This paper presents the results of Persian handwritten word recognition based on Mixture of Experts technique. In the basic form of ME the problem space is automatically divided into several subspaces for the experts, and the outputs of experts are combined by a gating NETWORK. In our proposed model, we used Mixture of Experts Multi Layered Perceptrons with Momentum term in the classification Phase. Applying this term makes three effects in our system: a) increase convergence rate, b) obtain the optimum performance in our system, c) and escape from the local minima on the error surface. We produce three different Mixture of Experts structure. Experimental result for proposed method show an error rate reduction of 6.42% compare to the mixture of MLPs experts. Comparison with some of the most related methods indicates that the proposed model yields excellent recognition rate in handwritten word recognition.

Estimation of evapotranspiration is essential for planning, designing and managing irrigation and drainage schemes, as well as water resources management. In this research, artificial NEURAL NETWORKs, NEURAL NETWORK wavelet model, multivariate regression and Hargreaves' empirical method were used to estimate reference evapotranspiration in order to determine the best model in terms of efficiency with respect to the existing data. The daily data of two meteorological stations of Shahrekord and Farrokhshahr airport in the dry and cold zones of Shahrekord during the period 2013-2004 was used; these included the minimum and maximum temperature, the average nominal humidity, wind speed at 2 meters height and sunshine hours. %75 of the data were validated, and %25 of the data was used for testing the models. Designed NETWORK is a predictive NEURAL NETWORK with an active sigmoid tangent function hidden in the layer. In the next step, different wavelets including Haar, db and Sym were applied on the data and the NEURAL NETWORK-wavelet was designed. To evaluate the models, the method was used by the Penman-Montith Fao and for all four methods, RMSE, MAE and R statistical indices were calculated and ranked. The results showed that the wave-let-NEURAL NETWORK with the db5 wavelet had a better performance than other wavelets, as well as the artificial NEURAL NETWORK, multivariate regression and the Hargreaves method. The results of wavelet NETWORK modelling with the db5 wavelet in the Farrokhshahr station were calculated to be 0. 2668, 0. 2067 and 0. 998, respectively; at the airport station, these were equal to 0. 2138, 0. 14 and 0. 9989, respectively. The results, therefore, showed that the NEURAL NETWORK-wavelet performance was more accurate than the other models studied in this study.

This paper represents a comparison between modified trainable NEURAL NETWORK ensemble with other trainable and non-trainable ENSEMBLES. The historical data available in this case study are from kharg petrochemical company in Tehran stock exchange. This company is one of the biggest producers of petrochemicals, including methanol, in Iran and its stock price is very much dependent on world methanol price. Therefore Kharg stock price reflects its financial information more clearly than others with no products for global exportation. The reason of choosing Kharg is related to its large data history and high rate of stock free float1. The results show how a modified trainable NEURAL NETWORK ensemble can overcome other trainable and non-trainable ENSEMBLES. This study also demonstrates how we can beat the market through our proposed model without the use of extensive market data or knowledge.