Inference-as-a-service: A situation inference service for context-aware computing

摘要

Context-aware computing is to provide situation-specific services, the situation is inferred from the available contexts, and the contexts are acquired from various sources such as sensors, environments, and SNS contents. However, building inference systems is costly due to the overhead of maintaining contexts repositories, transform contexts into machine-readable knowledge models, running suitable inference algorithms and learning from the inference results for further applications of inference tasks. In this paper, we present a cloud form of a situation inference platform, INFerence-as-a-Service (INFaaS), which provides a core set of common functionality required to develop various context-aware apps. A key challenge in developing such platforms is to manage variability on inference algorithms and inference requirements by applications. INFaaS provides novel approaches to handle the variability by constructing generic models among various inference methods, providing flexible adaptation mechanisms, and defining generic interfaces for invoking various algorithms. The key benefit of INFaaS is a significant reduction of efforts to develop contexts repositories and context-aware applications.