AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

GAIT-prop: A biologically plausible learning rule derived from backpropagation of error

Nasir Ahmad, Marcel A. J. van Gerven, Luca Ambrogioni

NeurIPS 2020 · Architectures and efficiency › RNNs and normalization · approach: Backprop-free credit assignment

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 weak orthogonal regularizer (lambda 0 -> 0.1) at lr 1e-4 (ablation)GAIT-prop · MNIST train, 4 hidden layersfinal train accuracy: 26.36 -> 100.00Improved
add orthogonal regularizer (lambda 0 -> 10) at lr 1e-3 (ablation)GAIT-prop · MNIST train, 4 hidden layersfinal train accuracy: 17.94 -> 93.79Improved
add orthogonal regularizer (lambda 0 -> 0.1) at lr 1e-4 (ablation)backpropagation · MNIST train, 4 hidden layersfinal train accuracy: 100.00 -> 100.00No clear effect
add orthogonal regularizer (lambda 0 -> 0.1) at lr 1e-3 (ablation)target propagation · MNIST train, 4 hidden layersfinal train accuracy: 10.22 -> 9.87No clear effect
raise orthogonal regularizer to lambda 1000 (ablation)GAIT-prop · MNIST train, 4 hidden layersfinal train accuracy: 100.00 -> 96.99Hurt
lower Adam lr from 1e-3 to 1e-5 for target propagation with lambda 10 (ablation)target propagation · MNIST train, 4 hidden layersfinal train accuracy: 11.45 -> 92.16Improved
replace GAIT-prop/BP with target propagation in non-linear networkstarget propagation · MNIST non-linear networksaccuracyHurt
compare GAIT-prop with backpropagation across depths and tasksGAIT-prop · MNIST, KMNIST, Fashion-MNIST, up to 4 hidden layerstest accuracyNo clear effect
increase depth to six or eight hidden layersGAIT-prop · MNIST, 6 and 8 hidden layerstest accuracyNo clear effect
use target propagation in linear networkstarget propagation · linear network, 4 hidden layers, no regularizeraccuracyNo clear effect
reduce hidden layer widths with auxiliary units at every layerGAIT-prop · MNIST variable-width networksaccuracyNo 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