Research map / Architectures and efficiency / Spiking networks
Neural Data Transformer 2: Multi-context Pretraining for Neural Spiking Activity
Joel Ye, Jennifer L Collinger, Leila Wehbe, Robert Gaunt
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 |
|---|---|---|---|
| add cross-subject human pretraining data (ablation) | NDT2 · Human reach intent decoding P2 | velocity R2: 0.487 -> 0.503 | Improved |
| add cross-task (online control) pretraining data (ablation) | NDT2 · Human reach intent decoding P2 | velocity R2: 0.444 -> 0.503 | Improved |
| add cross-task pretraining data (ablation) | NDT2 · Human reach intent decoding P3 | velocity R2: 0.493 -> 0.515 | Improved |
| add monkey data to pretraining or decoder preparation (ablation) | NDT2 · Human reach intent decoding | velocity R2: 0.503 -> 0.486 (P2), 0.515 -> 0.472 (P3) | Hurt |
| pretrain on other sessions/subjects/tasks (multi-context) vs single-session | NDT2 · RTT monkey Indy, 300 trials | velocity R2 / NLL | Improved |
| add stitching layers instead of context tokens for multi-context data | NDT1 · RTT sorted | velocity R2 / NLL | Hurt |
| use unsorted instead of sorted spikes | NDT2 · RTT | velocity R2 | Improved |
| add session context tokens (1 vs none) (ablation) | NDT2 · sorted multisession NLL | NLL | Improved |
| increase number of session/subject tokens beyond one (ablation) | NDT2 · multi-session / subject transfer | NLL | No clear effect |
| use smaller patches (8 vs 32 neurons) (ablation) | NDT2 · RTT | NLL / decoding R2 | Mixed |
| vary mask ratio (ablation) | NDT2 · 5 datasets | decoding | No clear effect |
Closest papers by meaning
- Neuroformer: Multimodal and Multitask Generative Pretraining for Brain Data (ICLR 2024 · similarity 0.90)
- A Scalable, Causal, and Energy Efficient Framework for Neural Decoding with Spiking Neural Networks (NeurIPS 2025 · similarity 0.88)
- Latent Diffusion for Neural Spiking Data (NeurIPS 2024 · similarity 0.87)
- Towards a "Universal Translator" for Neural Dynamics at Single-Cell, Single-Spike Resolution (NeurIPS 2024 · similarity 0.87)
- SPINT: Spatial Permutation-Invariant Neural Transformer for Consistent Intracortical Motor Decoding (NeurIPS 2025 · similarity 0.87)
- Population Transformer: Learning Population-level Representations of Neural Activity (ICLR 2025 · similarity 0.86)
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.
