Evaluation of Chaos in Plant Response and Its Identification Using Neural Networks

Abstract

Plant responses are characterized by complexity and uncertainty. In this study, a neural-network identification technique, taking the concept of chaos into consideration, is discussed. aiming at a more precise identification. In the method, the complexity of the data (diurnal change in the net photosynthetic rate of the plant grown in greenhouse) is quantitatively evaluated by introducing the concept of chaos before performing identification. Here, attractor of the data is described in the phase space, and the fracta1 dimension is calculated to measure the irregularity. Finally, the net photosynthetic rate, as affected by light intensity, are identified based on these results using a neural network. Relationships among the shapes of attractors, fraetal dimension and identification (or prediction) error were investigated. The result showed that the combinatorial uses of fraetal dimension and neural networks allowed a more precise nonlinear identification of the plant responses to be successfully achieved.