IntEvoBCI: A Neuro-Symbolic Framework for Evolving Human Intention from EEG Signals

A Hybrid Architecture Fusing Deep Memory and Dynamic Neural Fields for Intention Decoding.

Abstract

TBA

Keywords: Brain-computer interface, neuro-symbolic framework, intention decoding, EEG signals, cognitive simulation, drift-diffusion frameworks, uncertainty awareness, deliberation dynamics.

1. Introduction

TBA

🧠 Key Features

• Hybrid Core: Combines a 1D-CNN (Feature Extraction), GRU (Temporal Memory), and Neural Fields (Decision Dynamics).
• Interpretable Dynamics: Visualizes the "Thinking Process" via attractors, momentum, and entropy.
• Dynamic Interaction: Learns context-aware inhibition matrices ($\beta_{ij}$) to resolve conflicting intentions.
• Safety First: Optimizes for Expected Calibration Error (ECE), ensuring the system "knows when it doesn't know."

📂 Repository Structure

IntEvoBCI/
├── data/               # EEG Data (GDF format)
├── experiments/        # Execution scripts
│   ├── train.py        # Main training loop
│   ├── evaluate.py     # Cognitive evaluation
│   ├── ablation.py     # Component analysis
│   └── make_paper.py   # Asset generator
├── models/             # Pre-trained checkpoints (ignored)
├── results/
│   └── paper/          # Final tables and plots
├── src/
│   ├── cognitive_model.py     # IntEvoBCI Architecture
│   ├── cognitive_analysis.py  # Metrics (Stability, Entropy)
│   ├── cognitive_viz.py       # Plotting utilities
│   ├── baseline_cnn.py        # Comparison model
│   └── features.py            # CSP & Filterbank
└── requirements.txt

🚀 Getting Started

Prerequisites

• Python 3.8+
• PyTorch 2.0+

Installation

git clone https://github.com/your-username/IntEvoBCI.git
cd IntEvoBCI
pip install -r requirements.txt

🔬 Reproduction

1. Train the Model

   python experiments/train.py

   Trains the Advanced Cognitive BCI on Subject 1 (default).

2. Run Cognitive Evaluation

   python experiments/evaluate.py

   Generates "Thinking Plots" and computes latency/reversal metrics.

3. Replicate Paper Results

   python experiments/benchmark.py  # Compare vs CNN
   python experiments/ablation.py   # Run Ablation Study
   python experiments/make_paper.py # Generate Tables & JSON

📊 Results Summary

Model	Accuracy	ECE (Calibration)	Interpretability
Baseline CNN	~67%	0.31	Low
IntEvoBCI	~57%	0.22	High

IntEvoBCI offers a safer, more transparent alternative for real-world neurotechnology.

📜 Citation

If you use this code, please cite our paper:

IntEvoBCI: A Neuro-Symbolic Framework for Evolving Human Intention from EEG Signals. (2026).