Potential Applications of Artificial Neural Networks to Thermodynamics

The traditional GMDH method (Farlow, 1984 Ivakhnenko, 1971) is based on anrm-derlying assumption that the data can be modeled by using an approximation of the Volterra Series or Kolmorgorov-Gabor polynomial (Madala and Ivakhnenko, 1994) as shown in equation (1). 

It is possible to train a GMDH-type network to predict the output values y using training data that is 

This equation is tested for fit by determining the mean square error of the predicted y and actual values as shown in equation (4) using the set of testing data. 

General coimection between inputs and output variables can be expressed by equation (1). For most application the quadratic form of only two variables is used in the form to predict the output y. 

In the basic form of the GMDH algorithm, all the possibilities of two independent variables out of total n input variables are taken in order to construct the regression polynomial in the form of eqtiation (5) that best fits the dependent observations y i = 1, 2,. Ai) in a least-sqtiares sense. Conseqtiently, 

---

Sharma, R., Singhal, D., Ghosh, R., and Dwivedi, A. (1999). Potential applications of artificial neural networks to thermodynamics Vapor-liquid equilibrium predictions. Comput. Chem. Eng. 23, 385-390. 

Potential Applications of Artificial Neural Networks to Thermodynamics... 

---