SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2104.07446 · ICCV · 2021

Rehearsal revealed: The limits and merits of revisiting samples in continual learning

K Leuven, Matthias De Lange, Eli Verwimp

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 10 functions out of this paper's own repositories and ran 6 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
Mattdl/RehearsalRevealed canonical 5 of 9
copy not recorded — 1 of 1
FunctionStatusWhere it lives
ResNet18 Ran Mattdl/RehearsalRevealed/ridge_aversion_exp/model.py
code served (permissive licence) · get_code("b03f148b0213ab66")
calc_full_grad_norm Ran Mattdl/RehearsalRevealed/grad_norm_exp/grad_norms.py
code served (permissive licence) · get_code("5f3b1c7207af014b")
conv3x3 Ran Mattdl/RehearsalRevealed/ridge_aversion_exp/model.py
code served (permissive licence) · get_code("583f9780bdd00a45")
get_split_cifar10 Ran Mattdl/RehearsalRevealed/framework/cole/core.py
code served (permissive licence) · get_code("55bf3bcd2870630b")
pil_loader Ran this paper's copy was not recorded; identical code first harvested from neuroethology/bkind
pointer only · get_code("f321f54723433661")
rescale_model_coords Ran Mattdl/RehearsalRevealed/contour_exp/mode_connectivity_plot.py
code served (permissive licence) · get_code("3df54e3aaa048117")
get_single_label_mnist Not yet run Mattdl/RehearsalRevealed/framework/cole/core.py
code served (permissive licence) · get_code("4ac54a7e815a9d58")
get_split_mini_imagenet Not yet run Mattdl/RehearsalRevealed/ridge_aversion_exp/miniimagenet.py
code served (permissive licence) · get_code("31b613cae5ceb8b8")
get_split_mnist Not yet run Mattdl/RehearsalRevealed/framework/cole/core.py
code served (permissive licence) · get_code("8e18f937db588ef5")
merge_paths Not yet run Mattdl/RehearsalRevealed/grad_norm_exp/grad_norms_plot.py
code served (permissive licence) · get_code("d8513af311af3e75")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Learning from non-stationary data streams and overcoming catastrophic forgetting still poses a serious challenge for machine learning research. Rather than aiming to improve state-of-the-art, in this work we provide insight into the limits and merits of rehearsal, one of continual learning's most established methods. We hypothesize that models trained sequentially with rehearsal tend to stay in the same low-loss region after a task has finished, but are at risk of overfitting on its sample memory, hence harming generalization. We provide both conceptual and strong empirical evidence on three benchmarks for both behaviors, bringing novel insights into the dynamics of rehearsal and continual learning in general. Finally, we interpret important continual learning works in the light of our findings, allowing for a deeper understanding of their successes. 1 * Authors contributed equally.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2104.07446")
get_code_for_paper("2104.07446")
have("2104.07446")

Connect an agent — have() is free.