Universal Functions Originator—Part I: Design
Ali Ridha Al-Roomi, M.E. El-Hawary · 2019
Many computing systems have been introduced in the literature as tools to perform many applications, such as: pattern classification, function approximation, categorization/clustering, control, forecasting/prediction, and optimization. Such these tools are linear/nonlinear regression analysis and all the flavors of artificial neural networks (ANNs) and support vector machines (SVMs). Linear regression (LR) is used for simple data where the relation between its coefficients is linear, while the nonlinear regression (NLR) is used when that relation is nonlinear. ANNs and SVMs are more efficient and they can be used for complicated applications. However, each one of these approaches has its own strengths and weaknesses. In this study, a new computing system called a universal functions originator (UFO) is introduced; with a patent-pending status. This system can generate highly complicated mathematical models, as well as simplifying them, automatically through two optimization stages. Any arithmetic operator (including addition, subtraction, multiplication, and division) and any mathematical function (including exponential, logarithmic, trigonometric, hyperbolic functions) can be dragged into the pool of the search space. UFO has been designed and tested with a wide range of problems, and it shows an impressive performance with many promising uses and capabilities. The objective of this paper is to reveal its mechanism and the main concepts behind it.