Approaching (Almost) Any Machine Learning Classification Problem.

If you are doing sequence labeling, learning something about the data, tackling partially unlabeled data or time-varying data, you generally have to take a different approach.