SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2405.18758 · ICML · 2024

Learning to Continually Learn with the Bayesian Principle

Soochan Lee, Gunhee Kim, Hyeonseong Jeon, Jaehyeon Son

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 7 functions out of this paper's own repositories and ran 7 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
soochan-lee/SB-MCL canonical 7 of 7
FunctionStatusWhere it lives
get_continual_mask Ran soochan-lee/SB-MCL/models/continual_transformer.py
pointer only (licence: NONE) · get_code("2c6bc30a15479d82")
get_y Ran soochan-lee/SB-MCL/dataset.py
pointer only (licence: NONE) · get_code("87a56660bd4c8a47")
process_gnt Ran soochan-lee/SB-MCL/dataset.py
pointer only (licence: NONE) · get_code("43e280293e440b88")
resize_image Ran soochan-lee/SB-MCL/dataset.py
pointer only (licence: NONE) · get_code("20f1b95de7047461")
sample_test_attachment Ran soochan-lee/SB-MCL/models/continual_transformer.py
pointer only (licence: NONE) · get_code("77813c4947b36c1d")
sequential_bayes Ran soochan-lee/SB-MCL/models/sbmcl.py
pointer only (licence: NONE) · get_code("9fe72d2912ab2734")
slice_shots Ran soochan-lee/SB-MCL/models/sbmcl.py
pointer only (licence: NONE) · get_code("2c8587cbab576d85")

Repositories linked to this paper

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

Abstract

In the present era of deep learning, continual learning research is mainly focused on mitigating forgetting when training a neural network with stochastic gradient descent on a non-stationary stream of data. On the other hand, in the more classical literature of statistical machine learning, many models have sequential Bayesian update rules that yield the same learning outcome as the batch training, i.e., they are completely immune to catastrophic forgetting. However, they are often overly simple to model complex realworld data. In this work, we adopt the metalearning paradigm to combine the strong representational power of neural networks and simple statistical models' robustness to forgetting. In our novel meta-continual learning framework, continual learning takes place only in statistical models via ideal sequential Bayesian update rules, while neural networks are meta-learned to bridge the raw data and the statistical models. Since the neural networks remain fixed during continual learning, they are protected from catastrophic forgetting. This approach not only achieves significantly improved performance but also exhibits excellent scalability. Since our approach is domainagnostic and model-agnostic, it can be applied to a wide range of problems and easily integrated with existing model architectures.

For agents

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

get_harvested_code_for_paper("2405.18758")
get_code_for_paper("2405.18758")
have("2405.18758")

Connect an agent — have() is free.