We carried out an additional continual learning experiment on eight tasks (as in 1 [33, manuscript]) that consist of vision datasets with different domains: {CIFAR-10 / CIFAR-100 / MNIST / SVHN /