Answer.txt

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.