Abstract: In this work, we explore the benefits of Variance-Covariance Regularization in Continual Learning (CL). Neural networks suffer from abrupt performance loss when updated with additional data.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results