public class LibMatrixDNNConv2dHelper extends Object
Modifier and Type | Class and Description |
---|---|
static class |
LibMatrixDNNConv2dHelper.LoopedIm2ColConv2dAllChan
Performs convolution via: partialCopy1(filter %*% im2col(input)) = output
|
static class |
LibMatrixDNNConv2dHelper.LoopedIm2ColConv2dOneChan
Performs convolution via: partialCopy1(filter %*% im2col(input)) = output.
|
static class |
LibMatrixDNNConv2dHelper.LoopedIm2ColConv2dTransAllChan
This implementation is similar to LoopedIm2ColConv2dAllChan, except for using a
sparse-dense matrix multiplication with t(t(Xi) %*% t(F)) instead of a
dense-sparse matrix multiplication with Xi %*% F.
|
static class |
LibMatrixDNNConv2dHelper.SparseNativeConv2d
This operator is used only if native is enabled, filter is dense and input is sparse
|
Constructor and Description |
---|
LibMatrixDNNConv2dHelper() |
Copyright © 2017 The Apache Software Foundation. All rights reserved.