liveBook · Manning

Algorithmic statistics

As explained in Sutskever's List

Algorithmic statistics is the study of separating structure from noise in individual datasets using algorithmic descriptions and measures of typicality.

A black swan challenges the model “all swans are white”: it is an awkward misfit, while a model allowing black or white swans makes it typical. Algorithmic statistics studies how to separate structure from noise in an individual dataset by pairing a simple model with data that is typical under it. section 8.2.1

Kolmogorov complexity measures simplicity through the length of a description. Randomness deficiency measures typicality: low deficiency means the data fits what the model would normally produce; high deficiency signals a misfit. Together, these measures make the idea of a concise, unsurprising scientific explanation more precise. section 8.2.1 section 8.2.1

How Sutskever's List covers it

How deep it goes
Formalizes simplicity with Kolmogorov complexity and typicality with randomness deficiency, including an intuitive account of the latter.
Examples
Contrasts a black swan under an all-white-swans model with one allowing black or white swans.
What you can do after
Evaluate explanations by asking whether their models are concise and their data typical under those models.
How it connects
Connects Kolmogorov complexity and randomness deficiency to the algorithmic framing of scientific explanation.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Sutskever's List. The section links point to the parts of the book it draws on.