Random sample consensus: Difference between revisions
No edit summary |
No edit summary |
||
Line 4: | Line 4: | ||
The idea is to continuously randomly sample a subset of your data.<br> | The idea is to continuously randomly sample a subset of your data.<br> | ||
For each subset, you build a model and do cross-validation until you find a model with sufficiently small error. | For each subset, you build a model and do cross-validation until you find a model with sufficiently small error. | ||
You quit after <math>k</math> iterations or once your model has sufficiently low error. | |||
This is commonly used when you have a lot of points and building a model is worse than <math>O(n)</math>.<br> | This is commonly used when you have a lot of points and building a model is worse than <math>O(n)</math>.<br> |