Reusability in Artificial Neural Networks: An Empirical Study

Abstract

Machine learning, especially deep learning has aroused interests of researchers and practitioners for the last few years in development of intelligent systems such as speech, natural language, and image processing. Software solutions based on machine learning techniques attract more attention as alternatives to conventional software systems. In this paper, we investigate how reusability techniques are applied in implementation of artificial neural networks (ANNs). We conducted an empirical study with an online survey among experts with experience in developing solutions with ANNs. We analyze the feedback of more than 100 experts to our survey. The results show existing challenges and some of the applied solutions in an intersection between reusability and ANNs.