Cosmic sizing of machine learning image classifier software using neural networks

Abstract:

Development of machine learning software has now penetrated a large diversity of domains, in both academia and industry. From the initial realm of research with a focus on innovation and creativity, its scaling up in industry requires improved planning, monitoring and control of the development and implementation process. Such industry planning and monitoring is difficult without relevant measurement techniques adapted to the problem at hand. This paper illustrates how generic software functions can be extracted from machine learning (ML) system requirements and their functional size measured in COSMIC function points - ISO 19761. An application of these concepts is presented using an example of an ML image classifier software with a feedforward neural network.

Año de publicación:

2019

Keywords:

• ISO 19761
• COSMIC
• Machine learning
• Neural networks
• Function Points

Fuente:

scopus