predicting apparent metabolizable energy of wheat and corn based on the nutrient components and essential amino acids in broilers , using artificial neural network

three artificial neural networks (ann) models; general regression neural network (grnn), redial basis function (rbf) and three layer multiple perceptron network were carried out to evaluate the prediction of the apparent metabolizable energy (ame) of wheat and corn from its chemical composition in broiler. input variables included: gross energy (ge), crude protein (cp), crude fiber (cf), ether extract (ee), ash and phosphorous as well as essential amino acids profiles (arg, cys, his, ile, leu, lys, met, met+cys, phe, thr and trp). output variable was  ame of wheat or corn feedstuffs. the results showed that r2 ofthree layers perceptron neural network is higher than other two models in both wheat and corn. the best estimation for wheat and corn resulted from the cp (r2=0/89) and ge (r2=0/97) inputs, respectively. in wheat, rbf model had better estimation than grnn model in all inputs except for the amino acids input. the rbf model was poorly estimated only with gross energy input. in corn, grnn model has lower estimation than two other networks except gross energy input. thus it was concluded that the artificial neural networks can be a powerful tool for predicating metabolizable energy from its chemical composition than multiple linear regression in broilers