Skip to content

Conversation

fonnesbeck
Copy link
Member

@fonnesbeck fonnesbeck commented Oct 7, 2016

Currently the interval transformation does not properly transform the endpoints, making them inf and -inf. Also, adjust_precision has been modified to choose less extreme values if any Hessian elements are very large or small.

Closes #1428

mag = sqrt(abs(tau))

bounded = bound(log(mag), log(1e-10), log(1e10))
bounded = bound(log(mag), log(0.1), log(10))
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Doesn't this depend on the scale of the data?

@fonnesbeck
Copy link
Member Author

Ideally it should, yes, but if they are too large or small they don't make useful scalings. This is what caused the failure in the associated issue.

@fonnesbeck fonnesbeck merged commit 3eb171f into master Oct 8, 2016
@fonnesbeck
Copy link
Member Author

Dammit. Hit the merge button by accident. Can someone undo this? I am on a mobile device.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants