AI Paper Atlas

Research map / Optimization / Implicit bias and generalization

Understanding Straight-Through Estimator in Training Activation Quantized Neural Nets

Penghang Yin, Jiancheng Lyu, Shuai Zhang, Stanley Osher, Yingyong Qi, Jack Xin

ICLR 2019 · Optimization › Implicit bias and generalization · approach: Neural tangent kernel analysis · 60 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
Use clipped ReLU STE instead of identity STE (ablation)LeNet5 · MNIST 2-bit activationsvalidation accuracy: 98.49 -> 99.23Improved
Use clipped ReLU STE instead of identity STE (ablation)LeNet5 · MNIST 4-bit activationsvalidation accuracy: 98.98 -> 99.24Improved
Use clipped ReLU STE instead of vanilla ReLU STE (ablation)LeNet5 · MNIST 2-bit activationsvalidation accuracy: 99.24 -> 99.23No clear effect
Use clipped ReLU STE instead of vanilla ReLU STE (ablation)LeNet5 · MNIST 4-bit activationsvalidation accuracy: 99.32 -> 99.24Hurt
Use clipped ReLU STE instead of identity STE (ablation)VGG11 · CIFAR-10 2-bit activationsvalidation accuracy: 86.58 -> 90.92Improved
Use clipped ReLU STE instead of vanilla ReLU STE (ablation)VGG11 · CIFAR-10 2-bit activationsvalidation accuracy: 88.69 -> 90.92Improved
Use clipped ReLU STE instead of identity STE (ablation)VGG11 · CIFAR-10 4-bit activationsvalidation accuracy: 90.19 -> 92.08Improved
Use clipped ReLU STE instead of vanilla ReLU STE (ablation)VGG11 · CIFAR-10 4-bit activationsvalidation accuracy: 92.01 -> 92.08Improved
Use clipped ReLU STE instead of identity STE (ablation)ResNet20 · CIFAR-10 2-bit activationsvalidation accuracy: 46.52 -> 88.39Improved
Use clipped ReLU STE instead of vanilla ReLU STE (ablation)ResNet20 · CIFAR-10 2-bit activationsvalidation accuracy: 48.05 -> 88.39Improved
Use clipped ReLU STE instead of identity STE (ablation)ResNet20 · CIFAR-10 4-bit activationsvalidation accuracy: 54.16 -> 91.24Improved
Use clipped ReLU STE instead of vanilla ReLU STE (ablation)ResNet20 · CIFAR-10 4-bit activationsvalidation accuracy: 86.59 -> 91.24Improved

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