New to Flux! I want to compute the gradient and hessian for this loss function. loss(x,y) = -sum(y.*log.(model(x)) .- (1 .- y).*log.(1 .- model(x))) * 1 // size(y, 2) where model = Flux.Dense(in,out,σ) And how do you access the final values for hessain and gradient once the training is over

[image] [Flux] Issue while computing Hessian MethodError: objects of type Tracker.Grads are not callable Machine Learning I’m new to Flux and I’m trying to compute the hessian. Here’s the code: grad = Flux.gradient(() -> loss(x, y), Flux.params(model)) hess = Flux.gradie…

How to compute hessian

Specific Domains Machine Learning

baggepinnen August 26, 2019, 11:14pm 2

Topic		Replies	Views
[Flux] Issue while computing Hessian MethodError: objects of type Tracker.Grads are not callable Machine Learning flux	12	1439	August 28, 2019
Gradient of a Hessian dependent loss function Machine Learning flux	1	633	November 21, 2019
Is there an efficient way to compute the Hessian of a NN? Machine Learning flux	11	4819	November 12, 2019
Hessian inside a Flux loss function Machine Learning question	3	829	February 26, 2021
Hessian matrix of ML model General Usage flux , zygote , forwarddiff , reversediff	9	2361	April 28, 2021

How to compute hessian

Related topics