AI Paper Atlas

Research map / Architectures and efficiency / Quantization

AutoQ: Automated Kernel-Wise Neural Network Quantization

Qian Lou, Feng Guo, Minje Kim, Lantao Liu, Lei Jiang.

ICLR 2020 · Architectures and efficiency › Quantization · approach: Post-training weight quantization · 20 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
Assign a separate bitwidth to each weight kernel instead of one per layerAutoQ, resource-constrained quantization · ImageNet, ResNet-18/ResNet-50/SqueezeNetV1/MobileNetV2top-1 error (%)Improved
Assign a separate bitwidth to each weight kernel instead of one per layerAutoQ, resource-constrained quantization · ImageNet, ResNet-18/ResNet-50/SqueezeNetV1/MobileNetV2inference latency (ms)Improved
Assign a separate bitwidth to each weight kernel instead of one per layerAutoQ, accuracy-guaranteed quantization · ImageNet, ResNet-18/ResNet-50/SqueezeNetV1/MobileNetV2inference latency (ms): 42.2%Improved
Use kernel-wise rather than layer-wise quantizationAutoQ, accuracy-guaranteed quantization · ImageNet, ResNet-18/ResNet-50/SqueezeNetV1/MobileNetV2top-1 accuracy: -0.1%Mixed
Use hierarchical DRL with HLC goals and LLC kernel actions rather than one-level DDPGKernel-wise quantization search on MobileNetV2 · MobileNetV2search reward (inference accuracy): ~70% accuracy after only 200 episodes; 20% inference accuracy until 250 episodesImproved
Include extrinsic environment reward in the intrinsic reward, in addition to HLC-goal completionAutoQ LLC (relative to HIRO intrinsic reward) · MobileNetV2 kernel-wise searchsearch reward (inference accuracy): extra 200 episodes for HIRO to reach only 60% accuracyImproved
Use kernel-wise rather than layer-wise quantizationAutoQ on the spatial BitFunsion accelerator · CNN models evaluated on BitFunsioninference latency: 39.04%Improved
Use kernel-wise rather than layer-wise quantizationAutoQ on the spatial BitFunsion accelerator · CNN models evaluated on BitFunsioninference energy: 33.34%Improved
Quantize weight sub-kernels independently instead of whole kernels (ablation)AutoQ on the spatial BitFunsion accelerator · CNN models evaluated on BitFunsioninference latencyNo clear effect

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