Class ClubsTopology

Namespace

NanoQL.Experimental

Assembly

NanoQL.dll

The clubs topology, after H. M. Emara, "Adaptive Clubs-based Particle Swarm Optimization": each particle belongs to a default number of clubs, and its social best is the best of its clubs' members; a club's best leaves a random club (down to a minimum) and its worst joins one (up to a maximum), and every reset iterations the others return to the default: QuantLib's ClubsTopology.

[Experimental("NQL0001")]
public sealed class ClubsTopology : ParticleSwarmOptimizationTopology

Inheritance

object

ParticleSwarmOptimizationTopology

ClubsTopology

Inherited Members

object.GetType()

object.ToString()

object.Equals(object)

object.Equals(object, object)

object.ReferenceEquals(object, object)

object.GetHashCode()

Remarks

Fixes (Divergences/ParticleSwarmTopologyTests.cs): the initial clubs are drawn in [0, totalClubs − 1], where QuantLib's draws in [1, totalClubs] and reads one past each particle's clubs; a particle's social best is its best club's best particle, where QuantLib's takes the club's index for a particle; each club's worst is weighed over every member, where QuantLib's tests it only on a member that isn't the new best, so a club whose values fall in order names particle 0 its worst, a member or not; and a best particle leaves a club only above the minimum and a worst one joins only below the maximum, as the class documents, where QuantLib's never tests either, so a particle that stays best leaves every club and its social best is then the largest double. A club none of whose members passes QuantLib's comparisons, one with no members or whose personal bests are all NaN, has no best, and a particle reads only its clubs with a best, or its own personal best where none has one, where QuantLib's reads particle 0, a member or not, for such a club; a minimum below 1 is refused, as a particle in no club has no club-mates and so no social best, where QuantLib's best particle leaves every club and keeps particle 0's position at the largest double.

Its integers are std::uniform_int_distribution's on a std::mt19937 on the seed, drawn as MSVC draws them; the seed is required, QuantLib's default seeding from the clock. Each minimization starts the generator from the seed and the iterations from 0.

Constructors

ClubsTopology(int, int, int, int, int, long)

Initializes a new instance of the ClubsTopology class.

public ClubsTopology(int defaultClubs, int totalClubs, int maxClubs, int minClubs, int resetIteration, long seed)

Parameters

defaultClubs int

The clubs each particle starts in.

totalClubs int

The clubs.

maxClubs int

The most clubs a particle joins.

minClubs int

The fewest clubs a particle keeps.

resetIteration int

The iterations between resets to the default.

seed long

The seed, in [1, 4294967295].

Exceptions

ArgumentException

The club counts are out of order, with QuantLib's texts.

ArgumentOutOfRangeException

minClubs is below 1, or seed is outside [1, 4294967295].

Methods

Clone()

A copy for a minimization, unbound, its generators at their seeds.

public override ParticleSwarmOptimizationTopology Clone()

Returns

ParticleSwarmOptimizationTopology

The copy.

FindSocialBest()

Sets each particle's social best for the iteration (QuantLib's findSocialBest).

public override void FindSocialBest()

Exceptions

NanoQLException

The topology isn't bound to a minimization's swarm, where QuantLib reads clubs setSize hasn't made.

SetSize(int)

Initializes the state for the problem (QuantLib's setSize).

public override void SetSize(int M)

Parameters

M int

The swarm's size.