Research On DNA Models Of Artificial Neural Networks

DNA computing is a new subject. What the subject studies is how to solve mathematical problems using DNA molecules. Since Adleman solve the Hamilton path problem using DNA molecules in 1994, many mathematical problems were solved by DNA computing such as matrix multiplication, addition, symbolic determinants. What is more, in Nov. 2001 Y. Denenson simulated the transformation of states of finite autonomous. These facts demonstrate that DNA molecules computing by self-assembly according to Waston-Crick complement is powerful.In this paper, we developed three DNA-based models to realize basic functions of neural network. The idea, which is different from traditional techniques that are used to modify powers between neurons, is to pick out the appropriate powers combination from any possible combination . The central feature of these models is to apply the parallel logic.In addition to these DNA-based models , We illuminate the feasibilities of these models through a concrete example. Furthermore, these algorithms have some advantages compared with the existing algorithm which can achieve association .