Researchers use AI to catalog pi formulas

Just in time for PiDay, a team of researchers at the Israel Institute of Technology have announced the results of using AI-based computer tools to automatically catalog a large body of existing pi formulas.

In particular, these researchers found underlying structure in these formulas, in particular a “conservative matrix field” (CMF) functions as a sort of fingerprint, enabling researchers to automatically categorize pi formulas from 455,050 papers downloaded from the internet. These were first reduced to 385 distinct formulas, then analyzed using the CMF tool.

Their CMF signatures showed that 43% of these formulas all descend from similar roots; another 51% belong to other clusters, and only 6% remain as “orphans” with no clear connection to others.

Lyndie Chiou, a writer for Scientific American, has written a very nice article summarizing their work: HERE.

A preprint of the team’s technical paper, entitled “From Euler to AI: Unifying formulas for mathematical constants,” is available HERE.