init(input:output:convolutionWeights:convolutionBias:convolutionStride:convolutionDilationStride:convolutionPadding:normalization:normalizationBeta:normalizationGamma:normalizationMomentum:normalizationEpsilon:normalizationActivation:filterParameters:)
Returns a new fused, convolution normalization layer.
Declaration
convenience init?(input: BNNSNDArrayDescriptor, output: BNNSNDArrayDescriptor, convolutionWeights: BNNSNDArrayDescriptor, convolutionBias: BNNSNDArrayDescriptor?, convolutionStride: (x: Int, y: Int), convolutionDilationStride: (x: Int, y: Int), convolutionPadding: BNNS.ConvolutionPadding, normalization: BNNS.NormalizationType, normalizationBeta: BNNSNDArrayDescriptor, normalizationGamma: BNNSNDArrayDescriptor, normalizationMomentum: Float, normalizationEpsilon: Float, normalizationActivation: BNNS.ActivationFunction, filterParameters: BNNSFilterParameters? = nil)Parameters
- input:
The descriptor of the input.
- output:
The descriptor of the output.
- convolutionWeights:
The descriptor of the convolution weights.
- convolutionBias:
The descriptor of the convolution bias.
- convolutionStride:
The width and height increments of the input image.
- convolutionDilationStride:
The width and height increments between elements in the input image during convolution.
- convolutionPadding:
The padding, which is the number of virtual zeros added to the sides of each channel.
- normalization:
An enumeration that specifies the normalization type.
- normalizationBeta:
The descriptor of the normalization beta.
- normalizationGamma:
The descriptor of the normalization gamma.
- normalizationMomentum:
A value between 0 and 1 that the normalization operation uses to update the moving mean and moving variance during training.
- normalizationEpsilon:
The epsilon in the computation of the standard deviation.
- normalizationActivation:
The activation function that the layer applies to the output.
- filterParameters:
The filter runtime parameters.