Hi all.
Suppose you are looking at a scatterplot of 2-D data in matlab and you are envisioning some type of boundary to best separate the data. One thing you could do is train a logistic regression or SVM or random forest or whatever method you like to separate the data. However, you need to get your parameters correct if you want it to split the data nicely.
If you can already picture in your mind how you want to separate the data, then perhaps you would like to draw in the decision boundary yourself. There would also be code to test which side of the decision boundary a new point is on. So once you're done drawing the decision boundary, it will behave like an SVM in that you can get a binary answer of which class a new point is in.
I'm just trying to gauge the interest for this. If there is an interest I'll probably make a github account and put up some matlab code.
EDIT: this would be for a 2-D decision boundary only (and I realize some of those methods I mentioned are kind of overkill for 2-D)
[link] [comment]