Neural Decoder

Welcome to the BrainStorm 2026 Brain-Computer Interface (BCI) Hackathon! Build accurate, fast, and lightweight neural decoders for real-time auditory stimulus classification from ECoG recordings.

Quick Start

# 1. Install dependencies
make install
source .venv/bin/activate

# 2. Download data
uv run python -c "from brainstorm.download import download_train_validation_data; download_train_validation_data()"

# 3. Train and evaluate
uv run python examples/example_local_train_and_evaluate.py

Documentation

Getting Started

1. Overview - Understand the problem, constraints, and scoring
2. Installation - Set up your environment
3. Dataset - Learn about the ECoG data format

Building Your Solution

4. Defining a Model - Create custom models
5. Evaluation - Test locally and understand scoring
6. Submissions - Submit for test set evaluation
7. FAQ - Common questions

Project Structure

brainstorm2026-track1/
├── brainstorm/           # Core library
│   ├── ml/              # Model implementations
│   ├── loading.py       # Data loading
│   ├── spatial.py       # Spatial utilities
│   └── plotting.py      # Visualization
├── docs/                # Documentation
├── examples/            # Example scripts
└── tests/               # Test suite

Rules

✅ Allowed: Any Python libraries, ensemble models, AI coding tools, pre-trained models

❌ Not Allowed: Non-causal models, modifying evaluation code, models >25MB

See the FAQ for detailed rules.

---

Good luck, and happy hacking! 🧠⚡