.png){kind=icon}
.png){kind=icon}
Return
CM-CCL: Collaborative multi-scale concept-cognitive learning for knowledge discovery
.png){kind=icon}
PRE
AI
DOI：10.1016/j.patcog.2025.112268.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Concept-cognitive learning (CCL) provides an effective method for representing knowledge in data, with the use of concepts as knowledge carriers being its most significant characteristic. However, existing CCL models neglect the utilization of multi-scale information, resulting in insufficient representation capabilities of the learned concepts. Therefore, this paper proposes a novel multi-scale concept-cognitive learning model to address this issue. Firstly, a rational multi-scale data construction method is provided based on the characteristics of CCL. Then, a multi-scale feature selection method is introduced, which considers both the inter-scale correlations and intra-scale class distances. On this basis, progressive concepts are learned by integrating similar granular concepts at each scale to explicitly represent the knowledge in the data. Furthermore, a mechanism for the collaboration among progressive concepts at different scales is proposed to complete the classification task. Finally, a series of experiments are conducted to validate the effectiveness of the proposed CM-CCL model.
Keywords:
concept-cognitive learning
multi-scale information
feature selection
progressive concepts
knowledge representation
Journal
IF:
7.6
Papers:
1.3W
Citations:
4.5W