Finely tuned models sacrifice explanatory depth — mentioned by improbable22

Thu, 02 Jan 2020 04:13:04 GMT

Half the article is about the social observation that making your work unnecessarily complicated is, too often, a way to impress people. Especially those, like funding agencies, who try to judge without understanding it. Many will be impressed to see you flexing large calculational muscles, even if they can't quite see why. The other half appears to be an advert for the author's paper https://arxiv.org/abs/1910.13608 which advocates a particular measure they call "explanatory depth". It's not immediately obvious (to me) how this works, or how it relates to other bayesian & information-based…

Paper and discussion history. A mention is not an endorsement.

Finely tuned models sacrifice explanatory depth · HackerPapers discussions

Finely tuned models sacrifice explanatory depth — mentioned by improbable22