Implementation of Basic Math Processing Skills with Neural Arithmetic Expressions in One and Two Stage Numbers

When artificial intelligence technologies and algorithms, which are popular fields of study today, are examined, it is seen that most of the studies are based on classification and estimation results. Artificial intelligence studies carried out on this basis fall into the narrow field of artificial intelligence. However, the point to be reached in artificial intelligence technologies of the future is the general artificial intelligence concept, which predicts models that can comprehend and interpret events. It is predicted that the first step of the transition from the narrow concept of artificial intelligence to the general artificial intelligence technology is to transform the events into mathematics with neural arithmetic expressions. In this study, addition, subtraction, multiplication, division and exponentiation operations were applied on one-digit and two-digit numbers by teaching basic mathematical operations capabilities to an artificial intelligence model developed with the Long Sort-Term Memory (LSTM) algorithm. The data set required for learning is generated instantaneously and randomly. The data set required to learn one-step operations was created from 1000 data and subjected to 100 repetitive learning. The data set required to learn two-step operations was created from 50,000 data and subjected to 80 iterative learning. A success rate of 99.9% was achieved in operations performed with one-digit numbers, and 95.6% in operations performed with two-digit numbers.