基于联邦学习的分布式农业组织

Distributed agricultural organization based on federated learning

我国现阶段仍以小规模农业为主,如何在小农、小地块的农业生产经营方式下发展适宜的智慧农业颇具挑战。基于此,提出一种结合联邦学习与区块链技术的分布式农业人工智能框架,使得在数据不共享的情况下,能达到训练模型的目的,并可建立参与方的激励机制。该框架有助于充分利用农业多源异构数据,减少对用户数据量的要求,发展因地制宜的决策模型,促进小规模农业的产销衔接。

At present,small-scale agriculture is dominating in China.How to develop appropriate smart agriculture for agricultural management with small farmers and small plots is quite challenging.A distributed agricultural AI framework combining federal learning and blockchain technology was proposed,which can achieve the purpose of the training model and establish the incentive mechanism for participants without data sharing.This framework helped make full use of agricultural multi-source heterogeneous data,reducing user data requirements,developing decision-making models according to local conditions,and promoting the connection of production and marketing of small-scale agriculture.