AI Paper Atlas

Research map / Probabilistic and scientific modeling / Dynamical systems

Identifying Learning Rules From Neural Network Observables

Aran Nayebi, Sanjana Srivastava, Surya Ganguli, Daniel L Yamins

NeurIPS 2020 · Probabilistic and scientific modeling › Dynamical systems · approach: Synaptic plasticity learning rules · 7 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 aggregate statistics of weights, activations, or layer-wise activity changes across the full learning trajectoryRandom Forest learning-rule classifier · Architectures, tasks, and training curricula represented in the studylearning-rule identification accuracyImproved
Use a Random Forest instead of a linear SVM or Conv1D MLPLearning-rule identification · Full trajectories and observable measurestest accuracyImproved
Use activation statistics rather than weight or layer-wise activity-change statisticsLearning-rule identification with unit subsampling and measurement noise · FA vs. IA on ResNet-18, ImageNet and SimCLRtest accuracyImproved
Measure trajectory points further apart in timeLearning-rule identification from subsampled trajectories · Multiple observable measurestest accuracyImproved
Average observable statistics over the full learning trajectory instead of retaining the trajectory (ablation)SVM learning-rule classifier · All observable measurestest accuracyHurt
Average observable statistics over the full learning trajectory instead of retaining the trajectory (ablation)Random Forest learning-rule classifier · Weight observablestest accuracyNo clear effect
Train only on one consecutive portion of the learning trajectory rather than sampling across it (ablation)Random Forest learning-rule classifier · Trajectory undersamplinggeneralization performanceHurt
Use weight observables rather than other individual observable measuresLearning-rule identification with trajectory undersampling · Trajectory subsamplingtest accuracyImproved
Use activation observables rather than weight or layer-wise activity-change observablesLearning-rule identification · Unit undersampling and measurement noisetest accuracyImproved
Generalize from supervised cross-entropy training to self-supervised SimCLR (ablation)Random Forest learning-rule classifier · Held-out tasktest accuracyHurt
Use a linear classifier rather than a non-linear classifier (ablation)Learning-rule identification from individual observables · Consecutive portions of the learning trajectorytest accuracyHurt
Identify learning rules from activation trajectories rather than temporally averaged post-synaptic activitiesLearning-rule identification · Multiple learning-rule classesseparabilityImproved

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