms in the language of simple types. Substitutions are formally modeled as finite maps using the Coq library Coq.FSets.FMapInterface. Coq's method of functional induction is the main proof technique used in proving many of the axioms.

First-order unification using variable-free relational algebra

We present a framework for the representation and resolution of first-order unification problems and their abstract syntax in a variable-free relational formalism which is an executable variant of the Tarski-Givant relation algebra and of Freyd’s allegories restricted to the fragment necessary to compile and run logic programs. We develop a decision procedure for validity of relational terms, which corresponds to solving the original unification problem. The decision procedure is carried out by a conditional relational-term rewriting system. There are advantages over classical unification approaches. First, cumbersome and underspecified meta-logical procedures (name clashes, substitution, etc.) and their properties (invariance under substitution of ground terms, equality’s congruence with respect to term forming, etc.) are captured algebraically within the framework. Second, other unification problems can be accommodated, for example, existential quantification in the logic can be interpreted as a new operation whose effect is to formalize the costly and error prone handling of fresh names (renaming apart).

Normative conflict resolution in multi-agent systems

Norms (permissions, obligations and prohibitions) offer a useful and powerful abstraction with which to capture social constraints in multi-agent systems. Norms should exclude disruptive or antisocial behaviour without prescribing the design of individual agents or restricting their autonomy. An important challenge, however, in the design and management of systems governed by norms is that norms may, at times, conflict with one another; e.g, an action may be simultaneously prohibited and obliged for a particular agent. In such circumstances, agents no longer have the option of complying with these norms; whatever they do or refrain from doing will lead to a social constraint being broken. In this paper, we present mechanisms for the detection and resolution of normative conflicts. These mechanisms, based on first-order unification and constraint solving techniques, are the building blocks of more sophisticated algorithms we present for the management of normative positions, that is, the adoption and removal of permissions, obligations and prohibitions in societies of agents. We capture both direct and indirect conflicts between norms, formalise a practical concept of authority, and model conflicts that may arise as a result of delegation. We are able to formally define classic ways for resolving conflicts such as lex superior and lex posterior.

Source-tracking unification

We propose a path-based framework for deriving and simplifying source-tracking information for first-order term unification in the empty theory. Such a framework is useful for diagnosing unificatio...

Nominal unification

We present a generalisation of first-order unification to the practically important case of equations between terms involving binding operations. A substitution of terms for variables solves such an equation if it makes the equated terms α -equivalent, i.e. equal up to renaming bound names. For the applications we have in mind, we must consider the simple, textual form of substitution in which names occurring in terms may be captured within the scope of binders upon substitution. We are able to take a “nominal” approach to binding in which bound entities are explicitly named (rather than using nameless, de~Bruijn-style representations) and yet get a version of this form of substitution that respects α-equivalence and possesses good algorithmic properties. We achieve this by adapting two existing ideas. The first one is terms involving explicit substitutions of names for names, except that here we only use explicit permutations (bijective substitutions). The second one is that the unification algorithm should solve not only equational problems, but also problems about the freshness of names for terms. There is a simple generalisation of classical first-order unification problems to this setting which retains the latter's pleasant properties: unification problems involving α-equivalence and freshness are decidable; and solvable problems possess most general solutions.

Solving Equations by Graph Transformation

Abstract We review the concept of term graph narrowing as an approach for solving equations by transformations on term graphs. Term graph narrowing combines term graph rewriting with first-order term unification. This mechanism is complete for all term rewriting systems over which term graph rewriting is normalizing and confluent. This includes, in particular, all convergent term rewriting systems. Completeness means that for every solution of a given equation, term graph narrowing can find an equivalent or more general solution. The general motivation for using term graphs instead of terms is to improve efficiency: sharing common subterms saves space and avoids the repetition of computations.

First-order unification by structural recursion

First-order unification algorithms (Robinson, 1965) are traditionally implemented via general recursion, with separate proofs for partial correctness and termination. The latter tends to involve counting the number of unsolved variables and showing that this total decreases each time a substitution enlarges the terms. There are many such proofs in the literature (Manna & Waldinger, 1981; Paulson, 1985; Coen, 1992; Rouyer, 1992; Jaume, 1997; Bove, 1999). This paper shows how a dependent type can relate terms to the set of variables over which they are constructed. As a consequence, first-order unification becomes a structurally recursive program, and a termination proof is no longer required. Both the program and its correctness proof have been checked using the proof assistant LEGO (Luo & Pollack, 1992; McBride, 1999).

Unification of Higher-order Patterns modulo Simple Syntactic Equational Theories

We present an algorithm for unification of higher-order patterns modulo simple syntactic equational theories as defined by Kirchner [14]. The algorithm by Miller [17] for pattern unification, refined by Nipkow [18] is first modified in order to behave as a first-order unification algorithm. Then the mutation rule for syntactic theories of Kirchner [13,14] is adapted to pattern E-unification. If the syntactic algorithm for a theory E terminates in the first-order case, then our algorithm will also terminate for pattern E-unification. The result is a DAG-solved form plus some equations of the form λ øverlinex.F(øverlinex) = λ øverlinex. F(øverlinex^π ) where øverlinex^π is a permutation of øverlinex When all function symbols are decomposable these latter equations can be discarded, otherwise the compatibility of such equations with the solved form remains open.

Type inference with rank 1 polymorphism for type-directed compilation of ML

This paper defines an extended polymorphic type system for an ML-style programming language, and develops a sound and complete type inference algorithm. Different frdm the conventional ML type discipline, the proposed type system allows full rank 1 polymorphism , where polymorphic types can appear in other types such as product types, disjoint union types and range types of function types. Because of this feature, the proposed type system significantly reduces the value-only restriction of polymorphism, which is currently adopted in most of ML-style impure languages. It also serves as a basis for efficient implementation of type-directed compilation of polymorphism. The extended type system achieves more efficient type inference algorithm, and it also contributes to develop more efficient type-passing implementation of polymorphism. We show that the conventional ML polymorphism sometimes introduces exponential overhead both at compile-time elaboration and run-time type-passing execution, and that these problems can be eliminated by our type inference system. Compared with a more powerful rank 2 type inference systems based on semi-unification, the proposed type inference algorithm infers a most general type for any typable expression by using the conventional first-order unification, and it is therefore easily adopted in existing implementation of ML family of languages.

A Full Formalization of SLD-Resolution in the Calculus of Inductive Constructions

This paper presents a full formalization of the semantics of definite programs, in the calculus of inductive constructions. First, we describe a formalization of the proof of first-order terms unification: this proof is obtained from a similar proof dealing with quasi-terms, thus showing how to relate an inductive set with a subset defined by a predicate. Then, SLD-resolution is explicitely defined: the renaming process used in SLD-derivations is made explicit, thus introducing complications, usually overlooked, during the proofs of classical results. Last, switching and lifting lemmas and soundness and completeness theorems are formalized. For this, we present two lemmas, usually omitted, which are needed. This development also contains a formalization of basic results on operators and their fixpoints in a general setting. All the proofs of the results, presented here, have been checked with the proof assistant Coq.

Term graph narrowing

We introduce term graph narrowing as an approach for solving equations by transformations on term graphs. Term graph narrowing combines term graph rewriting with first-order term unification. Our main result is that this mechanism is complete for all term rewriting systems over which term graph rewriting is normalizing and confluent. This includes, in particular, all convergent term rewriting systems. Completeness means that for every solution of a given equation, term graph narrowing can find a more general solution. The general motivation for using term graphs instead of terms is to improve efficiency: sharing common subterms saves space and avoids the repetition of computations.