Study of RAN and its application in temperature compensation for sensors

Resource Allocating Network (RAN) is a famous on-line RBF network algorithm, which distributes hidden nodes dynamically as required in the course of learning stage. The main characteristic of RAN is establishing a network of structure tightly packed, and studying speed quicker. RAN can avoid effectively the difficulty of selecting initial parameters, such as hidden nodes and_expansion constant in RBF networks. And it can accomplish on-line learning. After verifying the validity by simulating experiment, we used RAN algorithm in the experiment of temperature compensation for pressure sensors. The results show that the convergence speed of RAN is superior to that of RBF networks, a satisfactory effect of error correction is acquired, and it can meet the requirement of practical application.