SCIENCE@home

Enriching Pre-Training Using Fuzzy Logic

Vansh Gupta,Vandana Bharti,Abhinav Kumar,Anshul Sharma,Sanjay Kumar Singh

Published 2025 in IEEE International Conference on Fuzzy Systems

ABSTRACT

Graph representation learning advances graph machine learning by encoding structural and relational information into feature vectors. This study introduces a fuzzy logic-based pre-processing layer that enhances node representations by adding semantic diversity and contextual understanding. The layer models uncertainty and captures abstract semantic characteristics in graph data, addressing the limitations of conventional methods that depend solely on structural attributes. By applying defuzzification, the layer refines embeddings, improving their robustness and effectiveness for a wide range of downstream tasks. To test its robustness, we introduce Gaussian noise ranging from 2% to 10% into the datasets, simulating real-world data imperfections. We evaluate the proposed layer on GraphMamba architecture, using DeepWalk and Node2Vec as baseline node feature generation algorithms. The results show consistent improvements in accuracy, F1 scores, precision, and recall across different noise levels. Our findings demonstrate the layer’s ability to preserve high representational quality, speed up convergence, and handle noisy representations effectively.

PUBLICATION RECORD

CITATION MAP

EXTRACTION MAP

CLAIMS

  • No claims are published for this paper.

CONCEPTS

  • No concepts are published for this paper.

REFERENCES

Showing 1-23 of 23 references · Page 1 of 1

CITED BY

  • No citing papers are available for this paper.

Showing 0-0 of 0 citing papers · Page 1 of 1