AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Locality-Aware Generalizable Implicit Neural Representation

Doyup Lee, Chiheon Kim, Minsu Cho, Wook-Shin Han

NeurIPS 2023 · Architectures and efficiency › Attention and transformers · approach: Transformer expressivity and circuits · 3 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 selective token aggregation (vs IPC weight modulation) (ablation)Locality-aware INR · ImageNette 178x178PSNR: 34.54 -> 37.46Improved
add multi-band feature modulation (vs simple MLP stack) (ablation)Locality-aware INR · ImageNette 178x178PSNR: 33.90 -> 37.46Improved
add selective token aggregation (ablation)Locality-aware INR · ShapeNet Lamp 3 viewsPSNR: 25.31 -> 26.00Improved
add multi-band feature modulation (ablation)Locality-aware INR · ShapeNet Lamp 3 viewsPSNR: 25.78 -> 26.00Improved
add selective token aggregation (ablation)Locality-aware INR · FFHQ 256x256PSNR: 34.52 -> 38.01Improved
add multi-band feature modulation (ablation)Locality-aware INR · FFHQ 256x256PSNR: 33.65 -> 38.01Improved
reverse bandwidth order to (32,128) (ablation)Locality-aware INR · ImageNette 178x178PSNR: 37.46 -> 35.00Hurt
use equal bandwidths (128,128) (ablation)Locality-aware INR · ImageNette 178x178PSNR: 37.46 -> 35.30Hurt
raise query bandwidth sigma_q from 16 to 128 (ablation)Locality-aware INR · ImageNette 178x178PSNR: 37.46 -> 35.58Hurt
increase modulation levels from L=1 to L=3 (ablation)Locality-aware INR · FFHQ 256x256PSNR: 37.09 -> 40.13Improved
increase modulation levels from L=3 to L=4 (ablation)Locality-aware INR · FFHQ 256x256PSNR: 40.13 -> 39.79No clear effect
add linear layers in selective token aggregation (ablation)Locality-aware INR · ImageNette 178x178PSNR: 31.57 -> 37.46Improved

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