Bias in Polynomials
November 15, 2021
I don’t want to do a mathy-type post. I want to point interested readers to the article “Mathematicians Find Structure in Biased Polynomials.” The idea is that certain expressions “favor particular outputs.” I worked through the write up and looked at a couple of the referenced papers.
Three observations:
- Biased polynomials may provide some insight into why certain machine learning methods go off the rails
- The “cementing” of certain outputs to a particular type of operation warrants review in the context of biased polynomials.
- Smart software may not be smart enough to recognize repetitive outputs, assuming — like some machine learning wizards — that math is objective. It is, but it may be biased.
Net net: An interesting area for inquiry by bright 17 year olds. Maybe there is a connection between biased polynomials and some of the surprising outputs from a Bayesian-dependent process.
Stephen E Arnold, November 15, 2021
Comments
-
- Subscribe to Beyond Search
Feature archive
News archive
-
