AI Paper Atlas

Research map / Architectures and efficiency / CNN design

SPIN: An Empirical Evaluation on Sharing Parameters of Isotropic Networks

Chien-Yu Lin, Anish Prabhu, Thomas Merth, Sachin Mehta, Anurag Ranjan, Maxwell Horton, Mohammad Rastegari

ECCV 2022 · Architectures and efficiency › CNN design · approach: Compact CNN architectures · 2 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
weight fusion init from pretrained net (channel weighted mean) instead of random init (ablation)WS-ConvMixer-768/32 share rate 2 · ImageNet-1ktop-1 accuracy (%): 73.23 -> 75.15Improved
fuse weights without pretrained initialization (Mean) (ablation)WS-ConvMixer · ImageNet-1ktop-1 accuracy (%): 74.29 -> 74.25No clear effect
initialize with pretrained weights when using Mean fusion (ablation)WS-ConvMixer · ImageNet-1ktop-1 accuracy (%): 74.25 -> 75.28Improved
add lightweight affine weight transformation (G=64) (ablation)WS-ConvMixer share rate 2 · ImageNet-1ktop-1 accuracy (%): 73.29 -> 73.36No clear effect
add affine weight transformation with G=1 (ablation)WS-ConvMixer · ImageNet-1ktop-1 accuracy (%): 73.29 -> 72.87Hurt
sequential instead of strided sharing mapping (ablation)WS-ConvMixer uniform · ImageNet-1ktop-1 accuracy (%): 72.80 -> 73.29Improved
random sharing mapping instead of sequential (ablation)WS-ConvMixer uniform · ImageNet-1ktop-1 accuracy (%): 73.29 -> DivergedHurt
share front layers instead of back layers (ablation)WS-ConvMixer sequential · ImageNet-1ktop-1 accuracy (%): 72.35 -> 73.31Improved
share only pointwise conv rather than also BN, bias, depthwise (ablation)WS-ConvMixer share rate 2 · ImageNet-1ktop-1 accuracy (%): 73.17 -> 73.29Improved
share BatchNorm as well as other components (ablation)WS-ConvMixer · ImageNet-1ktop-1 accuracy (%): 73.17 -> DivergedHurt
weight sharing (rate 8) instead of depth scaling at equal costConvMixer · ImageNet-1ktop-1 accuracy (%): 57.75 -> 67.19Improved
weight sharing rate 2 with fusion on ConvMixer-1536/20ConvMixer-1536/20 · ImageNet-1ktop-1 accuracy (%): 78.03 -> 78.47Improved

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