Research map / Probabilistic and scientific modeling / Dynamical systems
AMAG: Additive, Multiplicative and Adaptive Graph Neural Network For Forecasting Neuron Activity
Jingyuan Li, Leo Scholl, Trung Le, Pavithra Rajeswaran, Amy L Orsborn, Eli Shlizerman
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 changed | Where | Measured | Effect |
|---|---|---|---|
| remove both Add and Modulator modules (ablation) | AMAG · Monkey C multi-step | R2: 0.657 -> 0.424 | Hurt |
| remove self-connection in graph (ablation) | AMAG · Monkey C multi-step | R2: 0.657 -> 0.427 | Hurt |
| remove Add from graph operation (keep Modulator) (ablation) | AMAG · Monkey C multi-step | R2: 0.657 -> 0.647 | Hurt |
| remove Modulator (ablation) | AMAG · Monkey C multi-step | R2: 0.657 -> 0.652 | Hurt |
| remove sample-dependent Adaptor from Add (ablation) | AMAG · Monkey C multi-step | R2: 0.657 -> 0.655 | Hurt |
| keep only Adaptor in Add module (ablation) | AMAG · Monkey C multi-step | R2: 0.657 -> 0.648 | Hurt |
| make adjacency matrix non-learnable with random init (ablation) | AMAG · Monkey C multi-step | R2: 0.657 -> 0.575 | Hurt |
| make adjacency matrix non-learnable with correlation init (ablation) | AMAG · Monkey C multi-step | R2: 0.657 -> 0.617 | Hurt |
| use random instead of correlation initialization with learnable adjacency (ablation) | AMAG · Monkey C multi-step | R2: 0.657 -> 0.652 | No clear effect |
| use KNN pruning instead of similarity threshold 0.5 for DCRNN (ablation) | DCRNN · Monkey C multi-step | R2: 0.573 -> 0.606 | Improved |
| use KNN pruning instead of similarity threshold 0.8 for DCRNN (ablation) | DCRNN · Monkey C multi-step | R2: 0.445 -> 0.606 | Improved |
| use GRU-based GS4-G instead of state-space GS4-S temporal learning (ablation) | GraphS4mer · Monkey M multi-step | R2: 0.181 -> 0.268 | Improved |
Later papers that compared with this method
1 later accepted paper reports AMAG as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- DCHO: A Decomposition–Composition Framework for Predicting Higher-Order Brain Connectivity to Enhance Diverse Downstream Applications (AAAI 2026 · better on 8 of 8 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- POCO: Scalable Neural Forecasting through Population Conditioning (NeurIPS 2025 · similarity 0.86)
- A Unified, Scalable Framework for Neural Population Decoding (NeurIPS 2023 · similarity 0.85)
- Neural Latents Benchmark ‘21: Evaluating latent variable models of neural population activity (NeurIPS 2021 · similarity 0.85)
- BLEND: Behavior-guided Neural Population Dynamics Modeling via Privileged Knowledge Distillation (ICLR 2025 · similarity 0.85)
- Efficient and Effective Time-Series Forecasting with Spiking Neural Networks (ICML 2024 · similarity 0.84)
- Latent Diffusion for Neural Spiking Data (NeurIPS 2024 · similarity 0.83)
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.
