What is DeepHyper?

DeepHyper is an automated machine learning (AutoML) package for deep neural networks. It comprises two components: 1) Neural architecture search is an approach for automatically searching for high-performing the deep neural network search_space. 2) Hyperparameter search is an approach for automatically searching for high-performing hyperparameters for a given deep neural network. DeepHyper provides an infrastructure that targets experimental research in neural architecture and hyperparameter search methods, scalability, and portability across HPC systems. It comprises three modules: benchmarks, a collection of extensible and diverse benchmark problems; search, a set of search algorithms for neural architecture search and hyperparameter search; and evaluators, a common interface for evaluating hyperparameter configurations on HPC platforms.

Documentation

Deephyper documentation is on ReadTheDocs

Install instructions

From pip:

pip install deephyper

From github:

git clone https://github.com/deephyper/deephyper.git
cd deephyper/
pip install -e .

if you want to install deephyper with test and documentation packages:

# From Pypi
pip install 'deephyper[tests,docs]'

# From github
git clone https://github.com/deephyper/deephyper.git
cd deephyper/
pip install -e '.[tests,docs]'

Directory search_space

benchmark/
    a set of problems for hyperparameter or neural architecture search which the user can use to compare our different search algorithms or as examples to build their own problems.
evaluator/
    a set of objects which help to run search on different systems and for different cases such as quick and light experiments or long and heavy runs.
search/
    a set of algorithms for hyperparameter and neural architecture search. You will also find a modular way to define new search algorithms and specific sub modules for hyperparameter or neural architecture search.
hps/
        hyperparameter search applications
nas/
        neural architecture search applications

How do I learn more?

• Documentation: https://deephyper.readthedocs.io

• GitHub repository: https://github.com/deephyper/deephyper

Quickstart

Hyperparameter Search (HPS)

An example command line for HPS:

deephyper hps ambs --evaluator ray --problem deephyper.benchmark.hps.polynome2.Problem --run deephyper.benchmark.hps.polynome2.run --n-jobs 1

Neural Architecture Search (NAS)

An example command line for NAS:

deephyper nas ambs --evaluator ray --problem deephyper.benchmark.nas.polynome2Reg.Problem --n-jobs 1

Who is responsible?

Currently, the core DeepHyper team is at Argonne National Laboratory:

• Prasanna Balaprakash [email protected], Lead and founder
• Romain Egele [email protected]
• Misha Salim [email protected]
• Romit Maulik [email protected]
• Venkat Vishwanath [email protected]
• Stefan Wild [email protected]

Modules, patches (code, documentation, etc.) contributed by:

• Elise Jennings [email protected]
• Dipendra Kumar Jha [email protected]

Citing DeepHyper

If you are referencing DeepHyper in a publication, please cite the following papers:

• P. Balaprakash, M. Salim, T. Uram, V. Vishwanath, and S. M. Wild. DeepHyper: Asynchronous Hyperparameter Search for Deep Neural Networks. In 25th IEEE International Conference on High Performance Computing, Data, and Analytics. IEEE, 2018.

• P. Balaprakash, R. Egele, M. Salim, S. Wild, V. Vishwanath, F. Xia, T. Brettin, and R. Stevens. Scalable reinforcement-learning-based neural architecture search for cancer deep learning research. In SC ’19: IEEE/ACM International Conference on High Performance Computing, Network-ing, Storage and Analysis, 2019.

How can I participate?

Questions, comments, feature requests, bug reports, etc. can be directed to:

• Issues on GitHub

Patches through pull requests are much appreciated on the software itself as well as documentation. Optionally, please include in your first patch a credit for yourself in the list above.

The DeepHyper Team uses git-flow to organize the development: Git-Flow cheatsheet. For tests we are using: Pytest.

Acknowledgements

• Scalable Data-Efficient Learning for Scientific Domains, U.S. Department of Energy 2018 Early Career Award funded by the Advanced Scientific Computing Research program within the DOE Office of Science (2018--Present)
• Argonne Leadership Computing Facility: This research used resources of the Argonne Leadership Computing Facility, which is a DOE Office of Science User Facility supported under Contract DE-AC02-06CH11357.
• SLIK-D: Scalable Machine Learning Infrastructures for Knowledge Discovery, Argonne Computing, Environment and Life Sciences (CELS) Laboratory Directed Research and Development (LDRD) Program (2016--2018)

Copyright and license

Copyright © 2019, UChicago Argonne, LLC

DeepHyper is distributed under the terms of BSD License. See LICENSE

Argonne Patent & Intellectual Property File Number: SF-19-007