Mobile Edge Computing (MEC) combinedly with Federated Learning is con- sidered as most capable solutions to AI-driven services. Most of the studies focus on Federated Learning on security aspects and performance, but the re- search is lacking to establish an incentive mechanism for the devices that are connected with a server to perform different task. In MEC, edge nodes would not participate voluntarily in learning process, nodes differ in the accusation of multi-dimensional resources, which also affects the performance of federated learning. In a competitive market scenario, the auction game theory has been widely popular for designing efficient resource allocation mechanisms, as it particularly focuses on regulating the strategic interactions among the self-interested play- ers.In this thesis, I investigate auction-based approach that based on incentive mechanism and encourage nodes to share their resources and take part in train- ing process as well as to maximize the auction revenue. To achieve this research goal, I developed auction mechanism considering the network dynamics and neglecting the devices computation and design a novel generalized first price auction mechanism to encourage participation of connected devices. Furthermore, I studied the K top best-response bidding strategies that maximize the profits of the resource sellers and guarantee the stability and effectiveness of the auction by satisfying desired economic properties. To this end, I validate the performance of the proposed auction mechanisms and bidding strategies through numerical result analysis.

Mobile Edge Computing (MEC) combinedly with Federated Learning is con- sidered as most capable solutions to AI-driven services. Most of the studies focus on Federated Learning on security aspects and performance, but the re- search is lacking to establish an incentive mechanism for the devices that are connected with a server to perform different task. In MEC, edge nodes would not participate voluntarily in learning process, nodes differ in the accusation of multi-dimensional resources, which also affects the performance of federated learning. In a competitive market scenario, the auction game theory has been widely popular for designing efficient resource allocation mechanisms, as it particularly focuses on regulating the strategic interactions among the self-interested play- ers.In this thesis, I investigate auction-based approach that based on incentive mechanism and encourage nodes to share their resources and take part in train- ing process as well as to maximize the auction revenue. To achieve this research goal, I developed auction mechanism considering the network dynamics and neglecting the devices computation and design a novel generalized first price auction mechanism to encourage participation of connected devices. Furthermore, I studied the K top best-response bidding strategies that maximize the profits of the resource sellers and guarantee the stability and effectiveness of the auction by satisfying desired economic properties. To this end, I validate the performance of the proposed auction mechanisms and bidding strategies through numerical result analysis.

Resource Allocation through Auction-based Incentive Scheme for Federated Learning in Mobile Edge Computing

ASIF, JAWAD
2022/2023

Abstract

Mobile Edge Computing (MEC) combinedly with Federated Learning is con- sidered as most capable solutions to AI-driven services. Most of the studies focus on Federated Learning on security aspects and performance, but the re- search is lacking to establish an incentive mechanism for the devices that are connected with a server to perform different task. In MEC, edge nodes would not participate voluntarily in learning process, nodes differ in the accusation of multi-dimensional resources, which also affects the performance of federated learning. In a competitive market scenario, the auction game theory has been widely popular for designing efficient resource allocation mechanisms, as it particularly focuses on regulating the strategic interactions among the self-interested play- ers.In this thesis, I investigate auction-based approach that based on incentive mechanism and encourage nodes to share their resources and take part in train- ing process as well as to maximize the auction revenue. To achieve this research goal, I developed auction mechanism considering the network dynamics and neglecting the devices computation and design a novel generalized first price auction mechanism to encourage participation of connected devices. Furthermore, I studied the K top best-response bidding strategies that maximize the profits of the resource sellers and guarantee the stability and effectiveness of the auction by satisfying desired economic properties. To this end, I validate the performance of the proposed auction mechanisms and bidding strategies through numerical result analysis.
2022
Resource Allocation through Auction-based Incentive Scheme for Federated Learning in Mobile Edge Computing
Mobile Edge Computing (MEC) combinedly with Federated Learning is con- sidered as most capable solutions to AI-driven services. Most of the studies focus on Federated Learning on security aspects and performance, but the re- search is lacking to establish an incentive mechanism for the devices that are connected with a server to perform different task. In MEC, edge nodes would not participate voluntarily in learning process, nodes differ in the accusation of multi-dimensional resources, which also affects the performance of federated learning. In a competitive market scenario, the auction game theory has been widely popular for designing efficient resource allocation mechanisms, as it particularly focuses on regulating the strategic interactions among the self-interested play- ers.In this thesis, I investigate auction-based approach that based on incentive mechanism and encourage nodes to share their resources and take part in train- ing process as well as to maximize the auction revenue. To achieve this research goal, I developed auction mechanism considering the network dynamics and neglecting the devices computation and design a novel generalized first price auction mechanism to encourage participation of connected devices. Furthermore, I studied the K top best-response bidding strategies that maximize the profits of the resource sellers and guarantee the stability and effectiveness of the auction by satisfying desired economic properties. To this end, I validate the performance of the proposed auction mechanisms and bidding strategies through numerical result analysis.
Federated Learning
Incentive Mechanism
Auction
MEC
File in questo prodotto:
File Dimensione Formato  
Asif_Jawad_2013002.pdf

accesso aperto

Dimensione 3 MB
Formato Adobe PDF
3 MB Adobe PDF Visualizza/Apri

The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.12608/54131