5,337
edits
Deep Learning: Difference between revisions
→Nonconvex-nonconcave min-max opt
| Line 1,243: | Line 1,243: | ||
The goal is to find a local saddle point. | The goal is to find a local saddle point. | ||
====Stability==== | |||
If we drift away from <math>(x^*,y^*)</math> then the optimization is unstable. | If we drift away from <math>(x^*,y^*)</math> then the optimization is unstable. | ||
If we remain close, the optimization is stable even if we never converge. | If we remain close, the optimization is stable even if we never converge. | ||