mbahmani / openml-pimp

Parameter Importance according to OpenML

openml-pimp

This work aims to answer the questions:

• Which hyperparameters to optimize
• Over what ranges / priors

Short Video

(Explanation of the work in less than 3 minutes)

Data

All data in this paper can be reproduced using this notebook. Additionally, this folder contains arff files containing a large fraction of the meta-data generated for this project.

Dockerfile

For running all code in the best way, you need to install container which is inside Dockerfile: I want to explain what should you do step by step maybe that helps others. then run following code in the terminal where your Dockerfile is :

docker build . after some minutes the image file will download and your container should be installed for being sure: docker ps -a when you find it there is not! please run the following code

sudo docker run -d --name jupyter openml/jupyter-python

Now you should see your image: sudo docker ps -a

due to your Jupyter maybe run in a different port, it is a good idea to set it in your docker:

sudo docker stop jupyter
sudo docker rm jupyter

sudo docker run -d --name jupyter -p 8888:8888 openml/jupyter-python

Now you can import Openml in your Jupyter just run Jupyter with the following URL: 127.0.0.1:8888

Paper

Hyperparameter Importance Across Datasets by Jan N. van Rijn and Frank Hutter (ArXiv)

Bibtex

If you find this work useful, please cite:

@inproceedings{Rijn2018Hyperparameter,
  title        = {Hyperparameter Importance Across Datasets},
  author       = {van Rijn, Jan N. and Hutter, Frank},
  booktitle    = {Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining},
  year         = {2018},
  organization = {ACM},
  pages        = {2367--2376}
}