Publications

Since its birth in 1943, neural networks proved itself as a powerful and e - cient computational system. A system of neural networks is capable of handling tasks similar to the way a human handles it. Even though it is not yet widely recognized, many researchers around the globe are becoming interested in this eld because of the tendency towards arti cial intelligence it possess. There is a recent interest in the eld of arti cial neural networks (ANN) because of the new techniques and a better understanding of their capabilities. In order to measure the abilities of a neural design, it have been compared to a traditional logical design. Implementations of various computational designs in neural networks have been successfully done. This requires the construction of a neural design with exact computational function as the traditional logical design. The construction of an ANN is usually done manually, which is a time and e ort consuming process. But the constructed neural design may contain more than the needed nodes (neurons) which will a ect the testing results and time complexity of the neural design. Furthermore, manual neural networks implementation of big size designs, in terms of number of inputs and components, is extensively hard to accomplish. So we thought of a solution for this adversity by suggesting an automatic mechanism for implementing computational designs in neural networks. In our paper, an implementation algorithm for ANN is proposed for generat- ing the smallest possible and most e cient neural design while still conserving the function of the source logical design. The algorithm uses our self-prepared neural models of logic gates. It then checks for matches of these models to the logical model at hand. Matching to logical gates is very important for the size of the networks to get less at every match. After checking for matches, the algorithm makes a direct conversion of the remaining cases, if any, into neural analogous. Finally, after taking all components into consideration, it generates an ANN design.