Gated Fusion

How it works

Benefits

Highlighted claims

• The model concatenates three 64-dimensional graph-branch embeddings into a 192-dimensional per-node vector. — Learning Multi-Relational Graph Representations for DNA Methylation-Based Biological Age Estimation
• A gated MLP followed by softmax produces node-wise weights for the three graph branches. — Learning Multi-Relational Graph Representations for DNA Methylation-Based Biological Age Estimation
• The fused node representation is a weighted sum of the three branch embeddings. — Learning Multi-Relational Graph Representations for DNA Methylation-Based Biological Age Estimation
• The gate allows CpG nodes and samples to weight co-methylation, chromosomal, and gene-level information differently. — Learning Multi-Relational Graph Representations for DNA Methylation-Based Biological Age Estimation
• The gated fusion mechanism is interpreted as important because relation importance can vary by CpG node and sample. — Learning Multi-Relational Graph Representations for DNA Methylation-Based Biological Age Estimation