$3.$ A shortage of data. We have a binary classification problem with very high$-$dimensional data: there are

a million features. Unfortunately, we only have $1000$ training points. Nonetheless, we train a support

vector machine classifier and find that it works well in practice. What is a