Sparsity and Heterogeneous Dropout for Continual Learning in the Null Space of Neural Activations

Open in new window