If you use knn in package "class", you probabily will get a memory
error.... Many of members in our lab have encountered the same
problem.
If you write a knn algorithm by yourself and try to do a batch
process, you can only deal with around 80 data samples one time.
According to my computation time for 160 data samples, a knn wrote by
myself will need about 90 hours to process all data.......
I try to find ways to speed up the computation but all in vain.
If you use only the first 10 features, the computation time is
almost the same. (only slightly faster)
※ 引述《Tonnny (^~^)》之銘言：
: According to teacher's explantation in class,
: does it mean that we should't use the knn function?
: Instead, to write the nearest neighbor function by ourselves?