Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Issues with SR1 algorithm #221

Open
stefan-k opened this issue May 11, 2022 · 0 comments
Open

Issues with SR1 algorithm #221

stefan-k opened this issue May 11, 2022 · 0 comments
Labels
help wanted Extra attention is needed

Comments

@stefan-k
Copy link
Member

The symmetric rank-one (SR1) algorithm fails under certain conditions. In particular, it produces search directions which are not descent directions and therefore not accepted in the line searches. It is unclear to me whether this is a problem of the implementation or a general problem of this method. I was unable to find another implementation to compare to.

The example succeeds for this initial parameter vector:

let init_param: Array1<f64> = array![-1.2, 1.0, -5.0, 2.0, 3.0, 2.0, 4.0, 5.0];

but fails for this initial parameter vector:

let init_param: Array1<f64> = array![-1.2, 1.0, -5.0, 2.0, 3.0, 2.0, 4.0, 5.0];

The implementation includes a rule for skipping the Hessian update for cases where the denominator of the Hessian update is very small, but this rule does not seem to kick in for most if not all cases I've tried.

Any help on this is highly appreciated!

@stefan-k stefan-k added the help wanted Extra attention is needed label May 11, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
help wanted Extra attention is needed
Projects
None yet
Development

No branches or pull requests

1 participant