Towards Modular Machine Learning Solution Development: Benefits and Trade-offs
This addresses the problem of costly ML adoption for businesses, but it is incremental as it builds on existing modularity concepts.
The paper tackles the high cost and inefficiency of developing custom machine learning solutions due to monolithic architectures, and finds that modular approaches offer promising potential for better performance and data advantages in text and image tasks.
Machine learning technologies have demonstrated immense capabilities in various domains. They play a key role in the success of modern businesses. However, adoption of machine learning technologies has a lot of untouched potential. Cost of developing custom machine learning solutions that solve unique business problems is a major inhibitor to far-reaching adoption of machine learning technologies. We recognize that the monolithic nature prevalent in today's machine learning applications stands in the way of efficient and cost effective customized machine learning solution development. In this work we explore the benefits of modular machine learning solutions and discuss how modular machine learning solutions can overcome some of the major solution engineering limitations of monolithic machine learning solutions. We analyze the trade-offs between modular and monolithic machine learning solutions through three deep learning problems; one text based and the two image based. Our experimental results show that modular machine learning solutions have a promising potential to reap the solution engineering advantages of modularity while gaining performance and data advantages in a way the monolithic machine learning solutions do not permit.