AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

DISCOVER: Making Vision Networks Interpretable via Competition and Dissection

Konstantinos P. Panousis, Sotirios Chatzis

NeurIPS 2023 · Architectures and efficiency › RNNs and normalization · approach: Representation geometry and collapse · 0 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
Replace GELU with stochastic LWTA competition using 2 competitorsDeiT-T · ImageNet-1kaccuracy: 72.2 -> 72.5Improved
Replace GELU with stochastic LWTA competition using 16 competitorsDeiT-T · ImageNet-1kaccuracy: 72.2 -> 71.1Hurt
Replace GELU with stochastic LWTA competition using 2 competitorsDeiT-S · ImageNet-1kaccuracy: 77.0 -> 77.3Improved
Replace ReLU with stochastic LWTA competition using 2 competitorsResNet-18 · Places365accuracy: 52.25 -> 53.9Improved
Use stochastic LWTA competition with 16 competitors instead of GELUDeiT-T · ImageNet Val, Broden concept setCLIP cosine similarity: 0.6250 -> 0.6460Improved
Use CosSimCubed instead of SoftWPMI for concept matching (ablation)DeiT-T/16 · ImageNet Val, Broden concept setCLIP cosine similarity: 0.6460 -> 0.6543Improved
Use stochastic LWTA competition with 16 competitors instead of GELUDeiT-T · ImageNet Val, 3k concept setCLIP cosine similarity: 0.6226 -> 0.6650Improved
Use CosSimCubed instead of SoftWPMI for concept matching (ablation)DeiT-T/16 · ImageNet Val, 3k concept setCLIP cosine similarity: 0.6650 -> 0.6699Improved
Use stochastic LWTA competition with 8 competitors instead of GELUDeiT-T · CIFAR-100 Train, ImageNet concept setneuron identification accuracy: 53.00% -> 53.10%No clear effect
Use stochastic LWTA competition with 12 competitors instead of GELUDeiT-S · CIFAR-100 Train, ImageNet concept setneuron identification accuracy: 53.00% -> 56.10%Improved
Use stochastic LWTA competition instead of conventional GELUDeiT-T/16 · Average neuron identification accuracyneuron identification accuracy: 78.00% -> 78.30%Improved
Use stochastic LWTA competition instead of conventional GELUDeiT-S/2 · Average neuron identification accuracyneuron identification accuracy: 78.00% -> 78.70%Improved

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