AI Paper Atlas

Research map / Probabilistic and scientific modeling / Dynamical systems

Training stochastic stabilized supralinear networks by dynamics-neutral growth

Wayne WM Soo, Máté Lengyel

NeurIPS 2022 · Probabilistic and scientific modeling › Dynamical systems · approach: Latent models of neural activity · 5 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
shuffle weights within E/I quadrants of trained SSN (ablation)trained SSN 50:50 · MNISTproportion unstable: 0.0 -> 0.790Hurt
shuffle weights within E/I quadrants of trained SSN (ablation)trained SSN 80:20 · MNISTproportion unstable: 0.0 -> 0.794Hurt
replace trained weights with random Gaussian weights (ablation)trained SSN · MNISTproportion unstable: 0.0 -> 1.0Hurt
shuffle weights of trained general SSN (ablation)general SSN · GSM inferenceproportion unstable: 0.0 -> 0.446Hurt
use SSN instead of MLP readout networkSSN 50:50 · MNISTtest accuracy: 0.974 -> 0.949Hurt
use SSN instead of logistic regressionSSN 50:50 · MNISTtest accuracy: 0.914 -> 0.949Improved
use SSN instead of logistic regression on encoded-80SSN 80:20 · MNISTtest accuracy: 0.922 -> 0.952Improved
replace Gaussian-profile ring SSN with unconstrained SSNSSN 50:50 · MNISTtest accuracy: 0.223 -> 0.949Improved
train ring SSN with our method instead of Gaussian ring SSN [29]ring SSN · ring GSMnormalized cost: 1.0 -> 0.158Improved
use general unconstrained SSN instead of Gaussian ring SSNSSN · CIFAR-10-optimized GSMnormalized cost: ∼104 -> 0.745Improved
change E:I ratio from 50:50 to 80:20SSN · MNISTtest accuracy: 0.949 -> 0.952No 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