Adaptive Composition of Machine Learning as a Service (MLaaS) for IoT Environments

TL;DR

This paper introduces an adaptive framework for composing MLaaS in IoT environments that dynamically adjusts to changing data and system requirements, maintaining QoS efficiently.

Contribution

It presents a novel adaptive composition mechanism using multi-armed bandit optimization to handle IoT variability and reduce recomposition costs.

Findings

Effective in maintaining QoS in dynamic IoT settings

Reduces computational costs compared to static recomposition

Demonstrates improved performance on real-world datasets

Abstract

The dynamic nature of Internet of Things (IoT) environments challenges the long-term effectiveness of Machine Learning as a Service (MLaaS) compositions. The uncertainty and variability of IoT environments lead to fluctuations in data distribution, e.g., concept drift and data heterogeneity, and evolving system requirements, e.g., scalability demands and resource limitations. This paper proposes an adaptive MLaaS composition framework to ensure a seamless, efficient, and scalable MLaaS composition. The framework integrates a service assessment model to identify underperforming MLaaS services and a candidate selection model to filter optimal replacements. An adaptive composition mechanism is developed that incrementally updates MLaaS compositions using a contextual multi-armed bandit optimization strategy. By continuously adapting to evolving IoT constraints, the approach maintains Quality of Service (QoS) while reducing the computational cost associated with recomposition from scratch. Experimental results on a real-world dataset demonstrate the efficiency of our proposed approach.