AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Brains on Beats

Umut Güçlü, Jordy Thielen, Michael Hanke, Marcel van Gerven

NeurIPS 2016 · Architectures and efficiency › RNNs and normalization · approach: Representation geometry and collapse · 16 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 both time and frequency streams (DNN-TF) instead of time only (ablation)DNN music tagger · MagnaTagATunemean AUC: 0.8852 -> 0.8939Improved
use frequency-domain input instead of time-domain (ablation)DNN music tagger · MagnaTagATunemean AUC: 0.8852 -> 0.8905Improved
ensemble DNN-T and DNN-F predictionsDNN taggers · MagnaTagATunemean AUC: 0.8939 -> 0.8982Improved
ensemble all three DNN modelsDNN taggers · MagnaTagATunemean AUC: 0.8982 -> 0.9007Improved
replace task-optimized DNN-TF with random-weight DNN (ablation)DNN-TF · STG RSArepresentational similarity (delta r): 0.21Hurt
replace DNN-TF with mel-frequency spectrum features (ablation)DNN-TF · STG RSArepresentational similarity (delta r): 0.42Hurt
replace DNN-TF with MFCC features (ablation)DNN-TF · STG RSArepresentational similarity (delta r): 0.21Hurt
replace DNN-TF with low-quefrency mel spectrum (ablation)DNN-TF · STG RSArepresentational similarity (delta r): 0.44Hurt
replace DNN-TF with high-quefrency mel spectrum (ablation)DNN-TF · STG RSArepresentational similarity (delta r): 0.34Hurt
random weights instead of trained weights in searchlight (ablation)DNN-TF · STG layer gradientlayer-position correlation: 0.7255 -> -0.2175Hurt
tighten significance threshold from 0.05 to 0.01 in searchlightDNN-TF · STG voxelssurviving voxels (%) / gradient r: 85.53% -> 75.32%; r 0.7255 -> 0.7332No clear effect
go from shallow Layer 1 to deep Layer 8DNN-TF · STG ROI RSASpearman r: 0.6811 -> 0.4429Hurt

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