week 1: reflections

In machine learning by Parker TichkoLeave a Comment

Week 1 of our machine learning “course” entailed reading through the introduction and chapter 2 of An Introduction to Statistical Learning with Applications in R. In these opening chapters, the authors wove together statistical concepts, many that were covered in our required graduate-level statistics courses, with the basics of machine learning. Though I was familiar with some of the introductory content, I enjoyed how the authors framed important statistical and machine learning concepts as dichotomous pairs. Several are listed below:

  • Regression vs. classification
  • Supervised vs. unsupervised learning
  • Variance vs. bias
  • Inference vs. prediction
  • Interpretability vs. flexibility
  • Parametric vs. non-parametric

The introduction provided a great review, but one message really rose above the rest:

“There is no free lunch in statistics; no one method dominates all others over all possible data sets.” (pg. 29)

I suppose one could say this is the art of statistical modeling. Choosing an appropriate model for a data set is a constant balancing act between model flexibility and interpretability – an idea to keep in mind as we progress through the course.

Leave a Comment