Search

In practical multi-attribute group decision-making (MAGDM) problems, it is common to utilize heterogeneous representation forms to express distinct preference information for different experts, primarily due to their diverse backgrounds. To address the problem of reducing information loss from the aggregation of experts' opinions in heterogeneous MAGDM as well as improving the interpretability of the decision-making process, this paper introduces a concept lattice-based heterogeneous MAGDM approach. The heterogeneous multi-expert formal context is first proposed to capture the heterogeneous evaluation information of alternatives provided by different experts. Then, extended multi-expert concept lattices are constructed to aggregate evaluation information of alternatives by different experts. In this case, all concepts are considered to minimize information loss during the aggregation process. Second, to obtain more reasonable decision results, the distance between the concept intents and the heterogeneous positive and negative ideal solutions is considered, and the alternatives are ranked based on this measure. In addition, the MAGDM process for each alternative is visualized using the extended multi-expert concept lattices. This representation aids in identifying key concepts, their interdependencies, and the overall impact on the decision result. Finally, numerical examples and comparative analysis validate the validity and rationality of the proposed approach in heterogeneous MAGDM. [ABSTRACT FROM AUTHOR]

With the capability of modeling lighter, MLP-based models like the pNLP-Mixer and the HyperMixer demonstrate the potential for diverse tasks in NLP. However, these linguistic models are not optimized for the regularity of textual hierarchical abstraction. Here, this paper proposes the hidden bias attention (HBA), a novel attention mechanism that is lighter than self-attention and focuses on extracting hidden (topic) semanteme. Additionally, this paper introduces a series of lightweight deep learning architectures, HBA-Mixer based on HBA and MHBA-Mixers based on multi-head HBA, which both outperforms pNLP-Mixer and HyperMixer in accuracy with fewer parameters on 3 tasks, including text classification, natural language inference, and sentiment analysis. Compared with large pre-trained models, MHBA-Mixers achieve over 90% of their accuracy with one-thousandth of the parameters. [ABSTRACT FROM AUTHOR]