NFT @ 06032026: On the Alignment of Self-Supervised Graph Representation Learning Methods скачать в хорошем качестве

NFT @ 06032026: On the Alignment of Self-Supervised Graph Representation Learning Methods

#researchatnecstlab - #NECSTFridayTalk On Friday, March 6, 2026, we had a new talk for the series #NECSTFridayTalk. During this talk, we had, as speaker, Leonardo De Grandis, PhD at Dipartimento di Elettronica, Informazione e Bioingegneria. In the following, you can find the details about the talk: 📌 Title: On the Alignment of Self-Supervised Graph Representation Learning Methods 📌 Abstract: Graph Neural Networks are well suited for modeling heterogeneous knowledge from many domains, such as social networks, biology, and chemistry. However, the necessity of labels is often a challenge for their training, therefore self-supervised techniques represent a promising approach to learn powerful, generalizable, and transferable representations. This talk presents Graph-DINO, an adaptation of the well known DINO framework from computer vision, for self-supervised representation learning on graph-based data. The learned embeddings are also evaluated from an alignment perspective, leveraging complementary local and global alignment metrics, to examine its behavior with respect to other unsupervised methods. #NECSTLab #Computerscience