A Reward Response Game in the Federated Learning System

Decentral and Incentivized Federated Learning Frameworks: A Systematic Literature Review

“…1) Aggregation The aggregation of model-parameters, can be performed by a smart contract on top of blockchain [31], [40], [48], [52], [58], [59]. Since blockchain is assumed to be failure resistant, this strengthens the robustness against possible singlepoint of failure of an aggregation server.…”

“…Even though expansive, since all blockchain nodes store the same information in a redundant fashion, it might make sense to capitalize on the immutability and transparency feature of blockchain and store information where either a shared access among participants is required or where verifiability of the history is required to hold agents accountable for posterior reward calculations [36]. In particular, machine learning models [31], [37], [40], [44], [46], [47], [52], [53], [54], [58], [59], reputation scores [36], [42], User-information [30], [31], [38], [39], [58], [59] and Votes [30], [58], [59] are stored on-chain of the respective FLF.…”

“…Typical costs in the surveyed papers are computation, electricity (e.g. [40], [44], [45], [58], [59], [63]), data acquisition (e.g. pictures and sensor readings [61], [62], [63]) and privacy leakage due to model updates (e.g.…”

“…These kinds of two-stage optimization game are often formalized as the Stackelberg game. Jiang and Wu [40] and Chai et al [44] propose optimal incentive mechanisms with the Stackelberg game, in which two parties sequentially determine their actions according to other party's action. More specifically, a party called leader moves first and the other party called follower moves accordingly.…”

“…More specifically, a party called leader moves first and the other party called follower moves accordingly. Jian and Wu propose a Stackelberg-gamebased incentive mechanism for FL [40]. An aggregation server (a leader) first provides a task's deadline and rewards to workers (followers).…”