Merge branch 'master' of ssh://vo20.nl/home/git/repos/uva

modsim/ass3/Makefile

-CFLAGS=-Wall -Wextra -pedantic -std=c99 -D_GNU_SOURCE -g -O0
+CFLAGS=-Wall -Wextra -pedantic -std=c99 -D_GNU_SOURCE -g -ggdb -O0
 LDFLAGS=-lm
 
 PROGS=test main

statred/ass1/Makefile

+.PHONY: all clean
+
+all:
+
+clean:
+	rm -vf *.pyc q*.pdf

statred/ass1/q21_multivariate.py

 from pylab import array, eig, diagflat, dot, sqrt, randn, tile, \
         plot, subplot, axis, figure, clf, savefig
 
+# The used mu (mean vector) and cov (covariance matrix).
 mu = array([[3],
-              [4],
-              [5],
-              [6]])
+            [4],
+            [5],
+            [6]])
 
 cov  = array(
     [[ 3.01602775,  1.02746769, -3.60224613, -2.08792829],
@@ -12,10 +13,14 @@ cov  = array(
      [-3.60224613, -3.98616664, 13.04508284, -1.59255406],
      [-2.08792829,  0.48723704, -1.59255406,  8.28742469]])
 
+# Samples is the constant `N' which is the total amount of numbers to generate
+# according to the normal distribution.
 samples = 1000
 vector_size = 4
 
 def dataset():
+    # The covariance matrix is used to transform the generated dataset into a
+    # multivariant normal distribution dataset.
     d, U = eig(cov)
     L = diagflat(d)
     A = dot(U, sqrt(L))
@@ -23,11 +28,13 @@ def dataset():
     return dot(A,X) + tile(mu, samples)
 
 if __name__ == '__main__':
+    # Create a n*n grid of subplots and generate a new dataset.
     figure(vector_size**2)
     clf()
     Y = dataset()
     for i in range(vector_size):
         for j in range(vector_size):
+            # Skip the diagonal subplots since those are irrelevant.
             if i != j:
                 subplot(vector_size, vector_size, (i+1) + j*vector_size)
                 plot(Y[i], Y[j], 'x')

statred/ass1/q22_estimate.py

 from q21_multivariate import dataset
-from numpy import array, mean, tile, newaxis, dot
-from pylab import eigvals, diagflat, axis, figure, clf, show, plot, subplot
+from pylab import array, mean, tile, newaxis, dot, eigvals, \
+        axis, figure, clf, show, plot
 
 def eigenvalues(n):
+    """Return eigenvalues of unbiased estimators for the covariance matrix
+    Sigma (based on a pseudo-random generated dataset)."""
+
     Y = array([mean(dataset(), 1) for i in range(n)]).T
+
+    # Sigma = 1 / (n - 1) * Sum for i=1 to n: (x_i - x_mean) T(x_i - x_mean),
+    # where T(x) is the transpose of `x'. Mu = x_mean and
+    # Yzm = Sum for i=1 to n: x_i - x_mean.
     mu = mean(Y, 1)
     Yzm = Y - tile(mu[:,newaxis], n)
     S = dot(Yzm, Yzm.T) / (n - 1)
@@ -11,14 +18,17 @@ def eigenvalues(n):
 
 figure(1)
 clf()
-samples = range(2, 10000, 500)
+
+max_range = 10000
+
+samples = range(2, max_range, 500)
 data = [[] for i in range(4)]
+
 for n in samples:
     e = eigenvalues(n)
     for i in range(4):
         data[i].append(e[i])
 for i in range(4):
-    #subplot(2, 2, i+1)
     plot(samples, data[i], 'x')
-axis([0, 10000, 0., 0.025])
+axis([0, max_range, 0., 0.025])
 show()

statred/ass1/q23_iris.py

-from numpy import loadtxt
-from pylab import figure, plot, subplot, show, axis, clf
+from pylab import loadtxt, figure, plot, subplot, axis, clf, savefig
 
-def cnvt(s):
-    try:
-        return {'Iris-setosa': 0.0, 'Iris-versicolor': 1.0, \
-                'Iris-virginica': 2.0}[s]
-    except KeyError:
-        ireturn -1.0
+# The last column of the data sets is a label, which is used to distinguish the
+# three groups of data in the data sets. This label should be translated to a
+# floating point, or a conversion error will occur (since ``dtype=float'').
+cnvt_dict = {'Iris-setosa': 0.0, 'Iris-versicolor': 1.0, 'Iris-virginica': 2.0}
+data = loadtxt('iris.data', delimiter=',', dtype=float, \
+        converters={4: lambda s: not s in cnvt_dict and -1.0 or cnvt_dict[s]})
 
-data = loadtxt('iris.data', delimiter=',', dtype=float, converters={4: cnvt})
+# Transform the data set into
 graph_data = [[[] for i in range(3)] for j in range(16)]
-colors = ['r', 'g', 'b']
-figure(16)
-clf()
 for i in range(4):
     for j in range(4):
         if i != j:
             for d in data:
                 graph_data[i + j*4][int(d[4])].append((d[i], d[j]));
+
+colors = ['r', 'g', 'b']
+figure(16)
+clf()
+
 for i in range(4):
     for j in range(4):
         if i != j:
             subplot(4, 4, (i+1) + j*4)
             axis('equal')
+            # Plot the three data sets.
             for c in range(3):
                 tmp = zip(*graph_data[i + j*4][c])
                 plot(tmp[0], tmp[1], 'x' + colors[c])