← All work
published2023 – 2024

MGDPR

Multi-relational graph diffusion with parallel retention for stock-trend classification.

  • Multi-relational graphs
  • Retention
  • Financial ML

Problem

Financial graphs contain multiple relationship types whose relevance changes across stocks and time.

Approach

Combine relation-aware graph diffusion with parallel retention to preserve complementary relational signals.

Role

First author and implementation author

MGDPR models several stock-relation graphs in parallel and combines their representations through retention rather than collapsing every relation into one adjacency structure.