AI Paper Atlas

Research map / Continual learning / Online and replay-based

Overcoming Catastrophic Forgetting by Bayesian Generative Regularization

Pei-Hung Chen, Wei Wei, Cho-Jui Hsieh, Bo Dai

ICML 2021 · Continual learning › Online and replay-based · approach: Theoretical analysis of forgetting

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add EBM generative regularization to Bayesian VCL (ablation)VCL · Fashion-MNISTaverage classification accuracy (%): 76.9 -> 97.2Improved
add EBM generative regularization to Bayesian VCL (ablation)VCL · CUBaverage classification accuracy (%): 67.4 -> 78.8Improved
remove Bayesian parameter sampling and KL divergence, retaining the generative loss (ablation)BGR · Fashion-MNISTaverage classification accuracy (%): 97.2 -> 87.9Hurt
remove Bayesian parameter sampling and KL divergence, retaining the generative loss (ablation)BGR · CUBaverage classification accuracy (%): 78.8 -> 74.0Hurt
add generative EBM loss without Bayesian updates (ablation)VCL · Fashion-MNISTaverage classification accuracy (%): 76.9 -> 87.9Improved
add generative EBM loss without Bayesian updates (ablation)VCL · CUBaverage classification accuracy (%): 67.4 -> 74.0Improved
add L2 regularization to GEN (ablation)GEN · Fashion-MNISTaverage classification accuracy (%): 87.9 -> 90.9Improved
add L2 regularization to GEN (ablation)GEN · CUBaverage classification accuracy (%): 74.0 -> 72.8Hurt
use a small GAN of similar size instead of the larger VGR generator (ablation)VGR · Split-MNISTaverage classification accuracy (%): 88Hurt
reinitialize the classifier for each task in VGR-GAN-ONLY (ablation)VGR · Split-MNISTaverage classification accuracy (%): 66Hurt
add BGR generative regularization to iTAMLiTAML · Split-MNISTaverage classification accuracy (%): 97.8 -> 98.4Improved
train with generative regularization rather than SGD aloneBGR · Permuted-MNIST, digit task after another taskaccuracy (%): 99.7% to 54.2% -> 99.7% to 95.0%Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy