Autonomous DNA Neuron Learning Algorithm Based on DNA Strand Displacement

Abstract

DNA neuron learning, or weight update of DNA neurons, is an important research content in the construction of a DNA neural network. In this work, we propose a DNA reaction network to implement the autonomous weight update of DNA neurons based on DNA strand displacement. The DNA reaction network consists of four modules: weight update module, calculation module, synchronization module, and feedback adjustment module to achieve the effectiveness and consistency of multiple training data in DNA neuron learning. Especially, the learning and the testing of DNA neurons are in the same DNA strand displacement reaction system. The simulation results show that the DNA neuron can classify test data correctly, which proves that the algorithm adopted in this work is effective.