AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Implicit Neural Representations with Levels-of-Experts

Zekun Hao, Arun Mallya, Serge Belongie, Ming-Yu Liu

NeurIPS 2022 · Architectures and efficiency › RNNs and normalization · approach: Sparse dictionary representations · 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
use position-dependent hierarchical tiled weights instead of fixed weights with sine-cosine positional encodingMLP · 64MP image fittingPSNR: 32.34 -> 44.24Improved
replace learned coordinate embedding grid with tiled weightsPE + CE · 64MP image fittingPSNR: 39.65 -> 44.24Improved
periodic tiling without multi-scale arrangement (interleaved) (ablation)LoE · 64MP image fittingPSNR: 44.24 -> 33.80Hurt
partition input into uniform chunks with independent MLPs (ablation)LoE · 64MP image fittingPSNR: 44.24 -> 43.13Hurt
random affine weight tiling (ablation)LoE · 64MP image fittingPSNR: 44.24 -> 42.08Hurt
feed a constant vector instead of positional encoding (ablation)LoE · 64MP image fittingPSNR: 44.24 -> 39.48Hurt
bilinear interpolation of weights instead of piecewise constant (ablation)LoE · 64MP image fittingPSNR: 44.24 -> 44.85No clear effect
quadruple hidden channels of SIREN (SIREN-L) (ablation)SIREN · Video fittingPSNR: 29.13 -> 37.71Improved
tiled weights instead of SIREN-L at similar MACsSIREN-L · Video fittingPSNR: 37.71 -> 39.98Improved
tiled weights instead of coordinate embeddingPE + CE · Video fittingPSNR: 35.83 -> 39.98Improved
tiled weights without hierarchical arrangement (ablation)LoE · Novel view synthesis (Family)PSNR: 31.46 -> 27.73Hurt
hierarchical tiled weights instead of PE MLPMLP · Novel view synthesis (Family)PSNR: 30.50 -> 31.46Improved

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