Improving Hyperparameter Optimization By Planning Ahead

We provide here the source code for our paper: [Dataset2Vec: Learning Dataset Meta-Features](We provide here the source code for our paper: Improving Hyperparameter Optimization By Planning Ahead. ).

Usage

To meta-train the joint surrogate model, run the run-meta.py file.

python run-meta.py 

Use the weights with the best validation performance to initialize the surrogate for hyperparameter optimization.

python test-pets.py

Citing LookAhead-MPC

---

To cite LookAhead-MPC please reference our arXiv paper:

@article{jomaa2021improving,
  title={Improving Hyperparameter Optimization by Planning Ahead},
  author={Jomaa, Hadi S and Falkner, Jonas and Schmidt-Thieme, Lars},
  journal={arXiv preprint arXiv:2110.08028},
  year={2021}
}