One Logo Pool Ball
X
Github
Discord

System

Accelerate · IOS API REFERENCE

BNNS.ConvolutionLayer

class · iOS 14.0+ · SDK 27.1

class ConvolutionLayer

A convolution layer.

Members (2)

method · iOS 14.0+

applyBackward

TypeScript
applyBackward(batchSize: number, input: BNNSNDArrayDescriptor, output: BNNSNDArrayDescriptor, outputGradient: BNNSNDArrayDescriptor, inputGradient: BNNSNDArrayDescriptor, weightsGradient?: (BNNSNDArrayDescriptor) | null, biasGradient?: (BNNSNDArrayDescriptor) | null): void
Swift declaration
func applyBackward(batchSize: Int, input: BNNSNDArrayDescriptor, output: BNNSNDArrayDescriptor, outputGradient: BNNSNDArrayDescriptor, generatingInputGradient inputGradient: BNNSNDArrayDescriptor, generatingWeightsGradient weightsGradient: BNNSNDArrayDescriptor? = nil, generatingBiasGradient biasGradient: BNNSNDArrayDescriptor? = nil) throws

Applies the layer backward to generate input, output, weights, and bias gradients. - Parameter batchSize: The number of input-output pairs to process. - Parameter input: The input descriptor. - Parameter output: The output descriptor. - Parameter outputGradient: The output gradient descriptor. - Parameter inputGradient: The input gradient descriptor. - Parameter weightsGradient: The weights gradient descriptor. - Parameter biasGradient: The bias gradient descriptor.

init · iOS 14.0+

init

TypeScript
static init(convolutionType: BNNS_ConvolutionType, input: BNNSNDArrayDescriptor, weights: BNNSNDArrayDescriptor, output: BNNSNDArrayDescriptor, bias: (BNNSNDArrayDescriptor) | null, padding: BNNS_ConvolutionPadding, activation: BNNS_ActivationFunction, groupCount: number, stride: { "x": number; "y": number }, dilationStride: { "x": number; "y": number }, filterParameters?: (BNNSFilterParameters) | null): (BNNS_ConvolutionLayer) | null
Swift declaration
convenience init?(type convolutionType: BNNS.ConvolutionType, input: BNNSNDArrayDescriptor, weights: BNNSNDArrayDescriptor, output: BNNSNDArrayDescriptor, bias: BNNSNDArrayDescriptor?, padding: BNNS.ConvolutionPadding, activation: BNNS.ActivationFunction, groupCount: Int, stride: (x: Int, y: Int), dilationStride: (x: Int, y: Int), filterParameters: BNNSFilterParameters? = nil)

Creates a new convolution layer. - Parameter convolutionType: The convolution type. - Parameter input: The input descriptor. - Parameter weights: The weights descriptor. - Parameter output: The output descriptor. - Parameter bias: The bias descriptor. - Parameter padding: The number of zeros virtually added to the edges of the input. - Parameter activation: The activation function to apply to the result. - Parameter stride: Increment in the input image. - Parameter dilationStride: increment in the input image while convolving with the kernel. - Parameter filterParameters: Runtime filter parameters. - Returns: A new convolution layer.

Skipped declarations

No declarations were skipped in this snapshot.

All Accelerate types →