Daily Arxiv

This page organizes papers related to artificial intelligence published around the world.
This page is summarized using Google Gemini and is operated on a non-profit basis.
The copyright of the paper belongs to the author and the relevant institution. When sharing, simply cite the source.

Disentangled and Self-Explainable Node Representation Learning

Created by
  • Haebom

Author

Simone Piaggesi, Andre Panisson, Megha Khosla

Outline

This paper introduces the Disentangled and Self-Explainable Node Embedding (DiSeNE) framework, which generates node embeddings using unsupervised learning. DiSeNE uses separable representation learning to generate dimensionally interpretable embeddings, where each dimension corresponds to a distinct topological structure of the graph. This paper presents a novel objective function for separable and interpretable embeddings and new metrics for evaluating representation quality and human interpretability. Experiments on several benchmark datasets demonstrate the effectiveness of the proposed method.

Takeaways, Limitations

•
Takeaways:
◦
We present a novel framework for improving the interpretability of node embeddings in unsupervised learning environments.
◦
Securing dimension-specific interpretability through separable representation learning.
◦
Development of a new objective function that simultaneously optimizes interpretability and separability.
◦
Proposing a new metric to evaluate embedding quality and interpretability.
◦
Demonstrated excellent performance on various benchmark datasets.
•
Limitations:
◦
Lack of information about specific implementation details and algorithmic complexity.
◦
Additional information on real-world applications and performance comparisons is needed.
◦
Lack of comparative analysis with other separable representation learning techniques.
👍
Made with Slashpage