DeepNetQoE: Self-adaptive QoE Optimization Framework of Deep Networks

Published

June 24, 2021

Author(s)

Hamid Gharavi

Abstract

Future advances in deep learning and its impact on the development of artificial intelligence (AI) in all fields depends heavily on data size and computational power. Sacrificing massive computing resources in exchange for better precision rates of the network model is recognized by many researchers as the preferred approach. This leads to huge computing consumption and satisfactory results are not always expected when computing resources are limited. Therefore, it is necessary to make a balance between resources and the performance of the model to get satisfactory training results. This article proposes a self adaptive quality of experience (QoE) framework, DeepNetQoE, to guide the training of deep networks. A self-adaptive QoE model is a set up that relates to the accuracy of the model and the computing resources required for training to improve the experience value of the model during training and application. To maximize experience value when computer resources are limited, a resource allocation model and solutions need to be established. In addition, as a crowd counting example, we carry out experiments based on four network models to analyze the experience value, as well as joint optimization of resources and performance. Experimental results show that the proposed DeepNetQoE is capable of adaptively obtaining a high experience value according to user needs and guide users to determine the computational resources allocated to the network models.

Citation

IEEE Network

Volume

35

Issue

3

Pub Type

Journals

Download Paper

https://doi.org/10.1109/MNET.011.2000475

Local Download

Keywords

Deep Networks, QoE, Deep Learning, Artificial Intelligence.

Citation

Gharavi, H. (2021), DeepNetQoE: Self-adaptive QoE Optimization Framework of Deep Networks, IEEE Network, [online], https://doi.org/10.1109/MNET.011.2000475, https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=930707 (Accessed October 31, 2024)