AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Non-deep Networks

Ankit Goyal, Alexey Bochkovskiy, Jia Deng, Vladlen Koltun

NeurIPS 2022 · Architectures and efficiency › RNNs and normalization · approach: Efficient network compression and training · 8 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 rand-augment data augmentation (ablation)ParNet baseline · ImageNet valtop-1 accuracy: 75.02 -> 75.12No clear effect
add Skip-Squeeze-Excitation (SSE) (ablation)ParNet baseline · ImageNet valtop-1 accuracy: 75.12 -> 76.55Improved
replace ReLU with SiLU (ablation)ParNet with SSE · ImageNet valtop-1 accuracy: 76.55 -> 76.60No clear effect
use 3 parallel branches instead of a single stream (ablation)ParNet-M · ImageNet valtop-1 accuracy: 75.83 -> 76.6Improved
use VGG-style RepVGG-SSE blocks instead of ResNet blocks at depth 12 (ablation)ParNet-M · ImageNet valtop-1 accuracy: 73.52 -> 75.83Improved
add SSE to the wide ResNet12 (ablation)ResNet12-Wide · ImageNet valtop-1 accuracy: 73.52 -> 73.91Improved
use bottleneck ResNet block at depth 14 instead of basic block at depth 12 (ablation)ResNet-Wide · ImageNet valtop-1 accuracy: 73.52 -> 72.06Hurt
fuse 3x3 and 1x1 branches into one 3x3 conv for inferenceParNet-L · ImageNet inference speedspeed (samples/sec): 112 -> 154Improved
distribute the streams across 3 GPUsParNet-L · ImageNet inference latencylatency (ms): 6.50 -> 4.01Improved
longer training (200 epochs, cosine annealing)ParNet-XL · ImageNet valtop-1 accuracy: 78.55 -> 78.97Improved
train and test at resolution 320ParNet-XL · ImageNet valtop-1 accuracy: 78.97 -> 80.32Improved
add 10-crop testingParNet-XL · ImageNet valtop-1 accuracy: 80.32 -> 80.72Improved

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