One look at this paper and a reader
begins to wonder about the true scope of Machine Learning. For people who have
started on their Machine Learning journey, the content of this paper can prove
to be quite daunting and many of the terms absolutely alien. So, it is easy for
them to miss out the roadmap that Pedro Domingos has laid out for us to the
Machine Learning journey.
Pedro points out that there is a lot of “folk knowledge” that
is utilized behind building a good Machine Learning model. Knowledge of the different
techniques is a good start towards being successful on this journey, however,
it is only one of the skills that a Data Scientist needs to possess.
As many have come to realize (any many more will), the time
consumed in actually doing machine learning is very little compared to the time
consumed in gathering, integrating, cleaning & pre-processing data, and in
performing trial and error on feature design. The paper brings out 1) the
aspects necessary to build a model, 2) common traps and pitfalls that modelers
should avoid & 3) tips on how to bring out the best from a machine learning
exercise.
To give a snapshot of what was covered in the paper, below is
a graphical representation of the roadmap presented.
