CAREER: Adaptive Deep Learning Systems Towards Edge Intelligence

边缘智能将使用深度神经网络(DNN)的智能数据处理推向网络的边缘，更接近数据源。它使各种领域的应用成为可能，并引起了工业界和学术界的极大关注。然而，边缘平台上有限的资源，如边缘服务器和物联网设备，阻碍了对深度学习预测任务的查询提供快速准确响应的能力。因此，只有一些适合边缘部署的深度学习任务和较小的DNN模型是可行的。为了克服这一限制，本项目探索了一种新的自适应方法来构建深度学习系统。系统将根据三个关键维度--可变任务复杂性、波动的推理工作负载和多租户边缘环境中的资源竞争--产生的不同资源需求，对为预测任务执行的DNN进行实时调整。其目标是同时优化系统效率和精度。实现设想的适应性将促进深度学习技术在不同应用程序和环境中的有效部署。这项研究有可能为开发以前受到资源限制的新型边缘应用开辟新的可能性。它将使更广泛的深度学习任务能够在边缘平台上执行，以及更强大的DNN模型，这是充分释放边缘智能潜力的关键能力。该项目的实际影响将在各种应用程序中得到展示，特别是通过与马萨诸塞州老龄化和阿尔茨海默病互联护理技术中心的合作来加强老年人护理的应用程序。此外，该项目旨在培养一批在计算机系统和机器学习方面具有跨学科专业知识的熟练工程师和计算机科学家。将努力通过多样性计划和面向K-12学生的外展计划来招收代表性不足的学生。这一奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Edge intelligence pushes intelligent data processing using deep neural networks (DNNs) to the edge of the network, closer to data sources. It enables applications across various fields and has garnered significant attention from both industry and academia. However, the limited resources on edge platforms, such as edge servers and Internet of Things devices, hinder the ability to deliver fast and accurate responses to queries from deep learning prediction tasks. As a result, only some deep learning tasks and smaller DNN models suitable for edge deployment are feasible. To overcome this limitation, this project explores a new adaptive approach in building deep learning systems. The systems will make real-time adjustments to the DNNs executed for prediction tasks based on the varying resource demands arising from three critical dimensions -- variable task complexity, fluctuating inference workloads, and resource contention in multi-tenant edge environments. The goal is to optimize both system efficiency and accuracy. Realizing the envisioned adaptiveness will facilitate the effective deployment of deep learning techniques across diverse applications and environments. This research has the potential to open new possibilities for the development of novel edge applications that were previously limited by resource constraints. It will enable a broader range of deep learning tasks to be executed on edge platforms, along with more powerful DNN models, a capability critical in fully unleashing the potential of edge intelligence. The practical impact of the project will be demonstrated in a variety of applications, and in particular, applications that enhance elder care through collaboration with the Massachusetts AI and Technology Center for Connected Care in Aging and Alzheimer’s Disease. Moreover, this project aims to cultivate a pipeline of skilled engineers and computer scientists with interdisciplinary expertise in computer systems and machine learning. Efforts will be made to recruit underrepresented students through diversity programs and outreach initiatives to K-12 students.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.