A1118
Title: Deep neural networks on nonparametric regression for times series data
Authors: Lu Wang - Monash University (Australia) [presenting]
Abstract: Deep neural networks (DNN) for nonparametric regression on time series data are studied. Under a general smooth condition and a suitable restriction on the structure of the regression function, a new DNN estimator is proposed based on sparsely connected multilayer feedforward neural networks with ReLU activation function. An asymptotic normality has been established for the proposed estimator. Empirical performance via simulations and a real-data application on asset returns prediction also validate our theoretical findings.
