Adaptive interaction is a new approach to introduce adaptability into man-m
ade systems. In this approach, a system is decomposed into interconnected s
ubsystems that we call devices and adaptation occurs in the interactions. M
ore precisely, interaction weights among these devices will be adapted in o
rder to achieve the objective of minimizing a given cost function. The adap
tation algorithm developed is mathematically equivalent to a gradient desce
nt algorithm but requires only local information in its implementation. One
particular application of adaptive interaction that we study in this paper
is in neural networks. By applying adaptive interaction, we can achieve es
sentially the same adaptation as that using the well-known back-propagation
algorithm but without the need of a feedback network to propagate the erro
rs, which has many advantages in practice. A simulation is provided to show
the effectiveness of our approach. (C) 1999 Elsevier Science Inc. All righ
ts reserved.