Package org.nd4j.autodiff.samediff.ops

Class SDCNN

  • public class SDCNN
extends SDOps

    • Constructor Detail

      • SDCNN

        public SDCNN​(SameDiff sameDiff)

    • Method Detail

      • avgPooling2d

        public SDVariable avgPooling2d​(SDVariable input,
                               Pooling2DConfig Pooling2DConfig)
        2D Convolution layer operation - average pooling 2d
        Parameters:
        input - the input to average pooling 2d operation - 4d CNN (image) activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        Pooling2DConfig - Configuration Object
        Returns:
        output Result after applying average pooling on the input (NUMERIC type)

      • avgPooling2d

        public SDVariable avgPooling2d​(String name,
                               SDVariable input,
                               Pooling2DConfig Pooling2DConfig)
        2D Convolution layer operation - average pooling 2d
        Parameters:
        name - name May be null. Name for the output variable
        input - the input to average pooling 2d operation - 4d CNN (image) activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        Pooling2DConfig - Configuration Object
        Returns:
        output Result after applying average pooling on the input (NUMERIC type)

      • avgPooling3d

        public SDVariable avgPooling3d​(SDVariable input,
                               Pooling3DConfig Pooling3DConfig)
        3D convolution layer operation - average pooling 3d
        Parameters:
        input - the input to average pooling 3d operation - 5d activations in NCDHW format (shape [minibatch, channels, depth, height, width]) or NDHWC format (shape [minibatch, depth, height, width, channels]) (NUMERIC type)
        Pooling3DConfig - Configuration Object
        Returns:
        output after applying average pooling on the input (NUMERIC type)

      • avgPooling3d

        public SDVariable avgPooling3d​(String name,
                               SDVariable input,
                               Pooling3DConfig Pooling3DConfig)
        3D convolution layer operation - average pooling 3d
        Parameters:
        name - name May be null. Name for the output variable
        input - the input to average pooling 3d operation - 5d activations in NCDHW format (shape [minibatch, channels, depth, height, width]) or NDHWC format (shape [minibatch, depth, height, width, channels]) (NUMERIC type)
        Pooling3DConfig - Configuration Object
        Returns:
        output after applying average pooling on the input (NUMERIC type)

      • batchToSpace

        public SDVariable batchToSpace​(SDVariable x,
                               int[] blocks,
                               int[] croppingTop,
                               int... croppingBottom)
        Convolution 2d layer batch to space operation on 4d input.
        Reduces input batch dimension by rearranging data into a larger spatial dimensions
        Parameters:
        x - Input variable. 4d input (NUMERIC type)
        blocks - Block size, in the height/width dimension (Size: Exactly(count=2))
        croppingTop - (Size: Exactly(count=2))
        croppingBottom - (Size: Exactly(count=2))
        Returns:
        output Output variable (NUMERIC type)

      • batchToSpace

        public SDVariable batchToSpace​(String name,
                               SDVariable x,
                               int[] blocks,
                               int[] croppingTop,
                               int... croppingBottom)
        Convolution 2d layer batch to space operation on 4d input.
        Reduces input batch dimension by rearranging data into a larger spatial dimensions
        Parameters:
        name - name May be null. Name for the output variable
        x - Input variable. 4d input (NUMERIC type)
        blocks - Block size, in the height/width dimension (Size: Exactly(count=2))
        croppingTop - (Size: Exactly(count=2))
        croppingBottom - (Size: Exactly(count=2))
        Returns:
        output Output variable (NUMERIC type)

      • col2Im

        public SDVariable col2Im​(SDVariable in,
                         Conv2DConfig Conv2DConfig)
        col2im operation for use in 2D convolution operations. Outputs a 4d array with shape
        [minibatch, inputChannels, height, width]
        Parameters:
        in - Input - rank 6 input with shape [minibatch, inputChannels, kernelHeight, kernelWidth, outputHeight, outputWidth] (NUMERIC type)
        Conv2DConfig - Configuration Object
        Returns:
        output Col2Im output variable (NUMERIC type)

      • col2Im

        public SDVariable col2Im​(String name,
                         SDVariable in,
                         Conv2DConfig Conv2DConfig)
        col2im operation for use in 2D convolution operations. Outputs a 4d array with shape
        [minibatch, inputChannels, height, width]
        Parameters:
        name - name May be null. Name for the output variable
        in - Input - rank 6 input with shape [minibatch, inputChannels, kernelHeight, kernelWidth, outputHeight, outputWidth] (NUMERIC type)
        Conv2DConfig - Configuration Object
        Returns:
        output Col2Im output variable (NUMERIC type)

      • conv1d

        public SDVariable conv1d​(SDVariable input,
                         SDVariable weights,
                         SDVariable bias,
                         Conv1DConfig Conv1DConfig)
        Conv1d operation.
        Parameters:
        input - the inputs to conv1d (NUMERIC type)
        weights - weights for conv1d op - rank 3 array with shape [kernelSize, inputChannels, outputChannels] (NUMERIC type)
        bias - bias for conv1d op - rank 1 array with shape [outputChannels]. May be null. (NUMERIC type)
        Conv1DConfig - Configuration Object
        Returns:
        output result of conv1d op (NUMERIC type)

      • conv1d

        public SDVariable conv1d​(String name,
                         SDVariable input,
                         SDVariable weights,
                         SDVariable bias,
                         Conv1DConfig Conv1DConfig)
        Conv1d operation.
        Parameters:
        name - name May be null. Name for the output variable
        input - the inputs to conv1d (NUMERIC type)
        weights - weights for conv1d op - rank 3 array with shape [kernelSize, inputChannels, outputChannels] (NUMERIC type)
        bias - bias for conv1d op - rank 1 array with shape [outputChannels]. May be null. (NUMERIC type)
        Conv1DConfig - Configuration Object
        Returns:
        output result of conv1d op (NUMERIC type)

      • conv1d

        public SDVariable conv1d​(SDVariable input,
                         SDVariable weights,
                         Conv1DConfig Conv1DConfig)
        Conv1d operation.
        Parameters:
        input - the inputs to conv1d (NUMERIC type)
        weights - weights for conv1d op - rank 3 array with shape [kernelSize, inputChannels, outputChannels] (NUMERIC type)
        Conv1DConfig - Configuration Object
        Returns:
        output result of conv1d op (NUMERIC type)

      • conv1d

        public SDVariable conv1d​(String name,
                         SDVariable input,
                         SDVariable weights,
                         Conv1DConfig Conv1DConfig)
        Conv1d operation.
        Parameters:
        name - name May be null. Name for the output variable
        input - the inputs to conv1d (NUMERIC type)
        weights - weights for conv1d op - rank 3 array with shape [kernelSize, inputChannels, outputChannels] (NUMERIC type)
        Conv1DConfig - Configuration Object
        Returns:
        output result of conv1d op (NUMERIC type)

      • conv2d

        public SDVariable conv2d​(SDVariable layerInput,
                         SDVariable weights,
                         SDVariable bias,
                         Conv2DConfig Conv2DConfig)
        2D Convolution operation with optional bias
        Parameters:
        layerInput - the input to max pooling 2d operation - 4d CNN (image) activations in NCHW format (NUMERIC type)
        weights - Weights for the convolution operation. 4 dimensions with format [kernelHeight, kernelWidth, inputChannels, outputChannels] (NUMERIC type)
        bias - Optional 1D bias array with shape [outputChannels]. May be null. (NUMERIC type)
        Conv2DConfig - Configuration Object
        Returns:
        output result of conv2d op (NUMERIC type)

      • conv2d

        public SDVariable conv2d​(String name,
                         SDVariable layerInput,
                         SDVariable weights,
                         SDVariable bias,
                         Conv2DConfig Conv2DConfig)
        2D Convolution operation with optional bias
        Parameters:
        name - name May be null. Name for the output variable
        layerInput - the input to max pooling 2d operation - 4d CNN (image) activations in NCHW format (NUMERIC type)
        weights - Weights for the convolution operation. 4 dimensions with format [kernelHeight, kernelWidth, inputChannels, outputChannels] (NUMERIC type)
        bias - Optional 1D bias array with shape [outputChannels]. May be null. (NUMERIC type)
        Conv2DConfig - Configuration Object
        Returns:
        output result of conv2d op (NUMERIC type)

      • conv2d

        public SDVariable conv2d​(SDVariable layerInput,
                         SDVariable weights,
                         Conv2DConfig Conv2DConfig)
        2D Convolution operation with optional bias
        Parameters:
        layerInput - the input to max pooling 2d operation - 4d CNN (image) activations in NCHW format (NUMERIC type)
        weights - Weights for the convolution operation. 4 dimensions with format [kernelHeight, kernelWidth, inputChannels, outputChannels] (NUMERIC type)
        Conv2DConfig - Configuration Object
        Returns:
        output result of conv2d op (NUMERIC type)

      • conv2d

        public SDVariable conv2d​(String name,
                         SDVariable layerInput,
                         SDVariable weights,
                         Conv2DConfig Conv2DConfig)
        2D Convolution operation with optional bias
        Parameters:
        name - name May be null. Name for the output variable
        layerInput - the input to max pooling 2d operation - 4d CNN (image) activations in NCHW format (NUMERIC type)
        weights - Weights for the convolution operation. 4 dimensions with format [kernelHeight, kernelWidth, inputChannels, outputChannels] (NUMERIC type)
        Conv2DConfig - Configuration Object
        Returns:
        output result of conv2d op (NUMERIC type)

      • conv3d

        public SDVariable conv3d​(SDVariable input,
                         SDVariable weights,
                         SDVariable bias,
                         Conv3DConfig Conv3DConfig)
        Convolution 3D operation with optional bias
        Parameters:
        input - the input to average pooling 3d operation - 5d activations in NCDHW format (shape [minibatch, channels, depth, height, width]) or NDHWC format (shape [minibatch, depth, height, width, channels]) (NUMERIC type)
        weights - Weights for conv3d. Rank 5 with shape [kernelDepth, kernelHeight, kernelWidth, inputChannels, outputChannels]. (NUMERIC type)
        bias - Optional 1D bias array with shape [outputChannels]. May be null. (NUMERIC type)
        Conv3DConfig - Configuration Object
        Returns:
        output Conv3d output variable (NUMERIC type)

      • conv3d

        public SDVariable conv3d​(String name,
                         SDVariable input,
                         SDVariable weights,
                         SDVariable bias,
                         Conv3DConfig Conv3DConfig)
        Convolution 3D operation with optional bias
        Parameters:
        name - name May be null. Name for the output variable
        input - the input to average pooling 3d operation - 5d activations in NCDHW format (shape [minibatch, channels, depth, height, width]) or NDHWC format (shape [minibatch, depth, height, width, channels]) (NUMERIC type)
        weights - Weights for conv3d. Rank 5 with shape [kernelDepth, kernelHeight, kernelWidth, inputChannels, outputChannels]. (NUMERIC type)
        bias - Optional 1D bias array with shape [outputChannels]. May be null. (NUMERIC type)
        Conv3DConfig - Configuration Object
        Returns:
        output Conv3d output variable (NUMERIC type)

      • conv3d

        public SDVariable conv3d​(SDVariable input,
                         SDVariable weights,
                         Conv3DConfig Conv3DConfig)
        Convolution 3D operation with optional bias
        Parameters:
        input - the input to average pooling 3d operation - 5d activations in NCDHW format (shape [minibatch, channels, depth, height, width]) or NDHWC format (shape [minibatch, depth, height, width, channels]) (NUMERIC type)
        weights - Weights for conv3d. Rank 5 with shape [kernelDepth, kernelHeight, kernelWidth, inputChannels, outputChannels]. (NUMERIC type)
        Conv3DConfig - Configuration Object
        Returns:
        output Conv3d output variable (NUMERIC type)

      • conv3d

        public SDVariable conv3d​(String name,
                         SDVariable input,
                         SDVariable weights,
                         Conv3DConfig Conv3DConfig)
        Convolution 3D operation with optional bias
        Parameters:
        name - name May be null. Name for the output variable
        input - the input to average pooling 3d operation - 5d activations in NCDHW format (shape [minibatch, channels, depth, height, width]) or NDHWC format (shape [minibatch, depth, height, width, channels]) (NUMERIC type)
        weights - Weights for conv3d. Rank 5 with shape [kernelDepth, kernelHeight, kernelWidth, inputChannels, outputChannels]. (NUMERIC type)
        Conv3DConfig - Configuration Object
        Returns:
        output Conv3d output variable (NUMERIC type)

      • deconv2d

        public SDVariable deconv2d​(SDVariable layerInput,
                           SDVariable weights,
                           SDVariable bias,
                           DeConv2DConfig DeConv2DConfig)
        2D deconvolution operation with optional bias
        Parameters:
        layerInput - the input to deconvolution 2d operation - 4d CNN (image) activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        weights - Weights for the 2d deconvolution operation. 4 dimensions with format [inputChannels, outputChannels, kernelHeight, kernelWidth] (NUMERIC type)
        bias - Optional 1D bias array with shape [outputChannels]. May be null. (NUMERIC type)
        DeConv2DConfig - Configuration Object
        Returns:
        output result of deconv2d op (NUMERIC type)

      • deconv2d

        public SDVariable deconv2d​(String name,
                           SDVariable layerInput,
                           SDVariable weights,
                           SDVariable bias,
                           DeConv2DConfig DeConv2DConfig)
        2D deconvolution operation with optional bias
        Parameters:
        name - name May be null. Name for the output variable
        layerInput - the input to deconvolution 2d operation - 4d CNN (image) activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        weights - Weights for the 2d deconvolution operation. 4 dimensions with format [inputChannels, outputChannels, kernelHeight, kernelWidth] (NUMERIC type)
        bias - Optional 1D bias array with shape [outputChannels]. May be null. (NUMERIC type)
        DeConv2DConfig - Configuration Object
        Returns:
        output result of deconv2d op (NUMERIC type)

      • deconv2d

        public SDVariable deconv2d​(SDVariable layerInput,
                           SDVariable weights,
                           DeConv2DConfig DeConv2DConfig)
        2D deconvolution operation with optional bias
        Parameters:
        layerInput - the input to deconvolution 2d operation - 4d CNN (image) activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        weights - Weights for the 2d deconvolution operation. 4 dimensions with format [inputChannels, outputChannels, kernelHeight, kernelWidth] (NUMERIC type)
        DeConv2DConfig - Configuration Object
        Returns:
        output result of deconv2d op (NUMERIC type)

      • deconv2d

        public SDVariable deconv2d​(String name,
                           SDVariable layerInput,
                           SDVariable weights,
                           DeConv2DConfig DeConv2DConfig)
        2D deconvolution operation with optional bias
        Parameters:
        name - name May be null. Name for the output variable
        layerInput - the input to deconvolution 2d operation - 4d CNN (image) activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        weights - Weights for the 2d deconvolution operation. 4 dimensions with format [inputChannels, outputChannels, kernelHeight, kernelWidth] (NUMERIC type)
        DeConv2DConfig - Configuration Object
        Returns:
        output result of deconv2d op (NUMERIC type)

      • deconv3d

        public SDVariable deconv3d​(SDVariable input,
                           SDVariable weights,
                           SDVariable bias,
                           DeConv3DConfig DeConv3DConfig)
        3D CNN deconvolution operation with or without optional bias
        Parameters:
        input - Input array - shape [bS, iD, iH, iW, iC] (NDHWC) or [bS, iC, iD, iH, iW] (NCDHW) (NUMERIC type)
        weights - Weights array - shape [kD, kH, kW, oC, iC] (NUMERIC type)
        bias - Bias array - optional, may be null. If non-null, must have shape [outputChannels] (NUMERIC type)
        DeConv3DConfig - Configuration Object
        Returns:
        output result of 3D CNN deconvolution operation (NUMERIC type)

      • deconv3d

        public SDVariable deconv3d​(String name,
                           SDVariable input,
                           SDVariable weights,
                           SDVariable bias,
                           DeConv3DConfig DeConv3DConfig)
        3D CNN deconvolution operation with or without optional bias
        Parameters:
        name - name May be null. Name for the output variable
        input - Input array - shape [bS, iD, iH, iW, iC] (NDHWC) or [bS, iC, iD, iH, iW] (NCDHW) (NUMERIC type)
        weights - Weights array - shape [kD, kH, kW, oC, iC] (NUMERIC type)
        bias - Bias array - optional, may be null. If non-null, must have shape [outputChannels] (NUMERIC type)
        DeConv3DConfig - Configuration Object
        Returns:
        output result of 3D CNN deconvolution operation (NUMERIC type)

      • deconv3d

        public SDVariable deconv3d​(SDVariable input,
                           SDVariable weights,
                           DeConv3DConfig DeConv3DConfig)
        3D CNN deconvolution operation with or without optional bias
        Parameters:
        input - Input array - shape [bS, iD, iH, iW, iC] (NDHWC) or [bS, iC, iD, iH, iW] (NCDHW) (NUMERIC type)
        weights - Weights array - shape [kD, kH, kW, oC, iC] (NUMERIC type)
        DeConv3DConfig - Configuration Object
        Returns:
        output result of 3D CNN deconvolution operation (NUMERIC type)

      • deconv3d

        public SDVariable deconv3d​(String name,
                           SDVariable input,
                           SDVariable weights,
                           DeConv3DConfig DeConv3DConfig)
        3D CNN deconvolution operation with or without optional bias
        Parameters:
        name - name May be null. Name for the output variable
        input - Input array - shape [bS, iD, iH, iW, iC] (NDHWC) or [bS, iC, iD, iH, iW] (NCDHW) (NUMERIC type)
        weights - Weights array - shape [kD, kH, kW, oC, iC] (NUMERIC type)
        DeConv3DConfig - Configuration Object
        Returns:
        output result of 3D CNN deconvolution operation (NUMERIC type)

      • depthToSpace

        public SDVariable depthToSpace​(SDVariable x,
                               int blockSize,
                               DataFormat dataFormat)
        Convolution 2d layer batch to space operation on 4d input.
        Reduces input channels dimension by rearranging data into a larger spatial dimensions
        Example: if input has shape [mb, 8, 2, 2] and block size is 2, then output size is [mb, 8/(2*2), 2*2, 2*2]
        = [mb, 2, 4, 4]
        Parameters:
        x - the input to depth to space pooling 2d operation - 4d activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        blockSize - Block size, in the height/width dimension
        dataFormat - Data format: "NCHW" or "NHWC"
        Returns:
        output Output variable (NUMERIC type)

      • depthToSpace

        public SDVariable depthToSpace​(String name,
                               SDVariable x,
                               int blockSize,
                               DataFormat dataFormat)
        Convolution 2d layer batch to space operation on 4d input.
        Reduces input channels dimension by rearranging data into a larger spatial dimensions
        Example: if input has shape [mb, 8, 2, 2] and block size is 2, then output size is [mb, 8/(2*2), 2*2, 2*2]
        = [mb, 2, 4, 4]
        Parameters:
        name - name May be null. Name for the output variable
        x - the input to depth to space pooling 2d operation - 4d activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        blockSize - Block size, in the height/width dimension
        dataFormat - Data format: "NCHW" or "NHWC"
        Returns:
        output Output variable (NUMERIC type)

      • depthWiseConv2d

        public SDVariable depthWiseConv2d​(SDVariable layerInput,
                                  SDVariable depthWeights,
                                  SDVariable bias,
                                  Conv2DConfig Conv2DConfig)
        Depth-wise 2D convolution operation with optional bias
        Parameters:
        layerInput - the input to max pooling 2d operation - 4d CNN (image) activations in NCHW format (NUMERIC type)
        depthWeights - Depth-wise conv2d weights. 4 dimensions with format [kernelHeight, kernelWidth, inputChannels, depthMultiplier] (NUMERIC type)
        bias - Optional 1D bias array with shape [outputChannels]. May be null. (NUMERIC type)
        Conv2DConfig - Configuration Object
        Returns:
        output result of depthwise conv2d op (NUMERIC type)

      • depthWiseConv2d

        public SDVariable depthWiseConv2d​(String name,
                                  SDVariable layerInput,
                                  SDVariable depthWeights,
                                  SDVariable bias,
                                  Conv2DConfig Conv2DConfig)
        Depth-wise 2D convolution operation with optional bias
        Parameters:
        name - name May be null. Name for the output variable
        layerInput - the input to max pooling 2d operation - 4d CNN (image) activations in NCHW format (NUMERIC type)
        depthWeights - Depth-wise conv2d weights. 4 dimensions with format [kernelHeight, kernelWidth, inputChannels, depthMultiplier] (NUMERIC type)
        bias - Optional 1D bias array with shape [outputChannels]. May be null. (NUMERIC type)
        Conv2DConfig - Configuration Object
        Returns:
        output result of depthwise conv2d op (NUMERIC type)

      • depthWiseConv2d

        public SDVariable depthWiseConv2d​(SDVariable layerInput,
                                  SDVariable depthWeights,
                                  Conv2DConfig Conv2DConfig)
        Depth-wise 2D convolution operation with optional bias
        Parameters:
        layerInput - the input to max pooling 2d operation - 4d CNN (image) activations in NCHW format (NUMERIC type)
        depthWeights - Depth-wise conv2d weights. 4 dimensions with format [kernelHeight, kernelWidth, inputChannels, depthMultiplier] (NUMERIC type)
        Conv2DConfig - Configuration Object
        Returns:
        output result of depthwise conv2d op (NUMERIC type)

      • depthWiseConv2d

        public SDVariable depthWiseConv2d​(String name,
                                  SDVariable layerInput,
                                  SDVariable depthWeights,
                                  Conv2DConfig Conv2DConfig)
        Depth-wise 2D convolution operation with optional bias
        Parameters:
        name - name May be null. Name for the output variable
        layerInput - the input to max pooling 2d operation - 4d CNN (image) activations in NCHW format (NUMERIC type)
        depthWeights - Depth-wise conv2d weights. 4 dimensions with format [kernelHeight, kernelWidth, inputChannels, depthMultiplier] (NUMERIC type)
        Conv2DConfig - Configuration Object
        Returns:
        output result of depthwise conv2d op (NUMERIC type)

      • dilation2D

        public SDVariable dilation2D​(SDVariable df,
                             SDVariable weights,
                             int[] strides,
                             int[] rates,
                             boolean isSameMode)
        TODO doc string
        Parameters:
        df - (NUMERIC type)
        weights - df (NUMERIC type)
        strides - weights (Size: Exactly(count=2))
        rates - strides (Size: Exactly(count=2))
        isSameMode - isSameMode
        Returns:
        output Computed the grayscale dilation of 4-D input and 3-D filters tensors. (NUMERIC type)

      • dilation2D

        public SDVariable dilation2D​(String name,
                             SDVariable df,
                             SDVariable weights,
                             int[] strides,
                             int[] rates,
                             boolean isSameMode)
        TODO doc string
        Parameters:
        name - name May be null. Name for the output variable
        df - (NUMERIC type)
        weights - df (NUMERIC type)
        strides - weights (Size: Exactly(count=2))
        rates - strides (Size: Exactly(count=2))
        isSameMode - isSameMode
        Returns:
        output Computed the grayscale dilation of 4-D input and 3-D filters tensors. (NUMERIC type)

      • extractImagePatches

        public SDVariable extractImagePatches​(SDVariable input,
                                      int kH,
                                      int kW,
                                      int sH,
                                      int sW,
                                      int rH,
                                      int rW,
                                      boolean sameMode)
        Extract image patches
        Parameters:
        input - Input array. Must be rank 4, with shape [minibatch, height, width, channels] (NUMERIC type)
        kH - Kernel height
        kW - Kernel width
        sH - Stride height
        sW - Stride width
        rH - Rate height
        rW - Rate width
        sameMode - If true: use same mode padding. If false
        Returns:
        output The result is a 4D tensor which is indexed by batch, row, and column. (NUMERIC type)

      • extractImagePatches

        public SDVariable extractImagePatches​(String name,
                                      SDVariable input,
                                      int kH,
                                      int kW,
                                      int sH,
                                      int sW,
                                      int rH,
                                      int rW,
                                      boolean sameMode)
        Extract image patches
        Parameters:
        name - name May be null. Name for the output variable
        input - Input array. Must be rank 4, with shape [minibatch, height, width, channels] (NUMERIC type)
        kH - Kernel height
        kW - Kernel width
        sH - Stride height
        sW - Stride width
        rH - Rate height
        rW - Rate width
        sameMode - If true: use same mode padding. If false
        Returns:
        output The result is a 4D tensor which is indexed by batch, row, and column. (NUMERIC type)

      • im2Col

        public SDVariable im2Col​(SDVariable in,
                         Conv2DConfig Conv2DConfig)
        im2col operation for use in 2D convolution operations. Outputs a 6d array with shape
        [minibatch, inputChannels, kernelHeight, kernelWidth, outputHeight, outputWidth]
        Parameters:
        in - Input - rank 4 input with shape [minibatch, inputChannels, height, width] (NUMERIC type)
        Conv2DConfig - Configuration Object
        Returns:
        output Im2Col output variable (NUMERIC type)

      • im2Col

        public SDVariable im2Col​(String name,
                         SDVariable in,
                         Conv2DConfig Conv2DConfig)
        im2col operation for use in 2D convolution operations. Outputs a 6d array with shape
        [minibatch, inputChannels, kernelHeight, kernelWidth, outputHeight, outputWidth]
        Parameters:
        name - name May be null. Name for the output variable
        in - Input - rank 4 input with shape [minibatch, inputChannels, height, width] (NUMERIC type)
        Conv2DConfig - Configuration Object
        Returns:
        output Im2Col output variable (NUMERIC type)

      • localResponseNormalization

        public SDVariable localResponseNormalization​(SDVariable input,
                                             LocalResponseNormalizationConfig LocalResponseNormalizationConfig)
        2D convolution layer operation - local response normalization
        Parameters:
        input - the inputs to lrn (NUMERIC type)
        LocalResponseNormalizationConfig - Configuration Object
        Returns:
        output Result after Local Response Normalization (NUMERIC type)

      • localResponseNormalization

        public SDVariable localResponseNormalization​(String name,
                                             SDVariable input,
                                             LocalResponseNormalizationConfig LocalResponseNormalizationConfig)
        2D convolution layer operation - local response normalization
        Parameters:
        name - name May be null. Name for the output variable
        input - the inputs to lrn (NUMERIC type)
        LocalResponseNormalizationConfig - Configuration Object
        Returns:
        output Result after Local Response Normalization (NUMERIC type)

      • maxPoolWithArgmax

        public SDVariable[] maxPoolWithArgmax​(SDVariable input,
                                      Pooling2DConfig Pooling2DConfig)
        2D Convolution layer operation - Max pooling on the input and outputs both max values and indices
        Parameters:
        input - the input to max pooling 2d operation - 4d CNN (image) activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        Pooling2DConfig - Configuration Object

      • maxPoolWithArgmax

        public SDVariable[] maxPoolWithArgmax​(String[] names,
                                      SDVariable input,
                                      Pooling2DConfig Pooling2DConfig)
        2D Convolution layer operation - Max pooling on the input and outputs both max values and indices
        Parameters:
        names - names May be null. Arrays of names for the output variables.
        input - the input to max pooling 2d operation - 4d CNN (image) activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        Pooling2DConfig - Configuration Object

      • maxPooling2d

        public SDVariable maxPooling2d​(SDVariable input,
                               Pooling2DConfig Pooling2DConfig)
        2D Convolution layer operation - max pooling 2d
        Parameters:
        input - the input to max pooling 2d operation - 4d CNN (image) activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        Pooling2DConfig - Configuration Object
        Returns:
        output Result after applying max pooling on the input (NUMERIC type)

      • maxPooling2d

        public SDVariable maxPooling2d​(String name,
                               SDVariable input,
                               Pooling2DConfig Pooling2DConfig)
        2D Convolution layer operation - max pooling 2d
        Parameters:
        name - name May be null. Name for the output variable
        input - the input to max pooling 2d operation - 4d CNN (image) activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        Pooling2DConfig - Configuration Object
        Returns:
        output Result after applying max pooling on the input (NUMERIC type)

      • maxPooling3d

        public SDVariable maxPooling3d​(SDVariable input,
                               Pooling3DConfig Pooling3DConfig)
        3D convolution layer operation - max pooling 3d operation.
        Parameters:
        input - the input to average pooling 3d operation - 5d activations in NCDHW format (shape [minibatch, channels, depth, height, width]) or NDHWC format (shape [minibatch, depth, height, width, channels]) (NUMERIC type)
        Pooling3DConfig - Configuration Object
        Returns:
        output Result after applying max pooling on the input (NUMERIC type)

      • maxPooling3d

        public SDVariable maxPooling3d​(String name,
                               SDVariable input,
                               Pooling3DConfig Pooling3DConfig)
        3D convolution layer operation - max pooling 3d operation.
        Parameters:
        name - name May be null. Name for the output variable
        input - the input to average pooling 3d operation - 5d activations in NCDHW format (shape [minibatch, channels, depth, height, width]) or NDHWC format (shape [minibatch, depth, height, width, channels]) (NUMERIC type)
        Pooling3DConfig - Configuration Object
        Returns:
        output Result after applying max pooling on the input (NUMERIC type)

      • separableConv2d

        public SDVariable separableConv2d​(SDVariable layerInput,
                                  SDVariable depthWeights,
                                  SDVariable pointWeights,
                                  SDVariable bias,
                                  Conv2DConfig Conv2DConfig)
        Separable 2D convolution operation with optional bias
        Parameters:
        layerInput - the input to max pooling 2d operation - 4d CNN (image) activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        depthWeights - Separable conv2d depth weights. 4 dimensions with format [kernelHeight, kernelWidth, inputChannels, depthMultiplier] (NUMERIC type)
        pointWeights - Point weights, rank 4 with format [1, 1, inputChannels*depthMultiplier, outputChannels]. May be null (NUMERIC type)
        bias - Optional bias, rank 1 with shape [outputChannels]. May be null. (NUMERIC type)
        Conv2DConfig - Configuration Object
        Returns:
        output result of separable convolution 2d operation (NUMERIC type)

      • separableConv2d

        public SDVariable separableConv2d​(String name,
                                  SDVariable layerInput,
                                  SDVariable depthWeights,
                                  SDVariable pointWeights,
                                  SDVariable bias,
                                  Conv2DConfig Conv2DConfig)
        Separable 2D convolution operation with optional bias
        Parameters:
        name - name May be null. Name for the output variable
        layerInput - the input to max pooling 2d operation - 4d CNN (image) activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        depthWeights - Separable conv2d depth weights. 4 dimensions with format [kernelHeight, kernelWidth, inputChannels, depthMultiplier] (NUMERIC type)
        pointWeights - Point weights, rank 4 with format [1, 1, inputChannels*depthMultiplier, outputChannels]. May be null (NUMERIC type)
        bias - Optional bias, rank 1 with shape [outputChannels]. May be null. (NUMERIC type)
        Conv2DConfig - Configuration Object
        Returns:
        output result of separable convolution 2d operation (NUMERIC type)

      • separableConv2d

        public SDVariable separableConv2d​(SDVariable layerInput,
                                  SDVariable depthWeights,
                                  SDVariable pointWeights,
                                  Conv2DConfig Conv2DConfig)
        Separable 2D convolution operation with optional bias
        Parameters:
        layerInput - the input to max pooling 2d operation - 4d CNN (image) activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        depthWeights - Separable conv2d depth weights. 4 dimensions with format [kernelHeight, kernelWidth, inputChannels, depthMultiplier] (NUMERIC type)
        pointWeights - Point weights, rank 4 with format [1, 1, inputChannels*depthMultiplier, outputChannels]. May be null (NUMERIC type)
        Conv2DConfig - Configuration Object
        Returns:
        output result of separable convolution 2d operation (NUMERIC type)

      • separableConv2d

        public SDVariable separableConv2d​(String name,
                                  SDVariable layerInput,
                                  SDVariable depthWeights,
                                  SDVariable pointWeights,
                                  Conv2DConfig Conv2DConfig)
        Separable 2D convolution operation with optional bias
        Parameters:
        name - name May be null. Name for the output variable
        layerInput - the input to max pooling 2d operation - 4d CNN (image) activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        depthWeights - Separable conv2d depth weights. 4 dimensions with format [kernelHeight, kernelWidth, inputChannels, depthMultiplier] (NUMERIC type)
        pointWeights - Point weights, rank 4 with format [1, 1, inputChannels*depthMultiplier, outputChannels]. May be null (NUMERIC type)
        Conv2DConfig - Configuration Object
        Returns:
        output result of separable convolution 2d operation (NUMERIC type)

      • spaceToBatch

        public SDVariable spaceToBatch​(SDVariable x,
                               int[] blocks,
                               int[] paddingTop,
                               int... paddingBottom)
        Convolution 2d layer space to batch operation on 4d input.
        Increases input batch dimension by rearranging data from spatial dimensions into batch dimension
        Parameters:
        x - Input variable. 4d input (NUMERIC type)
        blocks - Block size, in the height/width dimension (Size: Exactly(count=2))
        paddingTop - Optional 2d int[] array for padding the result: values [[pad top, pad bottom], [pad left, pad right]] (Size: Exactly(count=2))
        paddingBottom - Optional 2d int[] array for padding the result: values [[pad top, pad bottom], [pad left, pad right]] (Size: Exactly(count=2))
        Returns:
        output Output variable (NUMERIC type)

      • spaceToBatch

        public SDVariable spaceToBatch​(String name,
                               SDVariable x,
                               int[] blocks,
                               int[] paddingTop,
                               int... paddingBottom)
        Convolution 2d layer space to batch operation on 4d input.
        Increases input batch dimension by rearranging data from spatial dimensions into batch dimension
        Parameters:
        name - name May be null. Name for the output variable
        x - Input variable. 4d input (NUMERIC type)
        blocks - Block size, in the height/width dimension (Size: Exactly(count=2))
        paddingTop - Optional 2d int[] array for padding the result: values [[pad top, pad bottom], [pad left, pad right]] (Size: Exactly(count=2))
        paddingBottom - Optional 2d int[] array for padding the result: values [[pad top, pad bottom], [pad left, pad right]] (Size: Exactly(count=2))
        Returns:
        output Output variable (NUMERIC type)

      • spaceToDepth

        public SDVariable spaceToDepth​(SDVariable x,
                               int blockSize,
                               DataFormat dataFormat)
        Convolution 2d layer space to depth operation on 4d input.
        Increases input channels (reduced spatial dimensions) by rearranging data into a larger channels dimension
        Example: if input has shape [mb, 2, 4, 4] and block size is 2, then output size is [mb, 8/(2*2), 2*2, 2*2]
        = [mb, 2, 4, 4]
        Parameters:
        x - the input to depth to space pooling 2d operation - 4d activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        blockSize - Block size, in the height/width dimension
        dataFormat - Data format: "NCHW" or "NHWC"
        Returns:
        output Output variable (NUMERIC type)

      • spaceToDepth

        public SDVariable spaceToDepth​(String name,
                               SDVariable x,
                               int blockSize,
                               DataFormat dataFormat)
        Convolution 2d layer space to depth operation on 4d input.
        Increases input channels (reduced spatial dimensions) by rearranging data into a larger channels dimension
        Example: if input has shape [mb, 2, 4, 4] and block size is 2, then output size is [mb, 8/(2*2), 2*2, 2*2]
        = [mb, 2, 4, 4]
        Parameters:
        name - name May be null. Name for the output variable
        x - the input to depth to space pooling 2d operation - 4d activations in NCHW format (shape [minibatch, channels, height, width]) or NHWC format (shape [minibatch, height, width, channels]) (NUMERIC type)
        blockSize - Block size, in the height/width dimension
        dataFormat - Data format: "NCHW" or "NHWC"
        Returns:
        output Output variable (NUMERIC type)

      • upsampling2d

        public SDVariable upsampling2d​(SDVariable input,
                               int scale)
        Upsampling layer for 2D inputs.
        scale is used for both height and width dimensions.
        Parameters:
        input - Input in NCHW format (NUMERIC type)
        scale - The scale for both height and width dimensions.
        Returns:
        output Upsampled input (NUMERIC type)

      • upsampling2d

        public SDVariable upsampling2d​(String name,
                               SDVariable input,
                               int scale)
        Upsampling layer for 2D inputs.
        scale is used for both height and width dimensions.
        Parameters:
        name - name May be null. Name for the output variable
        input - Input in NCHW format (NUMERIC type)
        scale - The scale for both height and width dimensions.
        Returns:
        output Upsampled input (NUMERIC type)

      • upsampling2d

        public SDVariable upsampling2d​(SDVariable input,
                               int scaleH,
                               int scaleW,
                               boolean nchw)
        2D Convolution layer operation - Upsampling 2d
        Parameters:
        input - Input in NCHW format (NUMERIC type)
        scaleH - Scale to upsample in height dimension
        scaleW - Scale to upsample in width dimension
        nchw - If true: input is in NCHW (minibatch, channels, height, width) format. False: NHWC format
        Returns:
        output Upsampled input (NUMERIC type)

      • upsampling2d

        public SDVariable upsampling2d​(String name,
                               SDVariable input,
                               int scaleH,
                               int scaleW,
                               boolean nchw)
        2D Convolution layer operation - Upsampling 2d
        Parameters:
        name - name May be null. Name for the output variable
        input - Input in NCHW format (NUMERIC type)
        scaleH - Scale to upsample in height dimension
        scaleW - Scale to upsample in width dimension
        nchw - If true: input is in NCHW (minibatch, channels, height, width) format. False: NHWC format
        Returns:
        output Upsampled input (NUMERIC type)

      • upsampling3d

        public SDVariable upsampling3d​(SDVariable input,
                               boolean ncdhw,
                               int scaleD,
                               int scaleH,
                               int scaleW)
        3D Convolution layer operation - Upsampling 3d
        Parameters:
        input - Input in NCHW format (NUMERIC type)
        ncdhw - If true: input is in NCDHW (minibatch, channels, depth, height, width) format. False: NDHWC format
        scaleD - Scale to upsample in depth dimension
        scaleH - Scale to upsample in height dimension
        scaleW - Scale to upsample in width dimension
        Returns:
        output Upsampled input (NUMERIC type)

      • upsampling3d

        public SDVariable upsampling3d​(String name,
                               SDVariable input,
                               boolean ncdhw,
                               int scaleD,
                               int scaleH,
                               int scaleW)
        3D Convolution layer operation - Upsampling 3d
        Parameters:
        name - name May be null. Name for the output variable
        input - Input in NCHW format (NUMERIC type)
        ncdhw - If true: input is in NCDHW (minibatch, channels, depth, height, width) format. False: NDHWC format
        scaleD - Scale to upsample in depth dimension
        scaleH - Scale to upsample in height dimension
        scaleW - Scale to upsample in width dimension
        Returns:
        output Upsampled input (NUMERIC type)