Version 0.9.2. Fixes to stl code tags

stl/feedfowardRICA.m

@@ -43,9 +43,7 @@
     filter = zeros(8,8); % You should replace this
     % Form W, obtain the feature (filterDim x filterDim) needed during the
     % convolution
-    % ---- YOUR CODE HERE ----
-    filter = squeeze(W(:,:,filterNum));
-    % ------------------------
+    %%% YOUR CODE HERE %%%
 
     % Flip the feature matrix because of the definition of convolution, as explained later
     filter = rot90(squeeze(filter),2);
@@ -56,16 +54,11 @@
     resp = zeros(convDim, convDim); % You should replace this
     % Convolve "filter" with "im" to find "resp"
     % be sure to do a 'valid' convolution
-    % ---- YOUR CODE HERE ----
-    resp = conv2(im,filter,'valid');
-    % ------------------------      
+    %%% YOUR CODE HERE %%%
     % Then, apply square-square-root pooling on "resp" to get the hidden
     % activation "act"
     act = zeros(convDim / poolDim, convDim / poolDim); % You should replace this
-    % ---- YOUR CODE HERE ---%
-    act=conv2(resp.^2,poolMat,'valid');
-    act = sqrt(act(1:poolDim:end,1:poolDim:end)+params.epsilon);
-    % ------------------------
+    %%% YOUR CODE HERE %%%
     features(:, :, filterNum, imageNum) = act;
   end
 end

stl/stlExercise.m

@@ -85,11 +85,8 @@
 %  Find opttheta by running the RICA on all the training patches.
 %  You will need to whitened the patches with the zca2 function 
 %  then call minFunc with the softICACost function as seen in the RICA exercise.
-%% ----------------- YOUR CODE HERE ----------------------
-[patches,V,E,D] = zca2(patches);
-% optimize
-[opttheta, cost, exitflag] = minFunc( @(theta) softICACost(theta, patches, params), randTheta, options); 
-%% -------------------------------------------------------
+%%% YOUR CODE HERE %%%
+
 % reshape visualize weights
 W = reshape(opttheta, params.numFeatures, params.n);
 display_network(W');
@@ -136,24 +133,14 @@
 options.MaxIter = 300;
 
 % optimize
-%% ----------------- YOUR CODE HERE ----------------------
-[opttheta_softmax, cost, exitflag] = minFunc( @(theta) softmax_regression_vec(theta, trainFeatures, trainLabels), randTheta2, options);
-
-%% -----------------------------------------------------
+%%% YOUR CODE HERE %%%
 
 
 %%======================================================================
 %% STEP 5: Testing 
 % Compute Predictions on tran and test sets using softmaxPredict
 % and softmaxModel
-%% ----------------- YOUR CODE HERE ----------------------
-W_softmax = reshape(opttheta_softmax, featureSize, numClasses);
-train_score = trainFeatures'*W_softmax;
-[~,train_pred] = max(train_score, [], 2); 
-score = testFeatures'*W_softmax;
-[~,pred] = max(score, [], 2); 
-%% -----------------------------------------------------
-
+%%% YOUR CODE HERE %%%
 % Classification Score
 fprintf('Train Accuracy: %f%%\n', 100*mean(train_pred(:) == trainLabels(:)));
 fprintf('Test Accuracy: %f%%\n', 100*mean(pred(:) == testLabels(:)));