mirror of
https://github.com/microsoft/qlib.git
synced 2026-06-06 05:51:17 +08:00
be4646b4b7
* Intermediate version * Fix yaml template & Successfully run rolling * Be compatible with benchmark * Get same results with previous linear model * Black formatting * Update black * Update the placeholder mechanism * Update CI * Update CI * Upgrade Black * Fix CI and simplify code * Fix CI * Move the data processing caching mechanism into utils. * Adjusting DDG-DA * Organize import
GATs
- Graph Attention Networks(GATs) leverage masked self-attentional layers on graph-structured data. The nodes in stacked layers have different weights and they are able to attend over their neighborhoods’ features, without requiring any kind of costly matrix operation (such as inversion) or depending on knowing the graph structure upfront.
- This code used in Qlib is implemented with PyTorch by ourselves.
- Paper: Graph Attention Networks https://arxiv.org/pdf/1710.10903.pdf