Skip to content

🧬 Transformer-based model for predicting reactivity profiles of RNA molecules.

Notifications You must be signed in to change notification settings

maudl3116/2023-Kaggle-Stanford-Ribonanza-RNA-Folding-uceemjl

Repository files navigation

2023-Kaggle-Stanford-Ribonanza-RNA-Folding-uceemjl

Builds upon the "best single model" shared by kaggler iafoss. Substantial performance improvement (public leaderboard position) was obtained by:

  • extracting base pairing probability matrices (bppm) predicted with contrafold2 to bias the attention.
  • replacing the sinusoidal positional encoding with a relative positional encoding.

About

🧬 Transformer-based model for predicting reactivity profiles of RNA molecules.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published