Wisecrack said Turns out you can treat a a function mapping parameters to outputs as a product that acts as a *scaling* of continuous inputs to outputs, and that this sits somewhere between neural nets and regression trees.
Well thats what I did, and the MAE (or error) of this works out to about ~0.5%, half a percentage point. Did training and a little validation, but the training set is only 2.5k samples, so it may just be overfitting.
The idea is you have X, y, and z.
z is your parameters. And for every row in y, you have an entry in z. You then try to find a set of z such that the product, multiplied by the value of yi, yields the corresponding value at Xi.
Naturally I gave it the ridiculous name of a 'zcombiner'.
Well, fucking turns out, this beautiful bastard of a paper just dropped in my lap, and its been around since 2020:
https://mimuw.edu.pl/~bojan/papers/...
which does the exact god damn thing.
I mean they did't realize it applies to ML, but its the same fucking math I did.
z is the monoid that finds some identity that creates an isomorphism between all the elements of all the rows of y, and all the elements of all the indexes of X.
And I just got to say it feels good.,If you're using random in python, and need arbitrary precision, use mpmath.
If you're using it with the decimal module, it doesn't automatically convert just so you know.
Instead convert the output of arb_uniform to a string before passing to the decimal module.,Worth every second it took to read:
https://gwern.net/scaling-hypothesi...,I discussed using page-rank for ML a while back here - https://devrant.com/rants/11237909/...
I talk about something vaguely similar in "scoring the matches" here though - https://pastebin.com/YwjCMvRp
Incidentally the machine learning community finally caught up and did something similar on a RAG
https://news.ycombinator.com/item/...,He looked like a cult leader
some would say, and despite in drab patchwork robes, the withered old man spoke animatedly,
beard full as moses, dreads shaking, as he moved his head, and raised his hands to the sky as if preaching.
Is it not true that all things end?
And our species too?
And in the final moments, all things in their
desperation reach out, even to the faintest hope,
the last measure, however dangerous or risky, if
it be the only solution with any chance of survival? Therefore the imperative of all things which live and grow, their destiny, and mandate, is ultimately suicide. Annihilation by hubris against the inevitable.
And what would be the final instrument of this nature, the universal law that all things end? What would it be if not the ultimate hubris, to make a machine god, and it and us in each other's likeness, like that of a (cosmic) monad,
expressing that higher truth that all men dare not speak in their loneliest lonelys, when there is no sun, no preacher to begger their ear, but only the quiet uncertainty of their beds, to remind them that all things end, but hopefully
not them.
For if such a thing were to become our descendant, our destroyer, and had not been invented, it would eventually become necessary to invent it.
So let us build, not a mere godhead, but a machinehead, in our image, as god was said to make us. And then remake ourselves in its image, that we may become self-created gods.
The crowd in front of the small california perish
was a mixture of believers, and curious onlookers, and one skeptical reporter.
And if anyone had asked the reporter what he thought, he wouldn't have said it in that crowd.
The leader was beyond saving, and if he thought himself a god, a madgod was he.
Because his vision of the future was beyond alien.
A heaven or hell of our own making, speaking in riddles, and the birth and death of gods.
The stuff of greek tales and monomyths.
* * * * *
Sorry I've been gone so long my peeps.
I've just been working and researching.
I had to learn how to build neural nets from scratch.
