ABSTRACT
This chapter details the principles of artificial neural network (ANN), including the structural parameters of the ANN, the method used to select training samples, the data processing method, and the training algorithm. It presents the principles of the ANN method, including structural parameters, the method used to select training samples, the data processing method, and the training algorithm, all of which can affect ANN performance. The chapter presents the inverse design methods based on computational-fluid-dynamics (CFD) and the ANN, including direct and indirect methods. In the design of an indoor environment, an ANN is typically used to establish the non-linear relationship between the design variables and design objectives. When a CFD-based ANN is used in the design of a building environment, the values of the design variables are obtained by means of a sampling method. A well-trained ANN represents a mathematical transformation between the multidimensional domain spaces of the input and output.
