provingground
package provingground
This is work towards automated theorem proving based on learning, using homotopy type theory (HoTT) as foundations and natural language processing.
The implementation of homotopy type theory is split into:
- the object HoTT with terms, types, functions and dependent functions, pairs etc
- the package induction with general inductive types and recursion/induction on these.
The learning package has the code for learning.
Scala code, including the
library, is integrated with homotopy type theory
in the scalahott packagespire
We have implemented a functor based approach to translation in the translation
package, used for
as well as nlp
and serialization
.parsing
The library package is contains basic structures implemented in HoTT.
- Alphabetic
- By Inheritance
- provingground
- AnyRef
- Any
- Hide All
- Show All
- Public
- Protected
Package Members
- package andrewscurtis
- package examples
- package induction
Much of the richness of HoTT is in the definitions of
(and their indexed counterparts) and of (dependent) functions on these by recursion and induction These are implemented using several layers of recursive definitions and diagonals (i.e., fixed points).Inductive types
Much of the richness of HoTT is in the definitions of
(and their indexed counterparts) and of (dependent) functions on these by recursion and induction These are implemented using several layers of recursive definitions and diagonals (i.e., fixed points). In HoTT, recursion and induction are applications of (dependent) functionsInductive types
rec_W,X
andind_W, Xs
to the
.definition data
It is useful to capture information regarding inductive types and the recursion and induction functions in scala types. Our implementation is designed to do this.
Inductive Type Definitions
Inductive types are specified by introduction rules. Each introduction rule is specified in ConstructorShape (without specifying the type) and ConstructorTL including the specific type. The full definition is in ConstructorSeqTL.
Recursion and Induction functions
These are defined recursively, first for each introduction rule and then for the inductive type as a whole. A subtlety is that the scala type of the
rec_W,X
andinduc_W, Xs
functions depends on the scala type of the codomainX
(or familyXs
). To make these types visible, some type level calculations using implicits are done, giving traits ConstructorPatternMap and ConstructorSeqMap that have recursive definition of the recursion and induction functions, the former for the case of a single introduction rule. Traits ConstructorSeqMapper and ConstructorPatternMapper provide the lifts.Indexed Versions
There are indexed versions of all these definitions, to work with indexed inductive type families.
- package interface
- package learning
- package library
- package scalahott
- package scratch
- package translation
Translation primarily using a functorial framework - see Translator$, for natural language processing as well as serialization, formatted output, parsing, interface with formal languages etc.
Translation primarily using a functorial framework - see Translator$, for natural language processing as well as serialization, formatted output, parsing, interface with formal languages etc.
Besides the Translator framework and helper typeclasses is Functors, several structures for concrete languages including our implementation of HoTT are in this package.
Type Members
- sealed trait Context extends AnyRef
- final case class FiniteDistribution[T](pmf: Vector[Weighted[T]]) extends AnyVal with ProbabilityDistribution[T] with Product with Serializable
Finite distributions, often supposed to be probability distributions, but may also be tangents to this or intermediates.
Finite distributions, often supposed to be probability distributions, but may also be tangents to this or intermediates.
- pmf
probability mass function, may have same object split.
- case class LinearStructure[A](zero: A, sum: (A, A) => A, mult: (Double, A) => A) extends Product with Serializable
- case class PickledWeighted(elem: String, weight: Double) extends Product with Serializable
- trait ProbabilityDistribution[A] extends Any
A probability distribution, from which we can pick values at random (the only abstract method).
A probability distribution, from which we can pick values at random (the only abstract method). We can obtain a random variable from this, which is an iterator.
- trait Subst[A] extends AnyRef
allows substitution of a
Term
by another. - trait SubstImplicits extends AnyRef
- sealed trait TermList[A] extends Subst[A]
allows substitution of a
Term
by another, as well as mapping to a vector of terms chiefly subtypes ofTerm
andHList
s of these; - trait TermListImplicits extends SubstImplicits
- case class Weighted[T](elem: T, weight: Double) extends Product with Serializable
Value Members
- object Base
- object Context
- object FiniteDistribution extends Serializable
- object Frankl
- object HoTT
Core of Homotopy Type Theory (HoTT) implementation.
Core of Homotopy Type Theory (HoTT) implementation. Includes: - terms : Term, - types : Typ - universes - functions and dependent functions (see [FuncLike], [Func]) - function types FuncTyp and pi-types PiDefn, - lambda definitions LambdaLike, - pairs PairTerm and dependent pairs DepPair - product types ProdTyp and sigma types SigmaTyp - Coproduct types PlusTyp, the Unit type Unit and the empty type Zero - recursion and induction functions for products, coproducts
General inductive types are
implemented here, but in the induction package.not
- object JvmUtils
- object LinearStructure extends Serializable
- object MereProposition
- object PickledWeighted extends Serializable
- object ProbabilityDistribution
- object Subst
- object TermList extends TermListImplicits
- object Utils
- object Weighted extends Serializable