AI Paper Atlas

Research map / Time series / Causal and generative time series

Seeing the forest and the tree: Building representations of both individual and collective dynamics with transformers

Ran Liu, Mehdi Azabou, Max Dabagia, Jingyun Xiao, Eva L Dyer

NeurIPS 2022 · Time series › Causal and generative time series · approach: Dynamical system reconstruction

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
separate individual module before population module (EIT vs baseline BM)Transformer · Synthetic 3-body, decoding interactionaccuracy (%): 19.99 -> 91.70Improved
separate individual module before population moduleTransformer · Synthetic 3-body, decoding all combinationsaccuracy (%): 2.03 -> 41.99Improved
separate individual module before population moduleTransformer · Synthetic 3-body, decoding typeaccuracy (%): 4.03 -> 37.76Improved
separate individual module before population moduleTransformer · Synthetic 2-body, decoding all combinationsaccuracy (%): 87.90 -> 94.72Improved
add interaction (population) module on top of individual temporal transformer (ablation)EIT · Mihi-Chewie T=2 M-2accuracy (%): 71.71 -> 81.61Improved
add interaction (population) module on top of individual temporal transformer (ablation)EIT · Jenkins Maze velocityR2 x100: 71.18 -> 82.15Improved
add interaction (population) module on top of individual temporal transformer (ablation)EIT · Jenkins Maze positionR2 x100: 87.09 -> 94.77Improved
EIT vs supervised NDT in-domain decodingEIT · Jenkins Maze positionR2 x100: 94.88 -> 94.77No clear effect
EIT vs supervised NDT in-domain decodingEIT · Mihi-Chewie T=6 C-2accuracy (%): 94.29 -> 92.59Hurt
EIT vs NDT-Sup in-domain decodingEIT · Mihi-Chewie T=2 M-1accuracy (%): 83.93 -> 86.51Improved
frozen EIT with linear readout vs retrained NDT on a new populationEIT · Mihi-Chewie T=2 M-2 (trained on C-2)accuracy (%): 73.06 -> 75.24Improved
frozen EIT with linear readout vs retrained NDT on a new populationEIT · Mihi-Chewie T=6 C-1accuracy (%): 85.24 -> 81.42Hurt

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