StudioInfinity/dyna3
2
0
Fork 2
Code Issues 65 Pull requests 2 Projects 3 Releases 4 Wiki Activity Actions

Use curly apostrophes

Vectornaut 2025-10-17 09:07:22 +00:00
parent 3aa5614f8b
commit c9167e0f6f
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
Numerical-optimization.md

@ -22,11 +22,11 @@ We saw above that the vanishingly rare gradient descent paths that lead to saddl
#### Affected optimization methods #### Affected optimization methods
Uniform regularization can be seen as an interpolation between Newton's method and gradient descent, which kicks in when lowest eigenvalue of the Hessian drops below zero and brings the search direction closer to the gradient descent direction as the lowest eigenvalue gets more negative. Since the Hessian is indefinite near a saddle point, Newton's method with uniform regularization should act at least sort of like gradient descent near a saddle point. This suggests that it could get bogged down near saddle points in the same way. Uniform regularization can be seen as an interpolation between Newton’s method and gradient descent, which kicks in when lowest eigenvalue of the Hessian drops below zero and brings the search direction closer to the gradient descent direction as the lowest eigenvalue gets more negative. Since the Hessian is indefinite near a saddle point, Newton’s method with uniform regularization should act at least sort of like gradient descent near a saddle point. This suggests that it could get bogged down near saddle points in the same way.
## Methods ## Methods
### Newton's method ### Newton’s method
_To be added_ _To be added_