package learning
- Alphabetic
- Public
- Protected
Type Members
- trait AdjDiffbleFunction[A, B] extends AnyRef
- case class AnswerFromPost[P, U, W, ID](func: (P) => U)(implicit pw: Postable[P, W, ID]) extends Product with Serializable
Looking up an answer to a query for
U
from posts of typeP
meant for querying for the latest post etc.Looking up an answer to a query for
U
from posts of typeP
meant for querying for the latest post etc.- func
transformation of the content of the post
- pw
postability
- trait ApplnInverse extends AnyRef
inverses under application with an without unification
- class BackPaths[A] extends AnyRef
- case class BasicDeducer(applnWeight: Double = 0.2, lambdaWeight: Double = 0.2, piWeight: Double = 0.2, varWeight: Double = 0.3, vars: Vector[Term] = Vector()) extends Product with Serializable
- trait BiPostable[P, W, ID] extends Postable[P, W, ID]
- case class BufferedRun[A](iter: Iterator[A]) extends Product with Serializable
- trait BuffersJson[W, B] extends AnyRef
- trait BuildPostable[W, B, P] extends AnyRef
typeclass for building HLists of postables (and other things) based on HLists of buffers, but formally just returns object of type
P
- trait BuildQuery[Q, W, ID, PList <: HList] extends AnyRef
- case class Collated[P](contents: Vector[P]) extends Product with Serializable
- class CompositeProver[D] extends AnyRef
- trait Conditioning[D[_]] extends AnyRef
typeclass for being able to condition
- trait ContextExport[Boat, D[_]] extends AnyRef
changes in the distribution-like object other than those induced by object level map while exporting from an island typically a change in context for variables representing terms in a context
changes in the distribution-like object other than those induced by object level map while exporting from an island typically a change in context for variables representing terms in a context
- Boat
the boat for an island
- D
the distribution-like object
- class CounterGlobalID extends GlobalID[(Int, Int)]
allows posting globally and keeps count without stroing anything
- trait DataGetter[P, W, ID] extends AnyRef
- case class Deducer(applnWeight: Double = 0.2, lambdaWeight: Double = 0.2, piWeight: Double = 0.2, varWeight: Double = 0.3, vars: Vector[Weighted[Term]] = Vector(), propDecay: Double = 0.5, cutoff: Double = 0.01, feedbackScale: Double = 0.1, abstractionWeight: Double = 0.3, genMemory: Double = 0.8) extends Product with Serializable
- class DeducerBuffer extends AnyRef
- class DeducerSource extends AnyRef
- class DerivedEquations extends AnyRef
- trait DiffbleFunction[A, B] extends (A) => B
- trait DistributionState[State, D[_]] extends StateDistribution[State, D]
typeclass for providing distributions from a state and modifying a state from distributions
- class DistributionStateEvolver[State, D[_]] extends StateEvolver[State]
- case class EgUnif(n: Int) extends Product with Serializable
- trait Empty[D[_]] extends AnyRef
- case class EntropyAtomWeight(h0: Double, kl0: Double, p0: Double, q0: Double, initWeight: Double, hW: Double, klW: Double, crossKL: Boolean = true) extends Product with Serializable
Spire gradient learning for tradeoff between generation entropy and theorem-proof relative entropy
Spire gradient learning for tradeoff between generation entropy and theorem-proof relative entropy
- h0
the initial generation entropy
- kl0
the initial relative entropy between theorems and proofs
- p0
the initial theorem weight of the element
- q0
the initial proof weight of the element
- initWeight
the weight of the initial distribution in generation of terms
- case class EqDistMemo[State](varDists: Map[(State, RandomVar[_]), (Double, FiniteDistribution[_], Set[EquationNode])], nodeDists: Map[(State, GeneratorNode[_]), (Double, FiniteDistribution[_], Set[EquationNode])]) extends Product with Serializable
- case class Equation(lhs: Expression, rhs: Expression) extends Product with Serializable
- class EquationExporter extends AnyRef
- case class EquationNode(lhs: Expression, rhs: Expression) extends Product with Serializable
- abstract class ErasablePostBuffer[P, ID] extends GlobalPost[P, ID]
- trait EvolvedEquations[State] extends AnyRef
- case class EvolvedState(init: TermState, result: TermState, params: TermGenParams, epsilon: Double) extends EvolvedStateLike with Product with Serializable
- trait EvolvedStateLike extends AnyRef
- class EvolverEquations[F] extends AnyRef
variables for probabilities and equations for consistency
- trait EvolverSupport extends AnyRef
Support of the final distribution of an
evolver
, i.e., a generative model for terms - sealed trait EvolverVariables extends AnyRef
Variables in an evolver; not all need to be used in a give case
- sealed trait Expression extends AnyRef
- class ExpressionEquationIndexifier extends AnyRef
Maps equations based on expressions in terms, randomvariables etc.
Maps equations based on expressions in terms, randomvariables etc. to equations based on indices only, for rapid computation. Also provides some helpers to relate index combinations and expressions.
- trait ExpressionEquationSolver extends AnyRef
Solve equations based on expressions (approximately), and compute various related quantities The actual solving is done after translating the equations to ones based on indices, for the sake of efficiency.
Solve equations based on expressions (approximately), and compute various related quantities The actual solving is done after translating the equations to ones based on indices, for the sake of efficiency. Other methods let one construct new expression-equation-solvers, including by export.
- trait ExstParMap extends AnyRef
- trait FallBackLookups extends AnyRef
- class FatExprEquations extends ExpressionEquationIndexifier
- case class FineDeducer(applnWeight: Double = 0.1, lambdaWeight: Double = 0.1, piWeight: Double = 0.1, varWeight: Double = 0.3, unifyWeight: Double = 0.5) extends FineEvolver with Product with Serializable
A refined deducer, i.e., evolution of terms and derivatives of evolution.
A refined deducer, i.e., evolution of terms and derivatives of evolution. Various evolutions are defined mutually recursively - of functions, of types, of terms of a type, of type families and of all terms. Derivatives are defined mutually recursively with the evolutions.
This is refined so that, for example, arguments are chosen conditionally from the domain of a function.
- varWeight
weight of a variable inside a lambda
- class FineDeducerStep[X[_]] extends AnyRef
Monad based step for deducer, to be used, for example, with Monix
Task
- class FineEvolver extends AnyRef
A refined evolver, i.e., evolution of terms and subclasses such as types.
A refined evolver, i.e., evolution of terms and subclasses such as types. Various evolutions are defined mutually recursively - of functions, of types, of terms of a type, of type families and of all terms. Derivatives are defined mutually recursively with the evolutions.
This is refined so that, for example, arguments are chosen conditionally from the domain of a function.
- class ForceDirected[A, V] extends AnyRef
- case class ForceDirectedVectors[A](vertices: Set[A], edges: Map[(A, A), Double], coulomb: Double, elasticity: Double, dim: Int, R: Double, scale: Double, decay: Double, steps: Int) extends ForceDirected[A, Vector[Double]] with Product with Serializable
- case class GatherMapPost[P](contents: Set[P]) extends Product with Serializable
- case class GatherPost[P](contents: Set[P]) extends Product with Serializable
- abstract class GenMonixFiniteDistribution[State] extends AnyRef
- abstract class GenMonixFiniteDistributionEq[State] extends AnyRef
- abstract class GenTruncatedFiniteDistribution[State] extends AnyRef
- case class GeneratorData[V](nodeCoeffs: Vector[(GeneratorNodeFamily[_ <: HList, _], V)], varsToResolve: Vector[RandomVar[_]]) extends Product with Serializable
- case class GeneratorEquations[State](nodeCoeffSeq: NodeCoeffSeq[State, Double], varWeight: Double, initState: State, finalState: State)(implicit sd: StateDistribution[State, FiniteDistribution]) extends EvolvedEquations[State] with Product with Serializable
- sealed trait GeneratorNode[+O] extends GeneratorNodeFamily[HNil, O]
A formal node for describing recursive generation.
A formal node for describing recursive generation. Can have several inputList, encoded as an HList, but has just one output.
- sealed trait GeneratorNodeFamily[Dom <: HList, +O] extends AnyRef
Family of (recursive) generation functions, either a function giving a family or a single {{GeneratorNode}}, which is the interesting case.
- case class GeneratorTF[State](nodeCoeffSeq: NodeCoeffSeq[State, Double], varWeight: Double, initState: State, finalState: State)(implicit sd: StateDistribution[State, FiniteDistribution]) extends Product with Serializable
- case class GeneratorVariables[State](nodeCoeffSeq: NodeCoeffSeq[State, Double], state: State)(implicit sd: StateDistribution[State, FiniteDistribution]) extends Product with Serializable
resolving a general specification of a recursive generative model as finite distributions, depending on truncation; the coefficients of the various generator nodes should be
Double
resolving a general specification of a recursive generative model as finite distributions, depending on truncation; the coefficients of the various generator nodes should be
Double
- State
scala type of the initial state
- nodeCoeffSeq
the various generator nodes with coefficients for various random variables and families
- sd
finite distributions from the initial state corresponding to random variables and families
- trait GlobalID[ID] extends AnyRef
- trait GlobalPost[P, ID] extends AnyRef
Allows posting any content, typically just returns an ID to be used by something else.
- class GraphEmbeddingLogisitic extends AnyRef
- trait HistoryGetter[W, B, ID] extends AnyRef
- class HoTTPostWeb extends AnyRef
Better building of post webs, without depending on separate lists for implicits and history (history to be done).
- class HoTTWebSession extends SimpleSession[HoTTPostWeb, (Int, Int)]
- class IndexEquationMapSolver extends ExpressionEquationIndexifier
Seek an approximately stable distribution for equations given in terms of indices, hence solving the equations (approximately).
Seek an approximately stable distribution for equations given in terms of indices, hence solving the equations (approximately). Here stable distribution is under the iteration to solve equations viewing them as fixed points. This is an implementation based on maps and keeping track of supports, which is not used as Vectors were more efficient.
- class IndexEquationSolver extends ExpressionEquationIndexifier
Seek an approximately stable distribution for equations given in terms of indices, hence solving the equations (approximately).
Seek an approximately stable distribution for equations given in terms of indices, hence solving the equations (approximately). Here stable distribution is under the iteration to solve equations viewing them as fixed points. A few quantitites related to the solutions, or the equations themselves, are also computed.
- class IndexedBacktrace extends AnyRef
Computation of probabilities of elements in equations for other elements, and of upper bounds on distances, based on equations in indices.
- case class IntElem(n: Int) extends SequenceElement with Product with Serializable
- trait IterantRunner[S, A, M] extends AnyRef
- case class LatestAnswer[Q, W, ID](answers: Vector[AnswerFromPost[_, Q, W, ID]])(implicit h: PostHistory[W, ID]) extends LocalQueryable[Q, W, ID] with Product with Serializable
Queries answered by taking the latest out of possibly many choices of posts to be looked up and transformed.
Queries answered by taking the latest out of possibly many choices of posts to be looked up and transformed.
- answers
various ways in which the query can be answered
- h
the history
- case class LatestTagged[Q, W, ID]()(implicit tag: scala.reflect.api.JavaUniverse.TypeTag[Q], h: PostHistory[W, ID]) extends LocalQueryable[Q, W, ID] with Product with Serializable
- case class LocalProver(initState: TermState = TermState(FiniteDistribution.empty, FiniteDistribution.empty), tg: TermGenParams = TermGenParams(), cutoff: Double = math.pow(10, -4), genMaxDepth: Option[Int] = None, limit: FiniteDuration = 12.minutes, maxRatio: Double = 1.01, resolution: Double = 0.0, scale: Double = 1.0, steps: Int = 10000, maxDepth: Int = 10, hW: Double = 1, klW: Double = 1, smoothing: Option[Double] = None, relativeEval: Boolean = false, stateFromEquation: Boolean = false, exponent: Double = 0.5, decay: Double = 1, maxTime: Option[Long] = None) extends LocalProverStep with Product with Serializable
Collect local/generative/tactical proving; this includes configuration and learning but excludes strategy and attention.
Collect local/generative/tactical proving; this includes configuration and learning but excludes strategy and attention. This can be called in a loop generating goals based on unproved theorems, using representation/deep learning with attention focussed or interactively.
- trait LocalProverStep extends AnyRef
- trait LocalQueryable[U, W, ID] extends AnyRef
Typeclass for being able to query W for a vector of elements of type Q at an index
- case class LocalTangentProver(initState: TermState = TermState(FiniteDistribution.empty, FiniteDistribution.empty), initEquations: Set[EquationNode], tangentState: TermState, tg: TermGenParams = TermGenParams(), cutoff: Double = math.pow(10, -4), genMaxDepth: Option[Int] = None, limit: FiniteDuration = 3.minutes, maxRatio: Double = 1.01, resolution: Double = 0.0, scale: Double = 1.0, steps: Int = 10000, maxDepth: Int = 10, hW: Double = 1, klW: Double = 1, smoothing: Option[Double] = None, relativeEval: Boolean = false, stateFromEquation: Boolean = false, exponent: Double = 0.5, decay: Double = 1, maxTime: Option[Long] = None) extends LocalProverStep with Product with Serializable
- case class MemoState[D[_], V, C](randVarVals: Set[Value[_, D]], genData: GeneratorData[V], context: C) extends Product with Serializable
- case class MonixFiniteDistribution[State](nodeCoeffSeq: NodeCoeffSeq[State, Double], varWeight: Double)(implicit sd: StateDistribution[State, FiniteDistribution]) extends GenMonixFiniteDistribution[State] with Product with Serializable
resolving a general specification of a recursive generative model as finite distributions, depending on truncation; the coefficients of the various generator nodes should be
Double
resolving a general specification of a recursive generative model as finite distributions, depending on truncation; the coefficients of the various generator nodes should be
Double
- State
scala type of the initial state
- nodeCoeffSeq
the various generator nodes with coefficients for various random variables and families
- sd
finite distributions from the initial state corresponding to random variables and families
- case class MonixFiniteDistributionEq[State](nodeCoeffSeq: NodeCoeffSeq[State, Double], varWeight: Double, limit: FiniteDuration = 3.minutes)(implicit sd: StateDistribution[State, FiniteDistribution]) extends GenMonixFiniteDistributionEq[State] with Product with Serializable
resolving a general specification of a recursive generative model as finite distributions, depending on truncation; the coefficients of the various generator nodes should be
Double
resolving a general specification of a recursive generative model as finite distributions, depending on truncation; the coefficients of the various generator nodes should be
Double
- State
scala type of the initial state
- nodeCoeffSeq
the various generator nodes with coefficients for various random variables and families
- sd
finite distributions from the initial state corresponding to random variables and families
- trait MonixSamples extends Samples[Task]
- case class MonixTangentFiniteDistribution[State](nodeCoeffSeq: NodeCoeffSeq[State, Double], varWeight: Double, baseState: State)(implicit sd: StateDistribution[State, FiniteDistribution]) extends GenMonixFiniteDistribution[State] with Product with Serializable
resolving a general specification of a recursive generative model as finite distributions, depending on truncation; the coefficients of the various generator nodes should be
Double
resolving a general specification of a recursive generative model as finite distributions, depending on truncation; the coefficients of the various generator nodes should be
Double
- State
scala type of the initial state
- nodeCoeffSeq
the various generator nodes with coefficients for various random variables and families
- sd
finite distributions from the initial state corresponding to random variables and families
- case class MonixTangentFiniteDistributionEq[State](nodeCoeffSeq: NodeCoeffSeq[State, Double], varWeight: Double, baseState: State, baseEquations: Set[EquationNode], limit: FiniteDuration = 3.minutes)(implicit sd: StateDistribution[State, FiniteDistribution]) extends GenMonixFiniteDistributionEq[State] with Product with Serializable
resolving a general specification of a recursive generative model as finite distributions, depending on truncation; the coefficients of the various generator nodes should be
Double
resolving a general specification of a recursive generative model as finite distributions, depending on truncation; the coefficients of the various generator nodes should be
Double
- State
scala type of the initial state
- nodeCoeffSeq
the various generator nodes with coefficients for various random variables and families
- sd
finite distributions from the initial state corresponding to random variables and families
- case class NextSample(p: FiniteDistribution[Term], ded: FineDeducer = FineDeducer(), vars: Vector[Term] = Vector(), size: Int = 1000, derTotalSize: Int = 1000, epsilon: Double = 0.2, inertia: Double = 0.3) extends Product with Serializable
- sealed trait NodeCoeffSeq[State, V] extends AnyRef
- sealed trait NodeCoeffs[State, V, RDom <: HList, Y] extends AnyRef
- class ParDistEq extends RecParDistEq
- class ParDistEqMemo extends AnyRef
- case class ParMapState(termDist: ParMap[Term, Double], typDist: ParMap[Typ[Term], Double], vars: Vector[Term] = Vector(), inds: ParMap[ExstInducDefn, Double] = ParMap(), goalDist: ParMap[Typ[Term], Double] = ParMap(), context: Context = Context.Empty) extends Product with Serializable
- class ParTangentDistEq extends RecParDistEq
- case class ParTermState(ts: TermState) extends Product with Serializable
- case class PickledTermPopulation(termsByType: Map[String, Vector[PickledWeighted]], types: Vector[PickledWeighted], thmsByProofs: Vector[PickledWeighted], vars: Vector[PickledWeighted], lambdaWeight: Double, piWeight: Double) extends Product with Serializable
- abstract class PostBuffer[P, ID] extends GlobalPost[P, ID]
A buffer for storing posts, extending
GlobalPost
which supplies an ID - case class PostData[P, W, ID](content: P, id: ID)(implicit pw: Postable[P, W, ID]) extends Product with Serializable
Data for a post, including the implicit saying it is postable
Data for a post, including the implicit saying it is postable
- content
the content of the post
- id
the index of the post, returned after posting
- trait PostDiscarder[P, ID] extends GlobalPost[P, ID]
- trait PostHistory[W, ID] extends AnyRef
- sealed trait PostMaps[P] extends AnyRef
- sealed trait PostResponse[W, ID] extends AnyRef
Response to a post, generating one or more posts or just a callback; this exists mainly for nicer type collections.
- trait Postable[P, W, ID] extends AnyRef
Typeclass for being able to post with content type P in W
- case class PreviousPosts[P](contents: Vector[P]) extends Product with Serializable
- case class ProductIndexExpression(constant: Double, indices: Vector[Int], negIndices: Vector[Int]) extends Product with Serializable
- case class ProofEntropies(gens: Map[Term, Double], theorems: Map[Typ[Term], Double], proofs: Map[Term, Double], genWeight: Double) extends Product with Serializable
- case class QueryBaseState(init: TermState) extends Product with Serializable
- case class QueryEquations(equations: Set[Equation]) extends Product with Serializable
- sealed trait QueryFromPosts[Q, PList <: HList] extends AnyRef
- case class QueryInitState(init: TermState) extends Product with Serializable
- case class QueryOptions[Q, W, ID](answers: (PostData[_, W, ID]) => Option[Q], modifiers: (PostData[_, W, ID]) => (Q) => Q) extends Product with Serializable
- case class QueryProver(lp: LocalProver) extends Product with Serializable
- trait Queryable[U, W] extends AnyRef
Typeclass for being able to query W for type Q
- class RandomVar[+O] extends RandomVarFamily[HNil, O]
A formal Random Variable up to equality of distribution.
A formal Random Variable up to equality of distribution. May actually have representations instead of distributions, for example.
- class RandomVarFamily[Dom <: HList, +O] extends AnyRef
A formal family of Random Variables up to equality of distribution.
A formal family of Random Variables up to equality of distribution. May actually have representations instead of distributions, for example.
- sealed trait RandomVarList[U <: HList] extends AnyRef
List of random variables, e.g.
List of random variables, e.g. input of simple generator nodes.
- abstract class RandomVarMemo[D[_]] extends RandomVarValues[D]
- trait RandomVarValues[D[_]] extends AnyRef
- case class RandomVector[X](base: RandomVar[X]) extends RandomVar[Vector[X]] with Product with Serializable
distributions of vectors from a base distribution
distributions of vectors from a base distribution
- X
scala type of the base
- base
the base distribution
- trait RecParDistEq extends AnyRef
- sealed trait RecursiveGeneratorNode[State, +O] extends GeneratorNode[O] with RecursiveGeneratorNodeFamily[HNil, State, O]
- sealed trait RecursiveGeneratorNodeFamily[Dom <: HList, State, +O] extends GeneratorNodeFamily[Dom, O]
- final case class Representation[T](rep: Vector[WeightVect[T]]) extends AnyVal with Product with Serializable
- class RepresentationLearner[A] extends AnyRef
- trait Samples[X[_]] extends TangSamples[X]
- trait SimpleDataGetter extends AnyRef
- class SimpleSession[W, ID] extends AnyRef
A simple session to post stuff, call responses and if these responses generate posts, recursively call itself for them.
- case class SnapShot[X, P](state: X, name: String, loops: Int, param: P) extends Product with Serializable
- case class SomePost[P](content: P) extends Product with Serializable
Wrapper to query for all posts, even after the query position or in a different thread
Wrapper to query for all posts, even after the query position or in a different thread
- content
the wrapped content
- sealed trait Sort[-S, +T] extends AnyRef
A
sort
, i.e.A
sort
, i.e. type refining a scala type. Can also be used for conditioning, giving one distribution from another.- T
scala type elements of this Sort
- sealed trait SortList[U <: HList] extends AnyRef
List of Sorts, used as domains for families.
- class SpireExprEquations extends ExpressionEquationIndexifier
- case class SpireGradient(vars: Vector[Expression], p: Map[Expression, Double], cost: Expression) extends Product with Serializable
- case class SplitPost[P, Q, W, ID](content: P, transformation: (Q) => P)(implicit evidence$10: scala.reflect.api.JavaUniverse.TypeTag[P], evidence$11: scala.reflect.api.JavaUniverse.TypeTag[Q], evidence$12: scala.reflect.api.JavaUniverse.TypeTag[W], evidence$13: scala.reflect.api.JavaUniverse.TypeTag[ID], pw: Postable[P, W, ID], qq: LocalQueryable[Q, W, ID]) extends Product with Serializable
Wrapper for post content that should be posted, with the previous elements of the same type also posted, in general with transformations (e.g.
Wrapper for post content that should be posted, with the previous elements of the same type also posted, in general with transformations (e.g. rescale)
- content
the content to be posted
- transformation
transformations of other posts, typically rescaling
- pw
postability of P
- trait StateDistribution[State, D[_]] extends AnyRef
typeclass for providing distributions from a state
- trait StateEvolver[State] extends AnyRef
- class StochasticLang[E] extends TruncatedDistributionLang[E]
- case class SumIndexExpression(terms: Vector[ProductIndexExpression]) extends Product with Serializable
- trait Support[D[_]] extends AnyRef
- case class TFData(vars: Map[VarVal[_], Double], varSets: Set[Set[VarVal[_]]], inFinalEvent: Map[Variable[_], Set[Variable[_]]], inFinalPairEvent: Map[Variable[_], Set[(Variable[_], Variable[_])]], equations: Set[Equation]) extends Product with Serializable
- case class TFDist[A](pmfMap: Map[A, Output[Double]]) extends Product with Serializable
- class TFEg extends AnyRef
- trait TangSamples[X[_]] extends AnyRef
- case class TangVec[+A](point: A, vec: A) extends Product with Serializable
- case class TangentFiniteDistribution[State](baseState: State, nodeCoeffSeq: NodeCoeffSeq[State, Double], varWeight: Double)(implicit sd: StateDistribution[State, FiniteDistribution]) extends GenTruncatedFiniteDistribution[State] with Product with Serializable
resolving a general specification of a recursive generative model as finite distributions, depending on truncation; the coefficients of the various generator nodes should be
Double
resolving a general specification of a recursive generative model as finite distributions, depending on truncation; the coefficients of the various generator nodes should be
Double
- State
scala type of the initial state
- nodeCoeffSeq
the various generator nodes with coefficients for various random variables and families
- sd
finite distributions from the initial state corresponding to random variables and families
- class TensorFlowExprEquations extends ExpressionEquationIndexifier
- class TensorFlowExprVarEquations extends ExpressionEquationIndexifier
- class TensorFlowFatExprEquations extends FatExprEquations
- case class TensorflowExpressions(tf: Ops, init: PartialFunction[Expression, Operand[TFloat32]]) extends Product with Serializable
- class TermBucket extends AnyRef
- trait TermEvolution extends AnyRef
- class TermEvolutionStep[X[_]] extends AnyRef
- class TermEvolver extends TermEvolution
- case class TermGenCost(ge: EvolvedEquations[TermState], varWeight: Double, hW: Double = 1, klW: Double = 1, eqW: Double = 1, epsilon: Double = math.pow(10, -5)) extends TermGenEqCost with Product with Serializable
- abstract class TermGenEqCost extends AnyRef
- case class TermGenParams(appW: Double = 0.1, unAppW: Double = 0.1, argAppW: Double = 0.1, lmW: Double = 0.1, piW: Double = 0.1, piTermW: Double = 0, termsByTypW: Double = 0.05, typFromFamilyW: Double = 0.05, sigmaW: Double = 0.05, recDefW: Double = 0, inducDefW: Double = 0, typAsCodW: Double = 0.05, targetInducW: Double = 0, varWeight: Double = 0.3, goalWeight: Double = 0.7, typVsFamily: Double = 0.5, negTargetW: Double = 0, solverW: Double = 0, contraW: Double = 0, solver: TypSolver = TypSolver.coreSolver) extends TermNodeCoeffSeq[TermState] with Product with Serializable
- case class TermGenParamsNodes(tg: TermGenParams) extends TermGeneratorNodes[TermState] with Product with Serializable
- class TermGeneratorNodes[InitState] extends AnyRef
Combining terms and subclasses to get terms, types, functions etc; these are abstract specifications, to be used for generating distributions, obtaining equations etc.
Combining terms and subclasses to get terms, types, functions etc; these are abstract specifications, to be used for generating distributions, obtaining equations etc.
- InitState
the initial state for the dynamics, equations etc
- case class TermLearner[F](supp: EvolverSupport, prob: (EvolverVariables) => F, apInv: ApplnInverse)(implicit evidence$3: Field[F], evidence$4: Trig[F]) extends EvolverEquations[F] with Product with Serializable
Adding equations from a simple generative model to EvolverEquations
- abstract class TermNodeCoeffSeq[State] extends AnyRef
- case class TermPopulation(termsByType: Map[Typ[Term], FiniteDistribution[Term]], types: FiniteDistribution[Typ[Term]], thmsByProofs: FiniteDistribution[Typ[Term]], vars: Vector[Weighted[Term]], lambdaWeight: Double, piWeight: Double, applnInvMap: InvMap = Vector()) extends Product with Serializable
- case class TermState(terms: FiniteDistribution[Term], typs: FiniteDistribution[Typ[Term]], vars: Vector[Term] = Vector(), inds: FiniteDistribution[ExstInducDefn] = FD.empty[ExstInducDefn], goals: FiniteDistribution[Typ[Term]] = FD.empty, context: Context = Context.Empty) extends TermsTypThms with Product with Serializable
A state, typically the initial state, for generating terms, types etc
A state, typically the initial state, for generating terms, types etc
- terms
distribution of terms
- typs
distribution of types
- vars
variables, over which we may take closures
- inds
inductive type definitions
- trait TermsTypThms extends AnyRef
- case class TheoremFeedback(fd: FiniteDistribution[Term], tfd: FiniteDistribution[Typ[Term]], vars: Vector[Term] = Vector(), scale: Double = 1.0, thmScale: Double = 0.3, thmTarget: Double = 0.2) extends Product with Serializable
feedback based on the term-type map as well as ensuring total weight of theorems is not small; various steps are explicit for exploration and debugging
- case class ThmEntropies(fd: FiniteDistribution[Term], varNames: Vector[Term] = Vector(), scale: Double = 1.0, thmScale: Double = 0.3, thmTarget: Double = 0.2) extends Product with Serializable
- final case class TruncDistVal[A](getFD: (Double) => Option[FiniteDistribution[A]]) extends AnyVal with Product with Serializable
- sealed trait TruncatedDistribution[A] extends AnyRef
- class TruncatedDistributionDomain[E] extends Domain[TruncatedDistribution[E]]
- class TruncatedDistributionExprPatterns[E] extends ExprPatterns[TruncatedDistribution[E]]
- class TruncatedDistributionLang[E] extends ExprLang[TruncatedDistribution[E]]
- case class TruncatedFiniteDistribution[State](nodeCoeffSeq: NodeCoeffSeq[State, Double], varWeight: Double)(implicit sd: StateDistribution[State, FiniteDistribution]) extends GenTruncatedFiniteDistribution[State] with Product with Serializable
resolving a general specification of a recursive generative model as finite distributions, depending on truncation; the coefficients of the various generator nodes should be
Double
resolving a general specification of a recursive generative model as finite distributions, depending on truncation; the coefficients of the various generator nodes should be
Double
- State
scala type of the initial state
- nodeCoeffSeq
the various generator nodes with coefficients for various random variables and families
- sd
finite distributions from the initial state corresponding to random variables and families
- trait TypSolver extends (Typ[Term]) => Option[Term]
- sealed abstract class TypedPostResponse[P, W, ID] extends PostResponse[W, ID]
Post response with type
P
of post as a type parameter - case class VarValueSet[D[_]](randVarVals: Set[Value[_, D]], randVarFamilyVals: Set[Value[_ <: HList, _, D]]) extends Product with Serializable
- case class WebState[W, ID](web: W, apexPosts: Vector[PostData[_, W, ID]] = Vector()) extends Product with Serializable
- case class WeightVect[T](elem: T, vect: Vector[Double]) extends Product with Serializable
- case class WeightedBackPaths[A, C](gens: Map[A, Vector[(C, Vector[A])]], coeffWeights: (C) => Double) extends BackPaths[A] with Product with Serializable
- case class WeightedBiPaths[A, C](gens: Map[A, Vector[(C, Vector[A])]], forwardGens: Map[A, Vector[(C, A)]], coeffWeights: (C) => Double, backWeight: Double) extends BackPaths[A] with Product with Serializable
- class WeightedTermBucket extends AnyRef
Deprecated Type Members
- case class ErasableWebBuffer[P, ID](buffer: ErasablePostBuffer[P, ID])(implicit pw: Postable[P, HoTTPost, ID]) extends Product with Serializable
- Annotations
- @deprecated
- Deprecated
(Since version soon) using HoTTPostWeb
- sealed trait GlobalProver[R] extends AnyRef
- Annotations
- @deprecated
- Deprecated
(Since version now) use composite prover
- class HoTTPost extends AnyRef
- Annotations
- @deprecated
- Deprecated
(Since version soon) migrating to HoTTPostWeb
- class QueryImplicit[Q, PList <: HList] extends QueryImplicitWrap[Q, Q, PList]
- Annotations
- @deprecated
- Deprecated
(Since version now) define query-from-posts and lift
- class QueryImplicitWrap[Q, T, PList <: HList] extends AnyRef
Have a wrapper type T for the query and make the companion object extend this class, giving query-from-posts as input (whose type should be deducable)
Have a wrapper type T for the query and make the companion object extend this class, giving query-from-posts as input (whose type should be deducable)
- Annotations
- @deprecated
- Deprecated
(Since version now) define query-from-posts and lift
- case class WebBuffer[P, ID](buffer: PostBuffer[P, ID])(implicit pw: Postable[P, HoTTPost, ID]) extends Product with Serializable
- Annotations
- @deprecated
- Deprecated
(Since version soon) using HoTTPostWeb
Value Members
- object AdjDiffbleFunction
- object ApplnInverse
- object BuffersJson
- object BuildPostable
- object BuildQuery
- case object Conditioning extends Product with Serializable
- object ContextExport
- object DE extends DerivedEquations
- object DataGetter extends SimpleDataGetter
- object Deducer extends Serializable
Generating terms from given ones using the main HoTT operations, and the adjoint of this generation.
Generating terms from given ones using the main HoTT operations, and the adjoint of this generation. This is viewed as deduction. Generation is a map on probability distributions, but the adjoint regards truncated distributions as the tangent space, and restricts domain to finite distributions.
- object DeducerSource
- object DerivedEquations
- object DiffbleFunction
- object DistributionState
- object Empty
- object EntropyAtomWeight extends Serializable
- object EqDistMemo extends Serializable
- object EquationExporter
- object EquationNode extends Serializable
- object EquationOps
- object ErasablePostBuffer
- object EvolverEquations
- object EvolverVariables
- object Expression
- object ExpressionEquationIndexifier
- object ExpressionEquationSolver
Working with expressions built from initial and final values of random variables, including in islands, given equations satisfied by these
- object ExstParMap
- object FieldGramSchmidt
- object FineDeducer extends Serializable
A refined deducer, i.e., evolution of terms and derivatives of evolution.
A refined deducer, i.e., evolution of terms and derivatives of evolution. Various evolutions are defined mutually recursively - of functions, of types, of terms of a type and of all terms. Derivatives are defined mutually recursively with the evolutions.
This is refined so that, for example, arguments are chosen conditionally from the domain of a function.
- object FineDeducerStep
- object FineProverTasks
A collection of functions to build provers; some are abstract methods for exploring, searching etc., while others generate terms and types, sometimes as derivatives.
A collection of functions to build provers; some are abstract methods for exploring, searching etc., while others generate terms and types, sometimes as derivatives. Some methods combine the two to give a ready to use function. These are based on TermEvolver and
Monix
. - object FiniteDistributionLearner
A combinator for learning systems with state finite distributions on vertices.
A combinator for learning systems with state finite distributions on vertices. Systems are built from components labeled by elements of a set M. The state also has weights for these components. The components are built from: moves (partial functions), partial combinations and islands.
- object GeneratorNode
- object GeneratorNodeFamily
- object GeneratorTF extends Serializable
- object GeneratorVariables extends Serializable
- object GeometricDistribution
An example, the geometric distribution in an abstract form
- object GramSchmidt
- object GraphEmbedding
- object HistoryGetter
- object HoTTBot
- object HoTTMessages
Messages to be posted for autonomous/interactive running.
Messages to be posted for autonomous/interactive running. These can be posted by a user, rule-based bot or a deep-learning system; from the point of view of deep learning, these are moves.
Some messages are instructions for what to do next; some report results and some provide data for use in the future.
Some types like
LocalProver
,ExpressionEval
,TermGenParams
etc can be posted in their raw form. - object HoTTPost
- object HoTTPostWeb
- object HoTTWebSession
- object HoTTgen
- object Hub
- object IndexEquationSolver
- object IndexedBacktrace
- object IterantRunner
- object IterateDyn
- object LemmaWeigths
- object LocalProver extends Serializable
- object LocalQueryable extends FallBackLookups
- object LocalTangentProver extends Serializable
- object MapVS
- object MemoState extends Serializable
- object MonixFieldGramSchmidt
- object MonixFiniteDistributionEq extends Serializable
- object MonixGramSchmidt
- object MonixProverTasks
A drop-in replacement for FineProverTasks to test the abstract generators approach; since testing is the goal redundancies and some comments have been removed.
- object MonixSamples
- object MonixTangentFiniteDistribution extends Serializable
- object MonixTangentFiniteDistributionEq extends Serializable
- object NodeCoeffSeq
- object NodeCoeffs
- object ParDistEq
- object ParGramSchmidt
- object ParMapState extends Serializable
- object ParStrategicProvers
- object ParTangentDistEq
- object ParTermState extends Serializable
- object PickledTermPopulation extends Serializable
- object PostBuffer
- object PostData extends Serializable
- object PostDiscarder
- object PostHistory
- object PostMaps
- object PostResponse
- object Postable
- object ProofEntropies extends Serializable
- object ProverTasks
- object QueryBaseState extends Serializable
- object QueryEquations extends Serializable
- object QueryFromPosts
- object QueryInitState extends Serializable
- object QueryOptions extends Serializable
- object QueryProver extends Serializable
- object Queryable
- object RandomVar
- object RandomVarFamily
- object RandomVarList
- object Representation extends Serializable
- object RepresentationLearner
- object Samples
- object SimpleEquations
- object Sort
- object SortList
- object SpireExprEquations
- object SpireGradient extends Serializable
- object SplitPost extends Serializable
- object StateDistribution
- object StrategicProvers
- object TFData extends Serializable
- object TFDist extends Serializable
- object TangVec extends Serializable
- object TangentFiniteDistribution extends Serializable
- object TensorFlowExprEquations
- object TermBucket
- object TermData
- object TermEvolutionStep
- object TermEvolver
- object TermGenJson
- object TermGenParams extends Serializable
- object TermGeneratorNodes
Combining terms and subclasses to get terms, types, functions etc; these are abstract specifications, to be used for generating distributions, obtaining equations etc.
Combining terms and subclasses to get terms, types, functions etc; these are abstract specifications, to be used for generating distributions, obtaining equations etc. The object contains helpers and other static objects.
- object TermNodeCoeffSeq
- object TermProver extends CompositeProver[TermResult]
- object TermRandomVars
- object TermState extends Serializable
- object TermsTypThms
- object TestCustomQuery
- object TruncDistVal extends Serializable
- object Truncate
- object TruncatedDistribution extends Functor[TruncatedDistribution]
- object TruncatedFiniteDistribution extends Serializable
- object TypSolver
- object TypedPostResponse
- object Unify
- object WeightVect extends Serializable
- object WeightedTermBucket
Deprecated Value Members
- object Collections
- Annotations
- @deprecated
- Deprecated
(Since version 17/5/2017) use spire
- object GlobalProver
- Annotations
- @deprecated
- Deprecated
(Since version now) use composite prover