API Reference - ConvolutionBlocks - AutomaticConvolution3D (AutoConvolution3D)

Constructors

new()

Creates a new convolution block object.


AutomaticConvolution3D.new({numberOfKernels: number, kernelDimensionSizeArray: {number}, strideDimensionSizeArray: {number}, learningRate: number, Optimizer: OptimizerObject, Regularizer: RegularizerObject, initializationMode: string}): ConvolutionBlockObject

Parameters:

  • numberOfKernels: The number of kernels to be used to extract the features from the input tensor.

  • kernelDimensionSizeArray: The dimension size for the kernel. The index of the array represents the dimension and the value represents the size for that particular dimension.

  • strideDimensionSizeArray: The dimension size for the stride. The index of the array represents the dimension and the value represents the size for that particular dimension. Note that the stride moves along one axis completely before incrementing to the next axis.

  • learningRate: The speed at which the model learns. Recommended that the value is set between (0 to 1).

  • Optimizer: The optimizer to be used.

  • Regularizer: The regularizer to be used.

  • initializationMode: The mode for the weights to be initialized. Available options are:

    • Zero

    • Random

    • RandomNormal

    • RandomUniformPositive

    • RandomUniformNegative

    • RandomUniformNegativeAndPositive

    • HeUniform

    • HeNormal

    • XavierUniform

    • XavierNormal

    • LeCunUniform

    • LeCunNormal

    • None

Returns:

  • ConvolutionBlock: The generated convolution block object.

Inherited From