3 ms·
This idea seems similar to local linear regression [1] in a way: approximate a non-linear model with multiple linear models that are fit to different data selec
by malcolmgreaves 10y ago
This idea seems similar to local linear regression [1] in a way: approximate a non-linear model with multiple linear models that are fit to different data selections.
Neat idea wrt. using a Naive Bayes classifier here!
[1] https://en.wikipedia.org/wiki/Local_regression https://en.wikipedia.org/wiki/Local_regression
- KennyCason 10y agoInteresting, I actually implemented Local Regression from that very Wiki page! https://github.com/kennycason/ml/tree/master/src/main/java/kenny/ml/regression/nonlinear/local https://github.com/kennycason/ml/tree/master/src/main/java/k... This idea originally came to me after explaining Random Forests to someone as a way to improve their Decision Tree. On a whim, I built out a prototype and it offered improved results. :)