search

ArdanaLabs / audit

0 stars 0 forks source link

[research] - any provable properties of Morgan's interval search? #29

Open quinn-dougherty opened 2 years ago

quinn-dougherty commented 2 years ago

Description

Early this week Morgan is implementing an original interval search algorithm, I have a description of it in mattermost, we'd like to know if there are any provable properties of it.

quinn-dougherty commented 2 years ago

we'll have post-hoc analysis (https://github.com/ArdanaLabs/audit/issues/30), this is for a priori insights.

quinn-dougherty commented 2 years ago

I took somewhere between 1 and 2.5 hours of math on this, I convinced myself that the assumptions of the solver are valid and I convinced myself that a = 0 is a sufficient initial guess for the left endpoint. I have not yet learned anything interesting about the behavior of the solver a priori, waiting on data