Finished ML toolbox

learning_curve.py

@@ -8,16 +8,43 @@
 
 data = load_digits()
 print data.DESCR
-num_trials = 10
+num_trials = 50
 train_percentages = range(5,95,5)
-test_accuracies = numpy.zeros(len(train_percentages))
+# test_accuracies = numpy.zeros(len(train_percentages))
 
 # train a model with training percentages between 5 and 90 (see train_percentages) and evaluate
 # the resultant accuracy.
 # You should repeat each training percentage num_trials times to smooth out variability
 # for consistency with the previous example use model = LogisticRegression(C=10**-10) for your learner
 
-# TODO: your code here
+test_accuracies = []
+
+model = LogisticRegression(C=10**-10)
+for i in range(18):
+    percentage = 0.05 * (i+1)
+    train_total = 0
+    test_total = 0
+    for i in range(num_trials):
+        X_train, X_test, y_train, y_test = train_test_split(data.data, data.target, train_size=percentage)
+        model.fit(X_train, y_train)
+        train_total += model.score(X_train,y_train)
+        test_total += model.score(X_test,y_test)
+    train_average = train_total/num_trials
+    test_average = test_total/num_trials
+    print percentage
+    print "Train accuracy %f" %train_average
+    print "Test accuracy %f"%test_average
+    test_accuracies.append(test_average)
+
+
+digits = load_digits()
+print digits.DESCR
+fig = plt.figure()
+for i in range(10):
+    subplot = fig.add_subplot(5,2,i+1)
+    subplot.matshow(numpy.reshape(digits.data[i],(8,8)),cmap='gray')
+
+plt.show()
 
 fig = plt.figure()
 plt.plot(train_percentages, test_accuracies)

questions.txt

@@ -0,0 +1,4 @@
+1.) Upward; generally, the more of the data is used in training, the more accurate the results are.
+2.) Yes, the smaller values. Less training data leads to more inconsistent results.
+3.) At num_trials = 50, the curve seems pretty consistently smooth.
+4.) At larger values of C, the calculations take significantly longer, but test accuracy increases more steeply with percentage of training data. It seems to asymptotically approach 90% accuracy. At smaller values, the calculations are fast, but the accuracy is terrible.