Optimization of Neural Network Input Through Genetic Algorithm
Ahmad M. Bataineh, Ameer Mubaslat · 2021
Big data is utilized and developed across a range of fields, such as; engineering, pharmaceutical, medical, ...etc. In the past decade, Reliable data acquisition has become increasingly in demand, and recent developments and advancements made with regard to Machine Learning (ML) technologies have served to only increase the interest in the proper acquisition, processing, and utilization of big data. These components are almost certainly dependent on the quality of available data. Accordingly, in this paper we introduce a new approach towards the utilization of the Genetic Algorithm (GA) systems and auto encoders alongside neural networks to aid in the tasks of machine vision and decision-making processes. The presented approach aims at improving neural network training parameters through the exploitation of real-coded GAs in the enhancement of the imagery and sensory inputs used in training the network. As such, The GA is utilized to adapt to several parameters of an input imagery and conforms to a local enhancement technique similar to statistical scaling. This would produce an image with enhanced sharpness, contrast and less noise represented in the choice of fitness criterion.