Tag: representation learning

Researchers have developed BHyGNN+, a self-supervised framework that learns representations of heterophilic hypergraphs without labels by contrasting each hypergraph against ...

A new adversarial training method called MI-MKGC dynamically balances text, image, and structural information to predict missing facts in multimodal ...

A large-scale study finds that pretraining tabular self-supervised models on carefully selected subsets of 100 to 1000 unlabeled examples often ...

A new graph convolution Transformer framework with self-distillation improves multi-view graph clustering by suppressing cross-view noise and exploiting complementary information ...

A new dual-branch deep learning framework called CrossBranch improves the estimation of cell-type composition from bulk, proteomic, and spatial omics ...

Researchers have unveiled a subspace clustering framework that preserves the structural fidelity of high-dimensional data while avoiding the prohibitive cost ...

A new survey in Artificial Intelligence Review unifies probabilistic, clustering and deep representation-learning approaches for inferring hidden system states from ...

Researchers have developed two frameworks that fuse network embeddings with spatial homophily to estimate social influence in location-based social networks ...

Researchers have developed CurvBi, a framework that unites four-dimensional bicomplex algebra with Fisher-Rao-based Riemannian fusion to improve multimodal knowledge graph ...

Researchers have developed a unified graph-based recommendation method that sequences online learning resources according to conceptual dependencies and learner behavior.