Tom Murphy's "GradIEEEnt half decent" paper

[Job van der Zwan]

Every year Tom Murphy writes an absurdly impractical compsci paper for SIGBOVIK, a conference that"[welcomes] the three neglected quadrants of research: joke realizations of joke ideas, joke realizations of serious ideas, and serious realizations of joke ideas. (In other words: SIGBOVIK is an evening of tongue-in-cheek academic presentations, a venue for silly ideas and/or executions.)"

For SIGBOVIK 2023 he explored how to abuse the rounding errors of IEEE-754 floating points to create arbitrary non-linear functions using only linear operations, and used the impractical ideas this resulted in for equally impractical purposes:

http://tom7.org/grad/

I'm not sure if the people in this group would find this amusing or a crime against numbers (or both), but either way it seems appropriate to share it given the chosen topic. I'm also curious if posits would be easier or more difficult to "abuse" in this way.

Kind regards,

Job van der Zwan

[John Gustafson]

Thanks for this, Job. Yes, it was both amusing and a crime against numbers. The exact same approach will work for posits, though of course the formulas will be quite different. If you were to use Type I unums it wouldn't work at all, though; the repeated linear functions would produce linear bands that get slowly wider and wider, not the desired nonlinear function. Similarly with Type II unums, SORNs, and valids (the interval form of posits).

I didn't know about Tom Murphy VII. Quite an amazing guy!

John

--
You received this message because you are subscribed to the Google Groups "Unum Computing" group.
To unsubscribe from this group and stop receiving emails from it, send an email to unum-computin...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/unum-computing/a752f902-ce50-4e3d-8472-e26ad41383bcn%40googlegroups.com.