An effective SteinGLM initialization scheme for training multi-layer feedforward sigmoidal neural networks. (July 2021)

• Record Type: Journal Article
• Authors: Yang, Zebin; Zhang, Hengtao; Sudjianto, Agus; Zhang, Aijun
• Is Part Of: Neural networks. Volume 139(2021)
• Page Start: 149
• Page End: 157
• Publication Date: 2021-07
• Subjects: Multi-layer feedforward neural network -- Initialization scheme -- Stein's identity -- Multi-index model -- Generalized linear model; Neural computers -- Periodicals; Neural networks (Computer science) -- Periodicals; Neural networks (Neurobiology) -- Periodicals; Nervous System -- Periodicals; Ordinateurs neuronaux -- Périodiques; Réseaux neuronaux (Informatique) -- Périodiques; Réseaux neuronaux (Neurobiologie) -- Périodiques; Neural computers; Neural networks (Computer science); Neural networks (Neurobiology); Periodicals; 006.32
• DOI: 10.1016/j.neunet.2021.02.014
• Languages: English
• ISSNs: 0893-6080

Abstract

Abstract: Network initialization is the first and critical step for training neural networks. In this paper, we propose a novel network initialization scheme based on the celebrated Stein's identity. By viewing multi-layer feedforward sigmoidal neural networks as cascades of multi-index models, the projection weights to the first hidden layer are initialized using eigenvectors of the cross-moment matrix between the input's second-order score function and the response. The input data is then forward propagated to the next layer and such a procedure can be repeated until all the hidden layers are initialized. Finally, the weights for the output layer are initialized by generalized linear modeling. Such a proposed SteinGLM method is shown through extensive numerical results to be much faster and more accurate than other popular methods commonly used for training neural networks.