Suppose we have a function f(x, y), and want to compute \partial_x f(x,y) for many y’s, ideally using no memory allocation. Using Zygote, I tried Zygote.forwarddiff(y->f(x,y), y). However, I believe this creates a new anonymous function each time we call it with a different x. So the call to foward…

[image] SEA: However, I believe this creates a new anonymous function each time we call it with a different x. So the call to fowarddiff is not efficient. Creating an anonymous function (closure) is more or less free after it’s compiled in Julia so I don’t think you need to worry about it. B…

How to compute partial derivatives efficiently with Zygote

Specific Domains Machine Learning

SEA September 13, 2019, 7:23pm 3

Thank you for the pointers, and let’s hope so!

Topic		Replies	Views
Automatic differentiation performance & computing derivatives of only a subset of the arguments Performance question , autodiff	6	970	October 2, 2021
Why does Zygote produce a wrong derivative? Machine Learning zygote	5	1009	August 29, 2020
Calculating specific components of gradients Specific Domains differentiation , zygote	4	1006	April 23, 2021
Allocations in Zygote for a simple derivative Machine Learning performance , zygote	0	695	February 19, 2020
Zygote dozens* of times slower than manually written function Performance zygote , forwarddiff	17	1832	April 21, 2022

How to compute partial derivatives efficiently with Zygote

Related topics