INFLUENCE OF THE RECEPTIVE FIELD ON THE SPEED AND ACCURACY OF RECOGNITION OF THE CONVOLUTIONAL NEURAL NETWORK

Authors

  • Orlenko Sergey National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", Ukraine
  • Timoshin Yuri National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", Ukraine

Abstract

The paper deals with the receptive field problems and methods of its definition with the help of changing topology of convolutional neural network. With examples shown ways how to achieve the same receptive field in different ways, also performed a number of studies and developed programs for datasets, MNIST and Cifar, which reflects the achieved recognition accuracy and time spent on training. 

Key words: convolutional neural network, receptive field, accuracy, time of study.

Fig.: 6. Bibl.: 3.

Published

2023-11-08

Issue

2018

Section

Machine learning, Big Data