You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on May 9, 2022. It is now read-only.
All of the weights presumably have the same magnitude. The current logic sums them in descending order, adding highest weights first, which maximizes the approximation error of floating-point arithmetic. With #24 as it stands now, they will be summed in randomized order. To minimize the approximation error, the smallest terms should be added first.
The text was updated successfully, but these errors were encountered:
Sign up for freeto subscribe to this conversation on GitHub.
Already have an account?
Sign in.
bat-publisher/index.js
Line 376 in 13686d8
All of the weights presumably have the same magnitude. The current logic sums them in descending order, adding highest weights first, which maximizes the approximation error of floating-point arithmetic. With #24 as it stands now, they will be summed in randomized order. To minimize the approximation error, the smallest terms should be added first.
The text was updated successfully, but these errors were encountered: