AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

CORNN: Convex optimization of recurrent neural networks for rapid inference of neural dynamics

Fatih Dinc, Adam Shai, Mark Schnitzer, Hidenori Tanaka

NeurIPS 2023 · Architectures and efficiency › RNNs and normalization · approach: Brain-inspired network comparison · 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
train with teacher forcing in BPTT (ablation)BPTT · Chaotic RNN reproduction (Fig. S1)convergence speedImproved
use cross-entropy loss instead of L2 loss (ablation)BPTT · Chaotic RNN reproduction (Fig. S1)reproduction accuracyImproved
replace FORCE with CORNNdRNN training · Random chaotic RNN, 5000 neuronstraining time: 4 orders of magnitudeImproved
add fixed-point initialization to CORNN (ablation)CORNN · Chaotic RNN reproduction (Fig. S6C)convergence speed and accuracyImproved
fixed-point initialization added to other algorithmsother solvers · Chaotic RNN reproductionstabilityImproved
run CORNN on GPU instead of CPUCORNN · Chaotic RNN reproduction (Fig. S5)efficiencyImproved
subsample observed neurons to 10% and jitter time-scalesCORNN · Timed-response taskreproduction of neural dynamicsNo clear effect
subsample observed neurons to 10% and jitter time-scalesCORNN · Timed-response tasksub-connectivity matrix reconstructionHurt
increase number of training trialsCORNN · Large generator networks (Fig. S8)reproduction accuracyImproved

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