NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:6333
Title:Gradient based sample selection for online continual learning

There is an interesting contribution in terms of formulating the continual learning setting into a geometry-based selection problem of samples from the experience replay. Reviewers would like to see more experiments on other standard benchmarks, as well as additional discussion about how each approximation step works and the relationship to other existing methods.