ROBUST COLLABORATIVE LEARNING FOR SEQUENCE MODELLING
Francois Buet-Golfouse, Hans Roggeman, Islam Utyagulov
-
Members: FreeSPS
IEEE Members: $11.00
Non-members: $15.00Length: 00:10:33
09 May 2022
Current deep learning techniques for RNA classification suffer from over-fitting and reproducibility. We show that, by introducing robustness by design in both CNN and RNN (with attention mechanism) algorithms, we are able to achieve standalone state-of-the-art accuracy on the most widely used dataset. By constructing model-agnostic robustness checks and reusing features obtained from both architectures, we build a collaborative framework that improves performance and stability.