Multivariate sigmoidal neural network approximation (Q553274)

The author considers multivariate neural network approximations to real and complex-valued continuous functions over boxes or over \({\mathbb R}^N\). The multivariate quasi-interpolation operators employ a tensor-product density function based on the logarithmic sigmoidal function. The paper establishes Jackson type inequalities in terms of the multivariate modulus of continuity of the function or its partial derivatives.