AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Reducing Flipping Errors in Deep Neural Networks

Xiang Deng, Yun Xiao, Bo Long, Zhongfei Zhang

AAAI 2022 · Architectures and efficiency › RNNs and normalization · approach: Efficient network compression and training · 5 citations (OpenAlex)

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
Restrict output-distribution changes on previously correctly classified samplesResNet-18 · CIFAR-100test accuracy: 77.31±0.33 -> 80.29±0.12Improved
Restrict output-distribution changes on previously correctly classified samplesResNet-18 · Tiny ImageNettest accuracy: 65.57±0.16 -> 67.72±0.32Improved
Restrict output-distribution changes on previously correctly classified samplesResNet-18 · CUBtest accuracy: 61.09±0.49 -> 67.65±0.25Improved
Restrict output-distribution changes on previously correctly classified samplesResNet-18 · Stanford Dogstest accuracy: 66.56±0.28 -> 70.64±0.25Improved
Restrict output-distribution changes on previously correctly classified samplesResNet-18 · CIFAR-100 with 60% label noiselast-epoch test accuracy: 32.22±0.32 -> 61.57±0.08Improved
Train representations with FERResNet-18 · CIFAR-100 to STL-10transfer accuracy: 69.71±0.14 -> 70.80±0.15Improved
Apply behavior regularization to misclassified samples as well (ablation)FER · CIFAR-100test accuracy: 80.29±0.12 -> 80.00±0.18Hurt
Apply behavior regularization to misclassified samples as well (ablation)FER · Stanford Dogstest accuracy: 70.64±0.25 -> 70.25±0.25Hurt
Use a single correct behavior instead of a weighted average of multiple behaviors (ablation)FER · CIFAR-100test accuracy: 80.29±0.12 -> 79.90±0.07Hurt
Use a single correct behavior instead of a weighted average of multiple behaviors (ablation)FER · Stanford Dogstest accuracy: 70.64±0.25 -> 70.08±0.18Hurt
Replace FER with LWR, removing both selective regularization and multi-epoch averaging (ablation)FER · CIFAR-100test accuracy: 80.29±0.12 -> 79.73±0.32Hurt
Use temperature τ = 10 instead of τ = 5 (ablation)FER with ResNet-18 · CIFAR-100test accuracy: 80.29±0.12 -> 79.63±0.15Hurt

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