Skip to content

jeanimal/double_descent_dar

Repository files navigation

double_descent_dar

Reproducing and playing with some ideas using code from this wonderful but highly mathematical paper:

Yehuda Dar, Muthukumar, V., & Baraniuk, R. (2021). A Farewell to the Bias-Variance Tradeoff? An Overview of the Theory of Overparameterized Machine Learning. https://arxiv.org/abs/2109.02355

Blog

The best summary of what I explore in this repo and WHY is in my blog:

https://medium.com/@jeanimal/how-double-descent-breaks-the-shackles-of-the-interpolation-threshold-5f5990dc02b2

But... it doesn't cover everything here yet.

In this repo

Notebooks:

Libraries:

Give me more!

Older explanations of double descent, focusing more on the necessary structure of the data:

More Code:

About

Reproducing ideas from the paper "Farewell to the bias-variance tradeoff" by Dar, Muthukumar, Baraniuk in python.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors