UFLDL教程Exercise答案(8)：Convolution and Pooling

教程地址：http://deeplearning.stanford.edu/wiki/index.php/UFLDL_Tutorial

Exercise地址：http://deeplearning.stanford.edu/wiki/index.php/Exercise:Convolution_and_Pooling

代码

Step 1: Load learned features ——cnnExercise.m

% --------------------------- YOUR CODE HERE --------------------------
% Train the sparse autoencoder and fill the following variables with 
% the optimal parameters:

optTheta =  zeros(2*hiddenSize*visibleSize+hiddenSize+visibleSize, 1);
ZCAWhite =  zeros(visibleSize, visibleSize);
meanPatch = zeros(visibleSize, 1);

load STL10Features.mat;

% --------------------------------------------------------------------

Step 2: Implement and test convolution and pooling 

Step 2a: Implement convolution——cnnConvolve.m

function convolvedFeatures = cnnConvolve(patchDim, numFeatures, images, W, b, ZCAWhite, meanPatch)
%cnnConvolve Returns the convolution of the features given by W and b with
%the given images
%
% Parameters:
%  patchDim - patch (feature) dimension
%  numFeatures - number of features特征数 hiddenSize隐藏层单元数
%  images - large images to convolve with, matrix in the form
%           images(r, c, channel, image number)  需要被卷积的大尺寸图像
%  W, b - W, b for features from the sparse autoencoder
%  ZCAWhite, meanPatch - ZCAWhitening and meanPatch matrices used for
%                        preprocessing
%
% Returns:
%  convolvedFeatures - matrix of convolved features in the form
%                      convolvedFeatures(featureNum, imageNum, imageRow, imageCol)

numImages = size(images, 4);  %图像数量
imageDim = size(images, 1);   %图像patch的row
imageChannels = size(images, 3);  %图像通道数

convolvedFeatures = zeros(numFeatures, numImages, imageDim - patchDim + 1, imageDim - patchDim + 1);

% Instructions:
%   Convolve every feature with every large image here to produce the 
%   numFeatures x numImages x (imageDim - patchDim + 1) x (imageDim - patchDim + 1) 
%   matrix convolvedFeatures, such that 
%   convolvedFeatures(featureNum, imageNum, imageRow, imageCol) is the
%   value of the convolved featureNum feature for the imageNum image over
%   the region (imageRow, imageCol) to (imageRow + patchDim - 1, imageCol + patchDim - 1)
%
% Expected running times: 
%   Convolving with 100 images should take less than 3 minutes 
%   Convolving with 5000 images should take around an hour
%   (So to save time when testing, you should convolve with less images, as
%   described earlier)

% -------------------- YOUR CODE HERE --------------------
% Precompute the matrices that will be used during the convolution. Recall
% that you need to take into account the whitening and mean subtraction
% steps

WT = W * ZCAWhite;  %ZCA白化并减去均值
bT = b - WT * meanPatch;  

% --------------------------------------------------------

convolvedFeatures = zeros(numFeatures, numImages, imageDim - patchDim + 1, imageDim - patchDim + 1);
for imageNum = 1:numImages
  for featureNum = 1:numFeatures

    % convolution of image with feature matrix for each channel
    convolvedImage = zeros(imageDim - patchDim + 1, imageDim - patchDim + 1);
    for channel = 1:3

      % Obtain the feature (patchDim x patchDim) needed during the convolution
      % ---- YOUR CODE HERE ----
      feature = zeros(8,8); % You should replace this
      feature = reshape(WT(featureNum,patchDim*patchDim*(channel-1)+1:patchDim*patchDim*channel),[patchDim,patchDim]);
      %WT（featureNum(hiddenSize)行，(channel*patchDim*patchDim)列）
      %取WT的对应第featureNum幅图像的第channel通道的图像信息，patchDim*patchDim个元素
            
      % ------------------------

      % Flip the feature matrix because of the definition of convolution, as explained later
      feature = flipud(fliplr(squeeze(feature)));   %翻转feature
                                                    %squeeze函数用于删除矩阵中的单一维（Remove singleton dimensions）
      
      % Obtain the image
      im = squeeze(images(:, :, channel, imageNum));  %大尺寸图像，一幅，单通道的图像信息，r*c矩阵

      % Convolve "feature" with "im", adding the result to convolvedImage
      % be sure to do a 'valid' convolution
      % ---- YOUR CODE HERE ----
      convolvedSingleChannel = conv2( im, feature, 'valid');  %对一幅图像，单通道进行卷积操作
      convolvedImage = convolvedImage + convolvedSingleChannel;  %将三通道的卷积值相加，作为最后的卷积结果
      
      % ------------------------

    end
    
    % Subtract the bias unit (correcting for the mean subtraction as well)
    % Then, apply the sigmoid function to get the hidden activation
    % ---- YOUR CODE HERE ----
    convolvedImage = sigmoid(convolvedImage + bT(featureNum));  %加上偏置项bias,用sigmoid函数计算激活值

    % ------------------------
    
    % The convolved feature is the sum of the convolved values for all channels
    convolvedFeatures(featureNum, imageNum, :, :) = convolvedImage;
  end
end

end

function sigm = sigmoid(x)  
    sigm = 1./(1+exp(-x)); 
end

Step 2b: Check your convolution——代码已给

Step 2c: Pooling——cnnPool.m

% -------------------- YOUR CODE HERE --------------------
resultDim = floor(convolvedDim / poolDim);  %池化后图像的维数
for imageNum = 1:numImages
  for featureNum = 1:numFeatures
      for poolRow = 1 : resultDim   %池化后图像的row
          for poolCol = 1 : resultDim  %%池化后图像的col
              patchRow = (poolRow - 1) * poolDim + 1;  %池化图像块的row起点
              patchCol = (poolCol - 1) * poolDim + 1;  %池化图像块的col起点
              poolPatch = convolvedFeatures(featureNum, imageNum, patchRow:(patchRow + poolDim - 1), patchCol:(patchCol + poolDim - 1));
              pooledFeatures(featureNum, imageNum, poolRow, poolCol) = mean(poolPatch(:));
          end
      end
  end
end

%---------------------------------------------------------------  



Step 2d: Check your pooling——代码已给
Step 3: Convolve and pool with the dataset——代码已给
Step 4: Use pooled features for classification——代码已给

Step 5: Test classifier——代码已给