Guido Fioretti and Bernd Süßmuth
Business Cycle Dynamics in a Neural Net Framework International Journal of Computer Research, 10 (2) 2001: 201-222.
Abstract
In this article, an unsupervised neural network with a negative feedback ---
also known as "novelty filter" in the litterature on neural networks ---
implements a disaggregated investments accelerator.
We ran the net with empirical data of disaggregated demand for the U.S. economy,
finding that our net is able to reproduce actual equipment investment series with good accuracy throughout the sample data.
Due to unfortunate misunderstandings, this article was published when it was
still at a preliminary version. The text presents the neural network as
having a positive (rather than a negative) feedback.