|
--- |
|
base_model: |
|
- autogluon/chronos-t5-large |
|
tags: |
|
- General-Purpose |
|
- Brain |
|
- Foundation Models |
|
- electroencephalogram |
|
- signal processing |
|
--- |
|
General-Purpose Brain Foundation Models for Time-Series Neuroimaging Data |
|
|
|
[[General-Purpose Brain Foundation Models for Time-Series Neuroimaging Data](https://openreview.net/forum?id=HwDQH0r37I)] |
|
## Model Details |
|
|
|
### Model Description |
|
|
|
<!-- Provide a longer summary of what this model is. --> |
|
|
|
- **Model type:** T5-large |
|
- **License:** [More Information Needed] |
|
- **Finetuned from model:** Chronos-t5-large |
|
|
|
## How to Get Started with the Model |
|
|
|
Here is [colab](https://colab.research.google.com/drive/1Dy6oUJeYuoi0cAo9imlfA2gZgueMun9h?usp=sharing) notebook for inference of the Moabb dataset. |
|
|
|
### Training Data: |
|
The training data used was the NMT EEG dataset. NMT is an open-source, annotated dataset comprising healthy and pathological EEG recordings. It consists of 2,417 recordings from unique participants, providing multichannel EEG data along with labels indicating the participants' pathological state, classified as normal or abnormal. Each EEG channel is treated as an independent time series, which is further divided into two segments: a context window for conditioning and a prediction target window. |
|
|
|
Hyperparameters are the same as those used in the Chronos paper. |
