Neural nets for economists. (English) Zbl 0703.90009
The economy as an evolving complex system, Proc. Evol. Paths Global Econ. Workshop, Santa Fe/NM 1987, Santa Fe Inst. Stud. Sci. Complexity 5, 33-48 (1988).
[For the entire collection see Zbl 0678.00029.]
This paper will review two aspects of recent neural networks research which may be of interest for economic modeling. First I will describe the original Hopfield model [see J. J. Hopfield, Proc. Nat. Acad. Sci. USA 79, 2554 (1982); ibid. 81, 3088 (1984)]. This has no direct relevance to economics, but is an example of one approach to modeling a complicated, poorly understood dynamical system, namely the brain. While the Hopfield model has a number of interesting features as a dynamical system, I will also remark that other ‘neural’ circuits may be designed which better fulfill its function of ‘associative memory’. My second subject will be the ‘back-propagation’ learning algorithm [see D. E. Rumelhart, G. E. Hinton and G. E. Williams, in: Parallel Distributed Processing, Vol. 1, D. E. Rumelhart and J. L. McClelland (eds.), Cambridge MA (1986); P. Werbos, “Beyond regression: New tools for prediction and analysis in the behavioral sciences”, Harvard Univ. Diss. (1974)]. As I will describe, this algorithm has potential as an economic predictor.
This paper will review two aspects of recent neural networks research which may be of interest for economic modeling. First I will describe the original Hopfield model [see J. J. Hopfield, Proc. Nat. Acad. Sci. USA 79, 2554 (1982); ibid. 81, 3088 (1984)]. This has no direct relevance to economics, but is an example of one approach to modeling a complicated, poorly understood dynamical system, namely the brain. While the Hopfield model has a number of interesting features as a dynamical system, I will also remark that other ‘neural’ circuits may be designed which better fulfill its function of ‘associative memory’. My second subject will be the ‘back-propagation’ learning algorithm [see D. E. Rumelhart, G. E. Hinton and G. E. Williams, in: Parallel Distributed Processing, Vol. 1, D. E. Rumelhart and J. L. McClelland (eds.), Cambridge MA (1986); P. Werbos, “Beyond regression: New tools for prediction and analysis in the behavioral sciences”, Harvard Univ. Diss. (1974)]. As I will describe, this algorithm has potential as an economic predictor.
