diff --git a/Answer.txt b/Answer.txt
new file mode 100644
index 0000000..851b8fd
--- /dev/null
+++ b/Answer.txt
@@ -0,0 +1,11 @@
+1.
+The Accuracy on Test Set increases as Percentage of Data Used for Training increases, but the slope keeps decreasing slightly. 
+2.
+When percentage used of data is relatively small, the curve is noisier. Standard deviation is relatively high when the dataset is relatively small.
+3.
+I set num_trails as 1000 and it looks smooth, but still need more to make it perfectly smooth
+4.
+As C gets bigger, the bending of learning curve becomes more obvious. C is the 'inverse of regularization strength', so bigger C means weaker regularization, which make the graph smoother. 
+
+
+
diff --git a/learning_curve.py b/learning_curve.py
index 2364f2c..90bd6bf 100644
--- a/learning_curve.py
+++ b/learning_curve.py
@@ -8,16 +8,21 @@
 
 data = load_digits()
 print data.DESCR
-num_trials = 10
+num_trials = 1000
 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
+curve_y = []
+model = LogisticRegression(C=10**-10)
+for i in train_percentages:
+	score = []
+	for j in range(num_trials):
+		X_train, X_test, Y_train, Y_test = train_test_split(data.data, data.target, train_size=i)
+		model.fit(X_train, Y_train)
+		score.append(model.score(X_test, Y_test))
+	test_accuracies[i/5-1] = sum(score)/num_trials
+
 
 fig = plt.figure()
 plt.plot(train_percentages, test_accuracies)