scripts/nn/layers/conv1d.dml

#-------------------------------------------------------------
#
# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements.  See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership.  The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License.  You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied.  See the License for the
# specific language governing permissions and limitations
# under the License.
#
#-------------------------------------------------------------

source("scripts/nn/util.dml") as util

forward = function(matrix[double] input, matrix[double] filter, int pad, int stride,
  int numInput, int numChannels, int inputWidth, int numFilters, int filterSize)
  return (matrix[double] out)
{
  /*
  * Computes the forward pass for a 1D spatial convolutional layer
  * by reshaping the input to fit conv2d.
  *
  * Inputs:
  *  - input: Inputs, of shape (N, C*W).
  *  - filter: Weights, of shape (F, C*W).
  *  - pad: Padding for left and right sides of input elements.
  *  - stride: Stride for moving filter.
  *  - numInput: Number of input elements N
  *  - numChannels: Number of input channels (dimensionality of input depth).
  *  - inputWidth: Input width.
  *  - numFilters: Number of filters F
  *  - filterSize: Filter width.
  *
  * Outputs:
  *  - out: Outputs, of shape (N, F*Wout).
  */
  out = conv2d(input, filter, padding=[0,pad], stride=[1, stride], 
	  input_shape=[numInput,numChannels,1,inputWidth], filter_shape=[numFilters,numChannels,1,filterSize])
}

backward_data = function(matrix[double] filter, matrix[double] dout, int pad, int stride,
  int numInput, int numChannels, int inputWidth, int numFilters, int filterSize)
  return (matrix[double] out)
{
  /*
  * Computes the backward pass regarding the input data for a 1D spatial convolutional layer
  * by reshaping the input to fit conv2d backward data pass.
  *
  * Inputs:
  *  - filter: Weights, of shape (F, C*W).
  *  - dout: Output of the forward pass
  *  - pad: Padding for left and right sides of input elements.
  *  - stride: Stride for moving filter.
  *  - numInput: Number of input elements N
  *  - numChannels: Number of input channels (dimensionality of input depth).
  *  - inputWidth: Input width.
  *  - numFilters: Number of filters F
  *  - filterSize: Filter width.
  *
  * Outputs:
  *  - out: gradients based on the input data of the convolution.
  */
  out = conv2d_backward_data(filter, dout, stride=[1,stride], padding=[0,pad], 
	  input_shape=[numInput,numChannels,1,inputWidth], filter_shape=[numFilters,numChannels,1,filterSize])
}

backward_filter = function(matrix[double] input, matrix[double] dout, int pad, int stride, 
  int numInput, int numChannels, int inputWidth, int numFilters, int filterSize)
  return (matrix[double] out)
{
  /*
  * Computes the backward pass regarding the filter for a 1D spatial convolutional layer
  * by reshaping the input to fit conv2d backward data pass.
  *
  * Inputs:
  *  - input: Inputs, of shape (N, C*W).
  *  - dout: Output of the forward pass
  *  - pad: Padding for left and right sides of input elements.
  *  - stride: Stride for moving filter.
  *  - numInput: Number of input elements N
  *  - numChannels: Number of input channels (dimensionality of input depth).
  *  - inputWidth: Input width.
  *  - numFilters: Number of filters F
  *  - filterSize: Filter width.
  *
  * Outputs:
  *  - out: gradients bsaed on the filter of the convolution.
  */
  out = conv2d_backward_filter(input, dout, stride=[1,stride], padding=[0,pad], 
	  input_shape=[numInput,numChannels,1,inputWidth], filter_shape=[numFilters,numChannels,1,filterSize])
}