Transformers learn through gradual rank increase

06/12/2023
by   Enric Boix-Adserà, et al.
0

We identify incremental learning dynamics in transformers, where the difference between trained and initial weights progressively increases in rank. We rigorously prove this occurs under the simplifying assumptions of diagonal weight matrices and small initialization. Our experiments support the theory and also show that phenomenon can occur in practice without the simplifying assumptions.

READ FULL TEXT

Please sign up or login with your details

Forgot password? Click here to reset