Supplementary Materials for " Generative vs Discriminative: Rethinking The Meta-Continual Learning "

In the above formulations, we assumed a uniform distribution on class labels for simplicity in notation. It is easy to generalize to the case of unbalanced class distributions.