0
$\begingroup$

An Introduction to Statistical Learning with Applications in R 2nd edition by Hastie et al. says that

Statistical learning refers to a set of tools for making sense of complex datasets.

How is it different from Machine Learning then?

Is Machine Learning a subset of Statistical Learning?

Improve this question
$\endgroup$
3
  • 3
    $\begingroup$ "statistics" refers to a set of tools for making sense of data (whether complex or not). I am not sure how the "learning" is really adding anything. Machine Learning, when done well, is a computationally focussed branch of statistics IMHO. $\endgroup$
    – Dikran Marsupial
    Commented Jun 3, 2023 at 16:14
  • 3
    $\begingroup$ I would describe it as a continuum with no clear demarcation but some topics which are generally more associated with one camp or the other. Some associated discussion on a thread about PCA: stats.stackexchange.com/a/584028/121522 $\endgroup$
    – mkt
    Commented Jun 3, 2023 at 16:14
  • $\begingroup$ There no consistent definitions. But the definitions I find most compelling emphasize that statistical learning is about improving your mental schema about how the world works whereas machine learning is about autonomously generating mappings without explicit schemas of the world, variables in it and relationships/interactions therein. $\endgroup$
    – jbuddy_13
    Commented Jun 3, 2023 at 22:24

1 Answer 1

Reset to default
5
$\begingroup$

I think any answers to this question will be verging on opinion-based, but I would say there is a gradient from

  • theoretical or pure statistics, focused on rigorous proofs of the properties of various statistical procedures or tests;
  • applied statistics, more interested in how procedures can be used with real data sets;
  • computational statistics, which focuses on algorithms and computational properties of procedures;
  • statistical learning, which asks how we can use computationally efficient, scaleable procedures to learn about patterns in data, but still using a statistical framework to understand how these procedures work;
  • machine learning, which is also interested in computationally efficient, scaleable procedures, but is less interested in the statistical properties of the answers;
  • artificial intelligence, which generalizes machine learning to a much broader framework of 'computer architectures to solve problems'.

Statistical learning and machine learning in particular are very similar, but statistical learning is a little closer to statistics and machine learning is a little closer to computer science. Someone who works in SL is more likely use confidence intervals to describe uncertainty, while someone who works in ML would (more likely) use risk bounds. People who do SL are generally interested in both prediction and inference, while ML tends to be more focused on prediction (although not exclusively: quantifying variable importance can be thought of as a form of inference). For what it's worth, Wikipedia says

Some statisticians have adopted methods from machine learning, leading to a combined field that they call statistical learning

There a big overlaps between each step in this gradient, and it's arguably not a strict gradient (for example, you could argue that computational statistics and statistical learning are overlapping subsets of applied statistics).

Improve this answer
$\endgroup$
1
  • $\begingroup$ Artificial intelligence is a bit difficult on this scale. There is an intersection between machine learning and artificial intelligence and in a way it is a generalization, but in another way it is a specialization. I would say that artificial intelligence is any artificial tool that performs human-like intelligent tasks. AI is a specialization of machine learning in that it is more specifically about intelligence tasks. It is a generalization in a way that it can relate to tools that have little to do with statistics, randomness, or learning (e.g. a TI-30 calculator) $\endgroup$
    – Sextus Empiricus
    Commented Jun 3, 2023 at 21:29

Not the answer you're looking for? Browse other questions tagged or ask your own question.