AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Dual Propagation: Accelerating Contrastive Hebbian Learning with Dyadic Neurons

Rasmus Høier, D. Staudt, Christopher Zach

ICML 2023 · Architectures and efficiency › RNNs and normalization · approach: Hebbian and predictive coding

Read the paper

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
replace BP with DPVGG16 · CIFAR10Top-1 accuracy: 92.26±0.23 -> 92.30±0.11No clear effect
replace BP with DPVGG16 · CIFAR100Top-1 accuracy: 69.63±0.24 -> 69.57±0.51No clear effect
replace BP with DPVGG16 · ImageNet32x32Top-1 accuracy: 41.28±0.19 -> 41.48±0.19No clear effect
use distinct feedback weights trained with Kolen-Pollack learning (ablation)DP with VGG16 · CIFAR10Top-1 accuracy: 92.30±0.11 -> 91.84±0.11Hurt
use distinct feedback weights trained with Kolen-Pollack learning (ablation)DP with VGG16 · CIFAR100Top-1 accuracy: 69.57±0.51 -> 70.40±0.25Improved
use distinct feedback weights trained with Kolen-Pollack learning (ablation)DP with VGG16 · CIFAR100Top-5 accuracy: 88.36±0.13 -> 88.57 ± 0.15Improved
increase βL from 1.0 to 10 (ablation)DP with MLP · MNISTtest accuracy (%): 98.38 -> 94.95Hurt
increase βL from 10 to 100 (ablation)DP with MLP · MNISTtest accuracy (%): 94.95 -> 85.42Hurt
perform five inference passes and update weights after each pass (ablation)DP with MLP · MNISTtest accuracy (%): 98.43 ±0.03 -> 98.40 ±0.02No clear effect
do not reset hidden and output units before each batch (ablation)DP with MLP · MNISTtest accuracy (%): 98.43 ±0.03 -> 98.42 ±0.07No clear effect
update all neurons in parallel (ablation)DP with MLP · MNISTtest accuracy (%): 98.43 ±0.03 -> 98.47 ±0.04No clear effect
randomly select layers for 100 updates (ablation)DP with MLP · MNISTtest accuracy (%): 98.43 ±0.03 -> 98.48 ±0.11No 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