Hello,
hello ,
I need a matlab code for using custom function(distance function) for distance parameter with ficknn or classificationknn.
In fact i want to define the distance below D(sq,si)=1/(1+exp(wi))*eucludien(xq,xi)
where
sq:the query instance
wi is a weight of instance i
matlab syntax for knn:model=fitcknn(instance_train,label_train,'Distance',distance function,....)
Thank you in advance
Hello Nedia,
You can create your own distance metric like this example
If Mdl is an ExhaustiveSearcher model object, then you can also specify a function handle for a custom distance metric using @ (for example, @distfun). The custom distance function must:
Have the form function D2 = distfun(ZI, ZJ).
Take as arguments:
A 1-by-K vector ZI containing a single row from X or from the query points Y
An m-by-K matrix ZJ containing multiple rows of X or Y
Return an m-by-1 vector of distances D2. D2(j) is the distance between the observations ZI and ZJ(j,:).
Example for a chi square distance with weights
w = [0.4; 0.6];
chiSqrDist = @(x,Z)sqrt((bsxfun(@minus,x,Z).^2)*w);
https://es.mathworks.com/help/stats/classification-using-nearest-neighbors.html#btap7nm
or check this other page to understand how to use the metrics or incorporate your own distance.
https://es.mathworks.com/help/stats/searcher.knnsearch.html
Regards Eduardo
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Writing