GCN

object GCN
Companion:
class
trait Product
trait Mirror
class Object
trait Matchable
class Any
GCN.type

Type members

Inherited types

type MirroredElemLabels <: Tuple

The names of the product elements

The names of the product elements

Inherited from:
Mirror
type MirroredLabel <: String

The name of the type

The name of the type

Inherited from:
Mirror

Value members

Concrete methods

def computeSparseAdjacency[S : Sc](valueOpt: STenOptions, edgeI: STen, edgeJ: STen, numNodes: Long): (Constant, Constant)
def gcn[S : Sc](in: Int, out: Int, tOpt: STenOptions, dropout: Double, nonLinearity: Boolean): GCN[ResidualModule[EitherModule[Variable, Variable, Seq4[Variable, Variable, Variable, Variable, Variable, Linear, BatchNorm, Fun, Dropout], Seq2[Variable, Variable, Variable, Linear, BatchNorm]]]]
def gcnAggregation[S : Sc](nodeFeatures: Variable, edgeI: STen, edgeJ: STen): Variable

Performs D^-0.5 (A+A'+I) D^-0.5 W where are node features (N x D) A is the asymmetric adjacency matrix without self loops, elements in {0,1} I is identity D is degree(A+I)

Performs D^-0.5 (A+A'+I) D^-0.5 W where are node features (N x D) A is the asymmetric adjacency matrix without self loops, elements in {0,1} I is identity D is degree(A+I)

Value parameters:
edgeList

N x 2 long tensor the edges in A (asymmetric, no diagonal)

nodeFeatures

N x D node features

Returns:

N x D aggregated features

def gcnAggregation[S : Sc](nodeFeatures: Variable, degrees: Variable, a: Variable): Variable

Implicits

Implicits

implicit def load[M <: Module : Load]: Load[GCN[M]]
implicit def trainingMode[M <: Module : TrainingMode]: TrainingMode[GCN[M]]