2009/5/14 Li Yang <lyshane(a)umich.edu>
Thanks Thomas for the reply. I was reading the paper,
learning algorithm by Aha and Kibler (1991), that Weka IB1 and IBk
classifiers implement. My impression was that in the paper, IB1, IB2, and
IB3 refer to three different instance-based algorithms. My guess was that I
could specify the number of neighbors, k, for each of these algorithms.
In other words, the number in the names IB1, IB2, and IB3 in the paper does
not seem to correspond with the number of neighbors I choose but denote the
three variations. So, which algorithm could the Weka IBk implementation be?
Maybe I should look at the code long enough to figure it out?
A quick look into the paper learns me the following:
"The IBL algorithms described in this paper employ either the nearest
neighbor or k-nearest neighbor classification function. The former
classifies an instance as being a member of the same concept as its most
similar instance. The latter does the same, but takes a majority vote among
its k most similar instances (we set k to 3)."
you can read the
"IBk - K-nearest neighbours classifier. Can select appropriate value of K
based on cross-validation. Can also do distance weighting"
with other words, Weka defines k as number of nearest neighbours, so IB3
would be IBk with 3 NN.
I did not find a reference to these classifiers in the mentioned paper,
although I saw they used some extensions and variations (which maybe totals
up to 3 different IBL algorithms).
Departement of Knowledge Engeneering
Faculty of Humanities & Science