NeuPart: Using Analytical Models to Drive Energy-Efficient Partitioning of CNN Computations on Cloud-Connected Mobile Clients

Manasi, Susmita Dey; Snigdha, Farhana Sharmin; Sapatnekar, Sachin S.

doi:10.1109/tvlsi.2020.2995135

Cited by 12 publications

(4 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Typically, publications in this topic discuss the problem of deploying AI models in a real scenario or in a scenario with peculiar constraints that challenge a standard approach. For example, deployment publications showcase the challenges of deploying energy‐efficient AI in FPGA (Tao et al, 2020), in Edge devices (Gondi & Pratap, 2021; Kim & Wu, 2020), and in mobile devices (Jayakodi et al, 2020; Manasi et al, 2020; Wang et al, 2022), and so on.…”

Section: Resultsmentioning

confidence: 99%

A systematic review of Green AI

Verdecchia

Sallou

Cruz

2023

WIREs Data Min & Knowl

View full text Add to dashboard Cite

With the ever‐growing adoption of artificial intelligence (AI)‐based systems, the carbon footprint of AI is no longer negligible. AI researchers and practitioners are therefore urged to hold themselves accountable for the carbon emissions of the AI models they design and use. This led in recent years to the appearance of researches tackling AI environmental sustainability, a field referred to as Green AI. Despite the rapid growth of interest in the topic, a comprehensive overview of Green AI research is to date still missing. To address this gap, in this article, we present a systematic review of the Green AI literature. From the analysis of 98 primary studies, different patterns emerge. The topic experienced a considerable growth from 2020 onward. Most studies consider monitoring AI model footprint, tuning hyperparameters to improve model sustainability, or benchmarking models. A mix of position papers, observational studies, and solution papers are present. Most papers focus on the training phase, are algorithm‐agnostic or study neural networks, and use image data. Laboratory experiments are the most common research strategy. Reported Green AI energy savings go up to 115%, with savings over 50% being rather common. Industrial parties are involved in Green AI studies, albeit most target academic readers. Green AI tool provisioning is scarce. As a conclusion, the Green AI research field results to have reached a considerable level of maturity. Therefore, from this review emerges that the time is suitable to adopt other Green AI research strategies, and port the numerous promising academic results to industrial practice. This article is categorized under: Technologies > Machine Learning

show abstract

Section: Resultsmentioning

confidence: 99%

A systematic review of Green AI

Verdecchia

Sallou

Cruz

2023

WIREs Data Min & Knowl

View full text Add to dashboard Cite

show abstract

“…where D tot (t) is defined through (7), (9), and (10), while E tot is defined through (8) and (11). Under feasibility assumption of (12), and i.i.d.…”

Section: A Proposed Solutionmentioning

confidence: 99%

“…In [7], the authors propose a strategy to partition the DNN under different network conditions to minimize the overall processing delay. Also, [8] minimizes the energy consumption on the client-side by partitioning Convolutional Neural Network (CNN) computations between the client and the cloud. Mao et al partition the DNN always after the first convolutional layer to minimize the cost of mobile devices and use the differentially private mechanism to preserve the privacy [9].…”

Section: Introductionmentioning

confidence: 99%

Energy-Efficient Cooperative Inference Via Adaptive Deep Neural Network Splitting at the Edge

Labriji,

Merluzzi,

Airod

et al. 2023

ICC 2023 - IEEE International Conference on Communications

View full text Add to dashboard Cite

Learning and inference at the edge is all about distilling, exchanging, and processing data in a cooperative and distributed way, to achieve challenging trade-offs involving energy, delay, and accuracy. This calls for a joint orchestration of radio and computing resources. We propose an online adaptive resource allocation algorithm to choose where to compute, and how to offload computations, exploiting the concept of Deep Neural Network (DNN) splitting. The latter allows a device to locally execute part of an inference related processing, and delegate the other portion to a nearby Mobile Edge Host (MEH), which receives intermediate results from the device via a time varying wireless communication channel. Our method deals with dynamic parameters involving wireless channels, data arrivals, and MEH's CPU availability, by taking online control actions including the best splitting point, and the uplink data rate to transfer raw data or intermediate results (e.g., extracted features). The decision is taken only based on instantaneous observations of context parameters, to minimize the long-term device energy consumption, while guaranteeing the end-to-end delay not to exceed a predefined threshold, on average and probabilistic sense. Besides a theoretical analysis, numerical simulations show the effectiveness of our adaptive method in selecting the best partial offloading decision (DNN splitting) under different network conditions. Differently from previous works on edge inference, we exploit recently developed empirical models for the energy consumption of NVIDIA ® edge boards, to evaluate the performance of DNN splitting at the edge, when exploring the typical offloading trade-off between energy and delay, both entailing communication and computing.

show abstract

“…Considered a major drawback, the "high computational complexity" with respect to compute explanations for a single prediction demands "minutes," even for the state-of-the-art model of LIME, GoogleNet (Samek & Müller, 2019). Although efforts are in place attempting to reduce energy consumption (Manasi et al, 2020), these types of models are still characterized by "high energy consumption" (Yang et al, 2017).…”

Section: As Explanation Messenger and Elucidatormentioning

confidence: 99%

Information Resilient Society in anAIWorld—IsXAISufficient?

Xie

Gao

Han

2022

Proceedings of the Association for Information Science and Tech

View full text Add to dashboard Cite

This paper discusses the role of organizational records management (ORM) with respect to the field of explainable artificial intelligence (XAI) and argues about its necessity and significance. The current trend is to utilize AI to explain AI, including both explanation production and provision. We argue that this kind of approach is not sufficient by itself as it lacks neutrality and localness to explanation recipients, thus ineffective in establishing public trust. We propose the addition of the ORM profession, as an informational 3rd party, to the current and future XAI. We envision that, by working together with XAI performers, ORM contributes to an information resilient AI society. To take on the related responsibilities, the ORM profession must improve its professional expertise and strengthen its professional independence. We call for consensus among individual information professionals and the advocacy of information professional networks to make this vision a reality.

show abstract

NeuPart: Using Analytical Models to Drive Energy-Efficient Partitioning of CNN Computations on Cloud-Connected Mobile Clients

Cited by 12 publications

References 27 publications

A systematic review of Green AI

A systematic review of Green AI

Energy-Efficient Cooperative Inference Via Adaptive Deep Neural Network Splitting at the Edge

Information Resilient Society in anAIWorld—IsXAISufficient?

Contact Info

Product

Resources

About