AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Deep Learning with Learnable Product-Structured Activations

Saanjali Maharaj, Prasanth B. Nair

ICLR 2026 · Architectures and efficiency › RNNs and normalization · approach: Theory of depth and width

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 product-structured learnable activations (LRNN) to SPDERSPDER · Cameraman image (~197k params)PSNR (dB): 49.0 -> 107.9Improved
add product-structured learnable activations (LRNN)SPDER · Audio representation: bachMSE loss (x1e-4): 1.12 -> 0.10Improved
add product-structured learnable activations (LRNN)SPDER · Audio representation: reggaeMSE loss (x1e-4): 24.8 -> 7.93Improved
add product-structured learnable activations (LRNN)SPDER · Audio representation: readingMSE loss (x1e-4): 8.88 -> 1.86Improved
add product-structured learnable activations (LRNN)SPDER · Audio representation: readingfrequency similarity rho_AG: 0.9324 -> 0.9862Improved
use LRNN-SPDER instead of SPDER/SIREN for 40 dB targetSIREN, SPDER · ImageNet 1000 images, 40 dB targetsuccess rate (%): 100% vs failures of 98.2% (SIREN) and 73.6% (SPDER)Improved
use LRNN instead of WIREWIRE · Sparse-view CT 256x256 chestPSNR (dB): 28.83 -> 29.13Improved
use LRNN instead of WIREWIRE · Sparse-view CT 256x256 chestSSIM: 0.6413 -> 0.7455Improved
use 16k-parameter LRNN instead of 132k-parameter SIRENSIREN · High-frequency Poisson PDE (n=2)MSE: 132k -> 16k parametersImproved
add LayerNorm after each LRNN layer (ablation)LRNN · deep LRNN trainingconvergenceImproved
use periodic activations in univariate components (ablation)LRNN · high-frequency signal representationspectral biasImproved
share activation functions across neurons in each layer (ablation)LRNN · image representationfidelity vs parameter countMixed

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