Embedding vs. reconstruction-level contrast (RLC).

<p>Top: the encoder consumes two inputs (the original graph and an augmented view) and contrasts their node embeddings in the embedding space. Bottom: DCOR uses dual autoencoders to reconstruct the adjacency and the attribute matrices for the original graph and the augmented view, then applies...

תיאור מלא

שמור ב:

מידע ביבליוגרפי
מחבר ראשי:	Hossein Rafieizadeh (22676722) (author)
מחברים אחרים:	Hadi Zare (20073000) (author), Mohsen Ghassemi Parsa (22676725) (author), Hocine Cherifi (8177628) (author)
יצא לאור:	2025
נושאים:	Cell Biology Science Policy Environmental Sciences not elsewhere classified Biological Sciences not elsewhere classified Information Systems not elsewhere classified intrusions across social reconstructions across views level contrastive learning dual contrastive learning across six benchmarks div >< p view discrepancies underutilized augmented graph views dcor improves auroc view discrepancies level contrast dual autoencoder augmented view specific information six datasets reduces auroc publicly available preserves fine physical domains performing non maximum gain leaving cross identifying threats financial fraud existing graph dcor reconstructs dcor ), contrasts reconstructions attributed networks attribute patterns
תגים:	הוספת תג אין תגיות, היה/י הראשונ/ה לתייג את הרשומה!

תיאור
סיכום:	<p>Top: the encoder consumes two inputs (the original graph and an augmented view) and contrasts their node embeddings in the embedding space. Bottom: DCOR uses dual autoencoders to reconstruct the adjacency and the attribute matrices for the original graph and the augmented view, then applies contrastive learning directly to the two sets of reconstructions, which preserves cross-view discrepancies that message passing may smooth out.</p>

Embedding vs. reconstruction-level contrast (RLC).

פריטים דומים