introduce evaluate function, add pretty print

docs/_build/html/_sources/julia_examples.txt

@@ -72,15 +72,15 @@ We can now frame this problem in Julia code and solve our problem using LLS:
 
   slope = Variable();
   offset = Variable();
-  line = offset .+ x_data * slope;
+  line = offset + x_data * slope;
   residuals = line - y_data;
   error = sum_squares(residuals);
   optval = minimize!(error);
 
   # plot the data and the line
   t = [0; 5; 0.1];
   plt.plot(x_data, y_data, "ro");
-  plt.plot(t, slope.value .* t .+ offset.value);
+  plt.plot(t, evaluate(slope) * t + evaluate(offset));
   plt.xlabel("x");
   plt.ylabel("y");
   plt.show();
@@ -113,9 +113,9 @@ Here is the Julia code to solve this problem using LLS and plot the quadratic:
   quadratic_coeff = Variable();
   slope = Variable();
   offset = Variable();
-  quadratic = offset .+ x_data * slope + quadratic_coeff * x_data .^ 2;
+  quadratic = offset + x_data * slope + quadratic_coeff * x_data .^ 2;
   residuals = quadratic - y_data;
-  error = sum_squares(error);
+  error = sum_squares(residuals);
   optval = minimize!(error);
 
   # Create some evenly spaced points for plotting, again replicate powers
@@ -124,7 +124,7 @@ Here is the Julia code to solve this problem using LLS and plot the quadratic:
 
   # Plot our regressed function
   plt.plot(x_data, y_data, "ro")
-  plt.plot(t, offset.value .+  t .* slope.value .+ t_squared * quadratic.value, "b")
+  plt.plot(t, evaluate(offset) + t * evaluate(slope) + t_squared * evaluate(quadratic_coeff), "b")
   plt.xlabel("x")
   plt.ylabel("y")
   plt.show()
@@ -254,8 +254,8 @@ outcome:
   optval = minimize!(mu * sum_squares(velocity) + sum_squares(force), constraints);
 
 
-  plt.plot(position.value[1, 1:2:T]', position.value[2, 1:2:T]', "r-", linewidth=1.5)
-  plt.quiver(position.value[1, 1:4:T], position.value[2, 1:4:T], force.value[1, 1:4:T-1]/2, force.value[2, 1:4:T-1]/2, width=0.002)
+  plt.plot(evaluate(position)[1, 1:2:T]', evaluate(position)[2, 1:2:T]', "r-", linewidth=1.5)
+  plt.quiver(evaluate(position)[1, 1:4:T], evaluate(position)[2, 1:4:T], evaluate(force)[1, 1:4:T-1]/2, evaluate(force)[2, 1:4:T-1]/2, width=0.002)
   plt.plot(0, 0, "bo", markersize=10)
   plt.plot(final_position[1], final_position[2], "go", markersize=10)
   plt.xlim([-5, 15])
@@ -334,7 +334,7 @@ The code and data for this example can be found `here <https://github.com/davidl
   objective = sum_squares(line_mat * x - line_vals);
   optval = minimize!(objective);
 
-  plt.imshow(reshape(x.value, img_size,img_size), cmap = get_cmaps()[29])
+  plt.imshow(reshape(evaluate(x), img_size,img_size), cmap = get_cmaps()[29])
 
 The final result of the tomography will look something like
 
@@ -427,10 +427,10 @@ Here is the LLS code:
   # print out the 5 words most indicative of sports and nonsports
   words = String[];
   f = open("largeCorpusfeatures.txt");
-  for i = 1:length(w.value)
+  for i = 1:length(evaluate(w))
     push!(words, readline(f))
   end
-  indices = sortperm(vec(w.value));
+  indices = sortperm(vec(evaluate(w)));
   for i = 1:5
     print(words[indices[i]])
   end
@@ -506,7 +506,7 @@ The following code uses LLS to find and plot the model:
   smoothing = 100;
   smooth_objective = sum_squares(yearly[1 : n - 1] - yearly[2 : n]);
   optval = minimize!(sum_squares(temps - yearly) + smoothing * smooth_objective, eq_constraints);
-  residuals = temps - yearly.value;
+  residuals = temps - evaluate(yearly);
 
   # Plot smooth fit
   plt.figure(1)
@@ -561,7 +561,7 @@ against the actual residual temperatures:
   # plot autoregressive fit of daily fluctuations for first few days
   plt.figure(3)
   plt.plot(residuals[ar_len + 1 : ar_len + 100], color="g")
-  plt.plot(residuals_mat[1:100, :] * ar_coef.value, color="r")
+  plt.plot(residuals_mat[1:100, :] * evaluate(ar_coef), color="r")
   plt.title("Autoregressive Fit of Residuals")
 
 
@@ -576,8 +576,8 @@ Melbourne:
   # plot final fit of data
   plt.figure(4)
   plt.plot(temps)
-  total_estimate = yearly.value
-  total_estimate[ar_len + 1 : end] += residuals_mat * ar_coef.value
+  total_estimate = evaluate(yearly)
+  total_estimate[ar_len + 1 : end] += residuals_mat * evaluate(ar_coef)
   plt.plot(total_estimate, color="r", alpha=0.5)
   plt.title("Total Fit of Data")
   plt.xlim([0, n])

docs/_build/html/_sources/julia_tutorial.txt

@@ -15,7 +15,7 @@ To install LLS, simply open up the Julia shell and run the command:
 
 .. code-block:: none
 
-  Pkg.clone("https://github.com/davidlizeng/LinearLeastSquares.jl.git")
+  Pkg.add("LinearLeastSquares")
 
 To use LLS in Julia, run the following command to import the library:
 

docs/_build/html/julia_examples.html

@@ -106,15 +106,15 @@ <h3>Linear Regression<a class="headerlink" href="#linear-regression" title="Perm
 We can now frame this problem in Julia code and solve our problem using LLS:</p>
 <div class="highlight-none"><div class="highlight"><pre>slope = Variable();
 offset = Variable();
-line = offset .+ x_data * slope;
+line = offset + x_data * slope;
 residuals = line - y_data;
 error = sum_squares(residuals);
 optval = minimize!(error);
 
 # plot the data and the line
 t = [0; 5; 0.1];
 plt.plot(x_data, y_data, &quot;ro&quot;);
-plt.plot(t, slope.value .* t .+ offset.value);
+plt.plot(t, evaluate(slope) * t + evaluate(offset));
 plt.xlabel(&quot;x&quot;);
 plt.ylabel(&quot;y&quot;);
 plt.show();
@@ -142,9 +142,9 @@ <h3>Quadratic Regression<a class="headerlink" href="#quadratic-regression" title
 <div class="highlight-none"><div class="highlight"><pre>quadratic_coeff = Variable();
 slope = Variable();
 offset = Variable();
-quadratic = offset .+ x_data * slope + quadratic_coeff * x_data .^ 2;
+quadratic = offset + x_data * slope + quadratic_coeff * x_data .^ 2;
 residuals = quadratic - y_data;
-error = sum_squares(error);
+error = sum_squares(residuals);
 optval = minimize!(error);
 
 # Create some evenly spaced points for plotting, again replicate powers
@@ -153,7 +153,7 @@ <h3>Quadratic Regression<a class="headerlink" href="#quadratic-regression" title
 
 # Plot our regressed function
 plt.plot(x_data, y_data, &quot;ro&quot;)
-plt.plot(t, offset.value .+  t .* slope.value .+ t_squared * quadratic.value, &quot;b&quot;)
+plt.plot(t, evaluate(offset) + t * evaluate(slope) + t_squared * evaluate(quadratic_coeff), &quot;b&quot;)
 plt.xlabel(&quot;x&quot;)
 plt.ylabel(&quot;y&quot;)
 plt.show()
@@ -263,8 +263,8 @@ <h2>Control<a class="headerlink" href="#control" title="Permalink to this headli
 optval = minimize!(mu * sum_squares(velocity) + sum_squares(force), constraints);
 
 
-plt.plot(position.value[1, 1:2:T]&#39;, position.value[2, 1:2:T]&#39;, &quot;r-&quot;, linewidth=1.5)
-plt.quiver(position.value[1, 1:4:T], position.value[2, 1:4:T], force.value[1, 1:4:T-1]/2, force.value[2, 1:4:T-1]/2, width=0.002)
+plt.plot(evaluate(position)[1, 1:2:T]&#39;, evaluate(position)[2, 1:2:T]&#39;, &quot;r-&quot;, linewidth=1.5)
+plt.quiver(evaluate(position)[1, 1:4:T], evaluate(position)[2, 1:4:T], evaluate(force)[1, 1:4:T-1]/2, evaluate(force)[2, 1:4:T-1]/2, width=0.002)
 plt.plot(0, 0, &quot;bo&quot;, markersize=10)
 plt.plot(final_position[1], final_position[2], &quot;go&quot;, markersize=10)
 plt.xlim([-5, 15])
@@ -331,7 +331,7 @@ <h3>Tomography<a class="headerlink" href="#tomography" title="Permalink to this
 objective = sum_squares(line_mat * x - line_vals);
 optval = minimize!(objective);
 
-plt.imshow(reshape(x.value, img_size,img_size), cmap = get_cmaps()[29])
+plt.imshow(reshape(evaluate(x), img_size,img_size), cmap = get_cmaps()[29])
 </pre></div>
 </div>
 <p>The final result of the tomography will look something like</p>
@@ -412,10 +412,10 @@ <h3>Binary Classification<a class="headerlink" href="#binary-classification" tit
 # print out the 5 words most indicative of sports and nonsports
 words = String[];
 f = open(&quot;largeCorpusfeatures.txt&quot;);
-for i = 1:length(w.value)
+for i = 1:length(evaluate(w))
   push!(words, readline(f))
 end
-indices = sortperm(vec(w.value));
+indices = sortperm(vec(evaluate(w)));
 for i = 1:5
   print(words[indices[i]])
 end
@@ -480,7 +480,7 @@ <h2>Time Series Analysis<a class="headerlink" href="#time-series-analysis" title
 smoothing = 100;
 smooth_objective = sum_squares(yearly[1 : n - 1] - yearly[2 : n]);
 optval = minimize!(sum_squares(temps - yearly) + smoothing * smooth_objective, eq_constraints);
-residuals = temps - yearly.value;
+residuals = temps - evaluate(yearly);
 
 # Plot smooth fit
 plt.figure(1)
@@ -525,7 +525,7 @@ <h2>Time Series Analysis<a class="headerlink" href="#time-series-analysis" title
 # plot autoregressive fit of daily fluctuations for first few days
 plt.figure(3)
 plt.plot(residuals[ar_len + 1 : ar_len + 100], color=&quot;g&quot;)
-plt.plot(residuals_mat[1:100, :] * ar_coef.value, color=&quot;r&quot;)
+plt.plot(residuals_mat[1:100, :] * evaluate(ar_coef), color=&quot;r&quot;)
 plt.title(&quot;Autoregressive Fit of Residuals&quot;)
 </pre></div>
 </div>
@@ -536,8 +536,8 @@ <h2>Time Series Analysis<a class="headerlink" href="#time-series-analysis" title
 <div class="highlight-none"><div class="highlight"><pre># plot final fit of data
 plt.figure(4)
 plt.plot(temps)
-total_estimate = yearly.value
-total_estimate[ar_len + 1 : end] += residuals_mat * ar_coef.value
+total_estimate = evaluate(yearly)
+total_estimate[ar_len + 1 : end] += residuals_mat * evaluate(ar_coef)
 plt.plot(total_estimate, color=&quot;r&quot;, alpha=0.5)
 plt.title(&quot;Total Fit of Data&quot;)
 plt.xlim([0, n])

docs/_build/html/julia_tutorial.html

@@ -60,7 +60,7 @@ <h2>Installing LLS<a class="headerlink" href="#installing-lls" title="Permalink
 For those new to Julia, the <a class="reference external" href="http://docs.julialang.org/en/release-0.2/">official Julia docs</a>
 are a good way to get acquainted with the language.</p>
 <p>To install LLS, simply open up the Julia shell and run the command:</p>
-<div class="highlight-none"><div class="highlight"><pre>Pkg.clone(&quot;https://github.com/davidlizeng/LinearLeastSquares.jl.git&quot;)
+<div class="highlight-none"><div class="highlight"><pre>Pkg.add(&quot;LinearLeastSquares&quot;)
 </pre></div>
 </div>
 <p>To use LLS in Julia, run the following command to import the library:</p>

docs/julia_examples.rst

@@ -72,15 +72,15 @@ We can now frame this problem in Julia code and solve our problem using LLS:
 
   slope = Variable();
   offset = Variable();
-  line = offset .+ x_data * slope;
+  line = offset + x_data * slope;
   residuals = line - y_data;
   error = sum_squares(residuals);
   optval = minimize!(error);
 
   # plot the data and the line
   t = [0; 5; 0.1];
   plt.plot(x_data, y_data, "ro");
-  plt.plot(t, slope.value .* t .+ offset.value);
+  plt.plot(t, evaluate(slope) * t + evaluate(offset));
   plt.xlabel("x");
   plt.ylabel("y");
   plt.show();
@@ -113,9 +113,9 @@ Here is the Julia code to solve this problem using LLS and plot the quadratic:
   quadratic_coeff = Variable();
   slope = Variable();
   offset = Variable();
-  quadratic = offset .+ x_data * slope + quadratic_coeff * x_data .^ 2;
+  quadratic = offset + x_data * slope + quadratic_coeff * x_data .^ 2;
   residuals = quadratic - y_data;
-  error = sum_squares(error);
+  error = sum_squares(residuals);
   optval = minimize!(error);
 
   # Create some evenly spaced points for plotting, again replicate powers
@@ -124,7 +124,7 @@ Here is the Julia code to solve this problem using LLS and plot the quadratic:
 
   # Plot our regressed function
   plt.plot(x_data, y_data, "ro")
-  plt.plot(t, offset.value .+  t .* slope.value .+ t_squared * quadratic.value, "b")
+  plt.plot(t, evaluate(offset) + t * evaluate(slope) + t_squared * evaluate(quadratic_coeff), "b")
   plt.xlabel("x")
   plt.ylabel("y")
   plt.show()
@@ -254,8 +254,8 @@ outcome:
   optval = minimize!(mu * sum_squares(velocity) + sum_squares(force), constraints);
 
 
-  plt.plot(position.value[1, 1:2:T]', position.value[2, 1:2:T]', "r-", linewidth=1.5)
-  plt.quiver(position.value[1, 1:4:T], position.value[2, 1:4:T], force.value[1, 1:4:T-1]/2, force.value[2, 1:4:T-1]/2, width=0.002)
+  plt.plot(evaluate(position)[1, 1:2:T]', evaluate(position)[2, 1:2:T]', "r-", linewidth=1.5)
+  plt.quiver(evaluate(position)[1, 1:4:T], evaluate(position)[2, 1:4:T], evaluate(force)[1, 1:4:T-1]/2, evaluate(force)[2, 1:4:T-1]/2, width=0.002)
   plt.plot(0, 0, "bo", markersize=10)
   plt.plot(final_position[1], final_position[2], "go", markersize=10)
   plt.xlim([-5, 15])
@@ -334,7 +334,7 @@ The code and data for this example can be found `here <https://github.com/davidl
   objective = sum_squares(line_mat * x - line_vals);
   optval = minimize!(objective);
 
-  plt.imshow(reshape(x.value, img_size,img_size), cmap = get_cmaps()[29])
+  plt.imshow(reshape(evaluate(x), img_size,img_size), cmap = get_cmaps()[29])
 
 The final result of the tomography will look something like
 
@@ -427,10 +427,10 @@ Here is the LLS code:
   # print out the 5 words most indicative of sports and nonsports
   words = String[];
   f = open("largeCorpusfeatures.txt");
-  for i = 1:length(w.value)
+  for i = 1:length(evaluate(w))
     push!(words, readline(f))
   end
-  indices = sortperm(vec(w.value));
+  indices = sortperm(vec(evaluate(w)));
   for i = 1:5
     print(words[indices[i]])
   end
@@ -506,7 +506,7 @@ The following code uses LLS to find and plot the model:
   smoothing = 100;
   smooth_objective = sum_squares(yearly[1 : n - 1] - yearly[2 : n]);
   optval = minimize!(sum_squares(temps - yearly) + smoothing * smooth_objective, eq_constraints);
-  residuals = temps - yearly.value;
+  residuals = temps - evaluate(yearly);
 
   # Plot smooth fit
   plt.figure(1)
@@ -561,7 +561,7 @@ against the actual residual temperatures:
   # plot autoregressive fit of daily fluctuations for first few days
   plt.figure(3)
   plt.plot(residuals[ar_len + 1 : ar_len + 100], color="g")
-  plt.plot(residuals_mat[1:100, :] * ar_coef.value, color="r")
+  plt.plot(residuals_mat[1:100, :] * evaluate(ar_coef), color="r")
   plt.title("Autoregressive Fit of Residuals")
 
 
@@ -576,8 +576,8 @@ Melbourne:
   # plot final fit of data
   plt.figure(4)
   plt.plot(temps)
-  total_estimate = yearly.value
-  total_estimate[ar_len + 1 : end] += residuals_mat * ar_coef.value
+  total_estimate = evaluate(yearly)
+  total_estimate[ar_len + 1 : end] += residuals_mat * evaluate(ar_coef)
   plt.plot(total_estimate, color="r", alpha=0.5)
   plt.title("Total Fit of Data")
   plt.xlim([0, n])

docs/julia_tutorial.rst

@@ -15,7 +15,7 @@ To install LLS, simply open up the Julia shell and run the command:
 
 .. code-block:: none
 
-  Pkg.clone("https://github.com/davidlizeng/LinearLeastSquares.jl.git")
+  Pkg.add("LinearLeastSquares")
 
 To use LLS in Julia, run the following command to import the library:
 

examples/control/control.jl

@@ -49,8 +49,8 @@ constraints += velocity[:, T] == 0;
 optval = minimize!(mu * sum_squares(velocity) + sum_squares(force), constraints);
 
 
-plt.plot(position.value[1, 1:2:T]', position.value[2, 1:2:T]', "r-", linewidth=1.5)
-plt.quiver(position.value[1, 1:4:T], position.value[2, 1:4:T], force.value[1, 1:4:T-1]/2, force.value[2, 1:4:T-1]/2, width=0.002)
+plt.plot(evaluate(position)[1, 1:2:T]', evaluate(position)[2, 1:2:T]', "r-", linewidth=1.5)
+plt.quiver(evaluate(position)[1, 1:4:T], evaluate(position)[2, 1:4:T], evaluate(force)[1, 1:4:T-1]/2, evaluate(force)[2, 1:4:T-1]/2, width=0.002)
 plt.plot(0, 0, "bo", markersize=10)
 plt.plot(final_position[1], final_position[2], "go", markersize=10)
 plt.xlim([-5, 15])

examples/machine_learning/document_classification.jl

@@ -39,10 +39,10 @@ optval = minimize!(objective);
 # print out the 5 words most indicative of sports and nonsports
 words = String[];
 f = open("largeCorpusfeatures.txt");
-for i = 1:length(w.value)
+for i = 1:length(evaluate(w))
   push!(words, readline(f))
 end
-indices = sortperm(vec(full(w.value)));
+indices = sortperm(vec(evaluate(w)));
 for i = 1:5
   print(words[indices[i]])
 end
@@ -51,9 +51,9 @@ for i = 0:4
 end
 
 # calculate training error
-yhat = sign(trainDocuments * w.value .+ v.value);
+yhat = sign(trainDocuments * evaluate(w) + evaluate(v));
 trainCE =  1/size(trainClasses,1)*sum(trainClasses .!= yhat)
 
 # calculate performance of our classifier on the test set
-yhat2 = sign(testDocuments * w.value .+ v.value);
+yhat2 = sign(testDocuments * evaluate(w) + evaluate(v));
 testCE = 1/size(testClasses,1)*sum(testClasses .!= yhat2)

examples/simple_lin_and_quad_reg/linear_regression.jl

@@ -3,6 +3,7 @@
 # Translated into LinearLeastSquares.jl by Karanveer Mohan and David Zeng
 
 using LinearLeastSquares
+import PyPlot.plt
 # Set the random seed to get consistent data
 srand(1)
 
@@ -19,14 +20,15 @@ y_data = x_data_expanded * true_coeffs + 0.5 * rand(n, 1);
 
 slope = Variable();
 offset = Variable();
-optval = minimize!(sum_squares(offset + x_data * slope - y_data));
-println("Slope = $(slope.value[1, 1]), offset = $(offset.value[1, 1])");
+line = offset + x_data * slope;
+residuals = line - y_data;
+error = sum_squares(residuals);
+optval = minimize!(error);
 
-import PyPlot.plt
-# Print results and plot
+# plot the data and the line
 t = [0; 5; 0.1];
 plt.plot(x_data, y_data, "ro");
-plt.plot(t, slope.value[1, 1] * t .+ offset.value[1, 1]);
+plt.plot(t, evaluate(slope) .* t .+ evaluate(offset));
 plt.xlabel("x");
 plt.ylabel("y");
 plt.show();