/TRL-HPO

Primary LanguagePython

Efficient Transformer-based Hyper-parameter Optimization for Resource-constrained IoT Environments

Tentative code: This code provides the implementations of Transformer-based Reinforcement Learning Hyper-parameter Optimization (TRL-HPO), which is the convergence of transformers and Actor-critic Reinforcement Learning. All the code documentation and variable definition mirrors the content of the manuscript published in IEEE Internet of Things Magazine (to be published). The arxiv file is as follows: I. Shaer, S. Nikan, and A. Shami, "Efficient Transformer-based Hyper-parameter Optimization for Resource-constrained IoT Environments, " arXiv preprint arXiv:2403.12237, 2024.

The link to the paper (arxiv): https://arxiv.org/abs/2403.12237

The functional scripts are as follows:

  1. Run run.py to train the model.
  2. Run analyze_results.py to evaluate the trained model.
  3. Run explainability_results.py to evaluate the trained model.
  4. Run flops_count.py to output the FLOPS of the model.

Requirements

The requirements are included in the requirements.txt file. To install the packages included in this file, use the following command: pip install -r requirements.txt

Contact-Info

Please feel free to contact me for any questions or research opportunities.