Let A and B be two independent sets, and ai, bj are any two elements of
set A and B respectively. For ai in set A, calculate the similarity
between ai and bj using a KERNEL and find out the most similar element to
ai in B, say bi', repeate this process for all elements in A (a1, a2,
a3...an) and add up all the similarities of the most similar pairs (ai,
bi').
For the elements in B(b1, b2, b3...bn), perform the same process to find
the most similar elements in A and also add up these similarities. Then
add up all the similarities obtained from the above steps. I'm wondering
if this result is still a kernel?
Thanks in advance for your reply.
|