A class U of recursive functions is said to be finitely (a, b) learnab
le if and only if for any b tuple of pairwise distinct functions from
U at least a of the b functions have been learned correctly from examp
les of their behavior after some finite amount of time. It is shown th
at this approach, called learning in parallel, is more powerful than n
onparallel learning. Furthermore, it is shown that imposing the restri
ction (called parallel super learning) on parallel learning that the l
earning algorithm also identiy on which of the input functions it is s
uccessful is still more powerful than nonparallel learning, A necessar
y and sufficient condition is derived for (a, b) superlearning and (c,
d) superlearning being the same power, Our new notion of parallel lea
rning is compared with other, previously defined notions of learning i
n parallel. Finally, we synthesize our notion of learning in parallel
with the concept of team learning and obtain some interesting trade-of
fs and comparisons. (C) 1995 Academic Press, Inc.