AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Attention-Gated Brain Propagation: How the brain can implement reward-based error backpropagation

Isabella Pozzi, Sander Bohte, Pieter Roelfsema

NeurIPS 2020 · Architectures and efficiency › RNNs and normalization · approach: Backprop-free credit assignment · 14 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
Use BrainProp reward-based learning instead of error backprop (Error-BP)BrainProp · Tiny ImageNet deep convaccuracy (%): 47.35 -> 47.50No clear effect
Use BrainProp instead of Error-BPBrainProp · CIFAR100 deep convaccuracy (%): 63.39 -> 59.58Hurt
Use BrainProp instead of Error-BPBrainProp · CIFAR10 deep convaccuracy (%): 88.48 -> 88.88No clear effect
Use BrainProp instead of Error-BPBrainProp · MNIST convaccuracy (%): 99.39 -> 99.31No clear effect
Use BrainProp instead of Error-BPBrainProp · MNIST denseaccuracy (%): 98.54 -> 98.68No clear effect
Use BrainProp instead of Error-BPBrainProp · CIFAR10 convaccuracy (%): 70.69 -> 69.30Hurt
Use BrainProp instead of Error-BPBrainProp · CIFAR100 convaccuracy (%): 36.56 -> 34.07Hurt
Use locally connected layers instead of shared-weight convolution (ablation)BrainProp · CIFAR10accuracy (%): 69.30 -> 64.92Hurt
Use BrainProp instead of Error-BPBrainProp · training epochs, CIFAR100 deep convepochs to converge: 101 -> 218Hurt

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