A Maximum Average Weight Matching Algorithm for Collective Tasks Allocation in the Collective Computing System

Collective computing is the latest generation of computing frameworks that have received a lot of attention recently. In this framework, different types of remote computing devices and even "human" are connected to the system as independent computing devices, which can be used to perform different types of complex tasks with huge amounts of resources. Therefore, the various computing devices have different computing capabilities that can be fully utilized in different tasks. However, research on collective computing is still relatively limited, and in particular, there is a gap in research on task matching methods for collective computing. This paper proposes a task matching method for collective computing systems. We first propose a new index to describe the evaluation of devices by tasks in collective computing systems. Afterwards, a maximum average weight matching algorithm is proposed for solving the optimal matching problem for weighted bipartite graphs in this scenario. Finally, we integrate the whole process into the collective computing system and perform related experiments with satisfactory results.