Network Together: Node Classification via Cross-Network Deep Network Embedding.

Area of Science:

• Machine Learning
• Network Science
• Data Mining

Background:

• Network embedding learns node representations but struggles with cross-network generalization.
• Existing methods fail to leverage information across different networks for node classification.

Purpose of the Study:

• To address the challenge of cross-network node classification by learning transferable features.
• To develop a novel model that incorporates domain adaptation into network embedding.

Main Methods:

• Proposed Cross-Network Deep Network Embedding (CDNE) model.
• Incorporated domain adaptation to learn network-invariant node representations.
• Leveraged network structures, node attributes, and labels for cross-network alignment.

Main Results:

• CDNE learns label-discriminative and network-invariant node vector representations.
• Demonstrated significant performance improvement over state-of-the-art methods in cross-network node classification tasks.
• Effectively transfers knowledge from source to target networks.

Conclusions:

• CDNE offers a robust solution for cross-network node classification.
• The proposed domain adaptation approach enhances feature transferability across networks.
• CDNE advances the field of network embedding for heterogeneous network analysis.