turning in project toolbox - machine learning

learning_curve.py

@@ -8,16 +8,20 @@
 
 data = load_digits()
 print data.DESCR
-num_trials = 10
+num_trials = 100
 train_percentages = range(5,95,5)
 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
+for i in range (len(train_percentages)):
+	for j in range(num_trials):
+		X_train, X_test, y_train, y_test = train_test_split(data.data, data.target, train_size=train_percentages[i])
+		model = LogisticRegression(C=10**1)
+		model.fit(X_train, y_train)
+		test_accuracies[i] += model.score(X_test,y_test)
 
 fig = plt.figure()
 plt.plot(train_percentages, test_accuracies)

questions.txt

@@ -0,0 +1,7 @@
+1. Generally, as the percentage of data used increases, so does the accuracy on the test set.
+
+2. It seemed to me that the whole curve appeared to be noisy, especially at a low number of trials. There didn't seem to be an area that was nosier than others.
+
+3. After around 700 or 800 trials, the curve starts to become smooth. 
+
+4. As C gets larger, the graph becomes smoother, and as C gets smaller, the graph becomes very noisy. 

questions.txt~

@@ -0,0 +1,7 @@
+1. Generally, as the percentage of data used increases, so does the accuracy on the test set.
+
+2. It seemed to me that the whole curve appeared to be noisy, especially at a low number of trials. There didn't seem to be an area that was nosier than others.
+
+3. After around 700 or 800 trials, the curve starts to become smooth. 
+
+4. As C gets larger, the graph becomes smoother, and as C gets smaller, the graph becomes very noisy.