<<<<<<< HEAD

🧠 Neuromorphic Decision Transformer (SNN-DT)

=======

Spiking Decision Transformer (SNN-DT)

631ccfaf032021291bf79dcf9b8aab028e5f4bd1

Local Plasticity, Phase-Coding, and Dendritic Routing for Low-Power Sequence Control <<<<<<< HEAD

631ccfaf032021291bf79dcf9b8aab028e5f4bd1

---

🌟 Overview

This repository contains the official PyTorch implementation of the Spiking Decision Transformer (SNN-DT), as presented in our flagship research paper:

"Spiking Decision Transformers: Local Plasticity, Phase-Coding, and Dendritic Routing for Low-Power Sequence Control" (Pandey & Biswas, 2025).

The SNN-DT architecture bridges the gap between the sequential modeling capabilities of dense Transformers and the extreme energy efficiency of Spiking Neural Networks (SNNs). By embedding Leaky Integrate-and-Fire (LIF) neurons within the block components, we secure state-of-the-art performance on continuous control tasks while reducing energy consumption by over four orders of magnitude ($\approx 40$ nJ).

---

✨ Core Neuromorphic Innovations

• 🕒 Phase-Coded Positional Encoding: Replaces generic float embeddings with rhythmic, spike phase-shifted encodings.
• 🌳 Dendritic-Style Routing MLP: Context-dependent gating coefficients dynamically prune attention heads without uniform averaging.
• 🧬 Three-Factor Local Plasticity: Elegantly implements STDP-like localized credit assignment rules, circumventing catastrophic unrolling.

Dynamic Dendritic Gating Coefficients

---

🚀 Quickstart & Installation

System Requirements: Linux/Windows, Python 3.8+, CUDA-enabled GPU (Recommended).

# Clone the repository
git clone https://github.com/Vishal-sys-code/neuromorphic_decision_transformer.git
cd neuromorphic_decision_transformer

# Install core dependencies natively
pip install -r requirements.txt

Tip

For deploying the documentation locally via Sphinx, execute make html inside the /docs directory!

---

💻 Experimental Workflows

1. Training the Architecture

Run the SNN-DT training pipeline, handling automated data orchestration and surrogate gradient optimization.

python snn-dt/scripts/train.py --model snn_dt --env "Pendulum-v1" --save-dir "results/snn_dt_pendulum"

Or execute the entire benchmarking suite across CartPole-v1, MountainCar-v0, and Acrobot-v1:

./run_all_experiments.sh

2. Neuromorphic Evaluation & Profiling

Evaluate checkpoint inference trajectories and monitor strict event-driven metrics (absolute spike outputs).

python eval_snn_dt.py \
    --env "Pendulum-v1" \
    --checkpoint_path "results/snn_dt_pendulum/best_model.pt" \
    --target_return -200

Note

The runtime reports normalized return against the expert policy baseline alongside the average hardware energy proxies (Spikes / timestep).

---

📖 Citation

If you build upon this architecture or framework, please consider citing our underlying research:

@article{pandey2025spiking,
  title={Spiking Decision Transformers: Local Plasticity, Phase-Coding, and Dendritic Routing for Low-Power Sequence Control},
  author={Pandey, Vishal and Biswas, Debasmita},
  journal={arXiv preprint arXiv:2508.21505},
  year={2025}
}

Maintained with ❤️ by the SNN-DT Research Team. Licensed under MIT.