Question 7 (10 points): Below is the learning curve of a classifier ("classification accuracy on test set" versus "the number of training samples"). The curve is generated by the following protocol: (1) the same classification model is being used with the same hyper-parameter setting for all experiments; the same set of test samples (independent from training set) are used for all experiments. (2) for each training sample size (value in x-axis), we randomly selected a number of training samples (based on that specific sample size), trained a classifier, evaluated the classifier's accuracy on the test set (value in y-axis); In this way, we collect a point on the learning curve. (3) for the same training sample size ( x-axis), we repeated the random training sampling selection process multiple times (e.g., 20 times), and evaluated the corresponding test accuracy. We plotted the mean (red points) and standard deviation (blue bar) in their test accuracy. For example, for training sample size of 2 (the left most point in the x-axis of the curve), the mean and standard deviation of test accuracy are around 62% and 13%, respectively. (In other words, the model has relatively lower test accuracy and is quite unstable.) Please mark out which ranges of x-axis corresponding to "underfitting" and "overfitting" respectively and explain why. If you answer that the curve does not show "underfitting" or "overfitting", please explain why