AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Learning rule influences recurrent network representations but not attractor structure in decision-making tasks

Brandon J McMahan, Michael Kleinman, Jonathan Kao

NeurIPS 2021 · Architectures and efficiency › RNNs and normalization · approach: Efficient recurrent architectures · 1 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
train with BPTT, GA or FF instead of Hebbian learningRNN representations · RDM and CDISVCCA MDS overlapMixed
vary learning rule (BPTT, GA, H, FF)RNN attractor topology · RDM and CDIattractor silhouette: -0.05; 0.004No clear effect
increase number of contexts N from 2 to 6BPTT vs GA representations · N-CDISVCCA silhouette: 0.10 -> 0.44Improved
increase number of contexts N from 2 to 6BPTT vs GA attractor structure · N-CDIattractor silhouette: 0.030 -> 7e-4No clear effect
increase task complexity beyond N=2H and FF learning · N-CDIreaching 90% accuracyHurt
increase input noise varianceattractor topology · RDMnumber of attractors at low inputImproved
increase input noise varianceuniversality across learning rules · RDMattractor silhouetteImproved
add high input noise (variance 1.0 vs 0.1)integration mechanism · RDMintegration typeMixed
increase N-CDI contexts beyond 3input-noise effect on attractor topology (BPTT) · N-CDIattractor similarity across noise 0.1, 0.5, 1.0Improved
change activation function (ablation)attractor topology · RDM vs CDIattractor topologyMixed
use single neuron readout for non-Hebbian RNNs (ablation)SVCCA representations · RDM and CDISVCCA clusteringNo clear effect

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