Test-Time Adaptive Composition for Machine Learning as a Service (MLaaS) in IoT Environments

2026-06-04

AI Summary

This work addresses the performance degradation of Machine Learning as a Service (MLaaS) compositions in dynamic Internet of Things (IoT) environments, where conventional adaptive approaches relying on service replacement or reconfiguration prove inefficient. For the first time, test-time adaptation (TTA) is introduced into MLaaS composition through a dynamic adjustment framework operating at inference time. The proposed method employs a TTA-aware composability model to evaluate service compatibility and integrates service-level adaptation mechanisms to enable rapid online optimization. Crucially, it achieves performance stabilization without replacing components or restructuring the system, thereby preserving architectural integrity while substantially reducing computational overhead. Experimental results demonstrate that the approach outperforms existing methods in maintaining the stability of composite service performance under dynamic conditions.

Abstract

The dynamic nature of Internet of Things (IoT) environments affects the long-term effectiveness of Machine Learning as a Service (MLaaS) compositions. Existing adaptive composition methods are mainly based on service replacement or re-composition, where identifying suitable substitutes is difficult and time-consuming. To address this, we propose a novel Test-Time Adaptive (TTA) composition framework for MLaaS in IoT environments. First, we introduce a TTA-aware composability model to determine whether adapted services remain compatible with the existing composition. Next, we design a service-level adaptation model to adjust individual services during inference while preserving composition performance. Experimental results demonstrate that the proposed framework reduces computational time more effectively than traditional adaptive approaches.

Problem

• Machine Learning as a Service
• IoT environments
• adaptive composition
• service replacement
• composability

Innovation

• Test-Time Adaptation
• MLaaS Composition
• IoT Environments
• Service-level Adaptation
• Composability Model