How to fit any dataset with a single parameter
1–10 of 155 posts
Re: How to fit any dataset with a single parameter
#2very silly. :)
Re: How to fit any dataset with a single parameter
#3more digits than I care for
Re: How to fit any dataset with a single parameter
#4The word “single parameter” is doing a lot of work in this statement.
Re: How to fit any dataset with a single parameter
#5>The purpose of this paper is to show that all the samples of any arbitrary dataset X can be reproduced by a simple differentiable equation: fα(x) = sin^2 (2^xτ arcsin √α)
So... use a PRNG?
Re: How to fit any dataset with a single parameter
#6This is beautiful. The real numbers are uncountable, so seems intuitive that could approximate any other space with them... basically a hash function. But this one has an inverse! It's so cool that someone actually implemented this.
Re: How to fit any dataset with a single parameter
#7This sounds like it could explain many ML models ;)
Re: How to fit any dataset with a single parameter
#8>The purpose of this paper is to show that all the samples of any arbitrary dataset X can be reproduced by a simple differentiable equation: fα(x) = sin^2 (2^xτ arcsin √α) So... use a PRNG?
Doesn't this exclude negative numbers?
Re: How to fit any dataset with a single parameter
#9First, this is a fun implementation and I love it.
Second, you could as easily embed an infinite size dataset into an infinitely long binary string and say that you've reproduced your dataset with a 'single' parameter! That's sort of what this is doing, with some extra steps.
Re: How to fit any dataset with a single parameter
#10The word “single parameter” is doing a lot of work in this statement.
intentionally:
> In addition to casting doubts on the validity of parameter-counting methods and highlighting the importance of complexity bounds based on Occam’s razor such as minimum description length (that trade off goodness-of-fit with expressive power), we hope that fα may also serve as entertainment for curious data scientists