AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Bit-Mixer: Mixed-Precision Networks With Runtime Bit-Width Selection

Adrian Bulat, Georgios Tzimiropoulos

ICCV 2021 · Architectures and efficiency › Quantization · approach: Mixed-precision quantization · 26 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
add Transitional Batch-Norm layers (ablation)Bit-Mixer ResNet-18 · ImageNet random layer-wise bitstop-1 accuracy (%): 8.8 -> 65.8Improved
add Transitional Batch-Norm layers (ablation)Bit-Mixer ResNet-18 · ImageNet 4-bittop-1 accuracy (%): 8.2 -> 69.2Improved
add Transitional Batch-Norm layers (ablation)Bit-Mixer ResNet-18 · ImageNet 2-bittop-1 accuracy (%): 10.2 -> 64.4Improved
add FP32 teacher distillation to Stage III (ablation)Bit-Mixer ResNet-18 · ImageNet 2-bittop-1 accuracy (%): 64.0 -> 64.4Improved
add Stage I teacher distillation to Stage II (ablation)Bit-Mixer ResNet-18 · ImageNet 2-bittop-1 accuracy (%): 65.1 -> 65.6Improved
add Stage I teacher distillation to Stage III (ablation)Bit-Mixer ResNet-18 · ImageNet random layer-wise bitstop-1 accuracy (%): 65.5 -> 65.7Improved
add layer-wise random bit training (Stage III) after Stage IIBit-Mixer ResNet-18 · ImageNet 2-bittop-1 accuracy (%): 65.6 -> 64.4Hurt
include 1-bit in the shared 4-3-2-1 meta-networkBit-Mixer ResNet-18 · ImageNet 1-bittop-1 accuracy (%): 59.0 -> 57.3Hurt
run Stage III with binary-ternary 1.5-bit weightsBit-Mixer ResNet-18 · ImageNet 1.5-bittop-1 accuracy (%): 64.0 -> 62.1Hurt
use asymmetric instead of symmetric quantization (ablation)Bit-Mixer ResNet-18 · ImageNet 2-bittop-1 accuracy (%): 65.1 -> 65.2No clear effect
use per bit-width scaling factors instead of shared scale-based quantization (ablation)Bit-Mixer ResNet-18 · ImageNettop-1 accuracy (%): 5--10% worseHurt
sample random sub-paths instead of progressive Stage III training (ablation)Bit-Mixer ResNet-18 · ImageNettop-1 accuracy (%)Hurt

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