A review of sequential three-way decision and multi-granularity learning

The concept of three-way decision, interpreted and described as thinking, problem solving, and information processing in threes, has been widely studied and applied in machine learning and data engineering in recent years. In open-world environment, the connection and interaction of dynamic and uncertainty by multi-granularity learning gives more vitality to three-way decision. In this paper, we investigate and summarize the initial and development models of three-way decision. Then we revisit the historical line of sequential three-way decision from rough set to granular computing. Besides, we focus on exploring a unified framework of three-way multi-granularity learning with four crucial problems on mining uncertain region continually. Finally, we give some proposals on three-way decision associated with open-continual learning.(c) 2022 Elsevier Inc. All rights reserved.

Automated summary

The concept of three-way decision, focusing on thinking, problem solving, and information processing in threes, has been extensively studied and applied in the fields of machine learning and data engineering in recent years. The integration of dynamic and uncertainty through multi-granularity learning in an open-world environment has brought new vitality to three-way decision. This paper investigates and summarizes the initial and development models of three-way decision, traces the historical line of sequential three-way decision from rough set to granular computing, and proposes a unified framework of three-way multi-granularity learning with four crucial problems on mining uncertain regions continuously. Additionally, proposals on three-way decision associated with open-continual learning are provided.