Abstract
XXI century is the age of global automation and digitization.There is high demand for optical recognition software, including character recognition.There are different approaches in solution optical recognition problem.Some of them based on classical feature extraction methods.Other based on machine learning algorithms.In this work, we observed related works in machine learning field and propose the plan for further research.The work relies on two research studies that describe basics and fundamentals of machine learning.These research include various experiments in this field.We conducted several experiments to get acquainted with methods and techniques and to identify key features that are affecting on optical character recognition process.We analyzed two main architectural structures: multilayer perceptron and convolutional neural network.In conclusion, we identified key points of machine learning techniques and composed our own strategy for further researches.The result shows the difference in performance of different convolutional neural network models under the same conditions.Further work will cover researches and experiments on performance of several architectures.In addition, we observed latest tools, software programs and environments for the most convenient way to organize implementation process.