DLOPT: Deep Learning Optimization

Python library for artificial neural network (NN) optimization.

Index

This respository contains several deliverables associated to the main topic (NN optimization). The main structure is as follows:

• data: datasets used in publications.
• dlopt: main library
• docs: miscellaneous documents
• etc: analysis, useful scripts, and other stuff.
• examples: a quick guide on using DLOPT
• publications: stand alone deliverables related to scientific publications. Please refer to Publications.

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.

Prerequisites

The main folder contains a requirements.txt file listing the Python packages needed to tun the code.

Installing

Clone the repository on your local machine, install the dependencies (you may want to use a virtual environment), run the setup.py file:

python setup.py install

and have fun!

Building

To build a python library, you may use the setup.py file:

python setup.py build

To build a binary application you could use pyinstaller. Remember to add the path to required libraries and any hidden import, for example:

pyinstaller optimizer.py -F --path ../env/lib/python3.5/ --hidden-import algorithms

How to cite DLOPT

We encourage authors of scientific papers including results generated using DLOPT to cite:

• Camero, A., Toutouh, J., and Alba, E. DLOPT: Deep Learning Optimization Library. arXiv preprint arXiv:1807.03523 (july 2018)

Related Publications

• Camero, A., Toutouh, J., Stolfi, D.H., and Alba, E. Evolutionary Deep Learning for Car Park Occupancy Prediction in Smart Cities. To appear in Proc. of Learning and Intelligent OptimizatioN Conference (LION 12). 2018.
• Camero, A., Toutouh, J., and Alba, E. Low-cost recurrent neural network expected performance evaluation. arXiv preprint arXiv:1805.07159 (may 2018)
• Camero, A., Toutouh, J., and Alba, E. DLOPT: Deep Learning Optimization Library. arXiv preprint arXiv:1807.03523 (july 2018)
• Camero, A., Toutouh, J., and Alba, E. Comparing Deep Recurrent Networks Based on the MAE Random Sampling, a First Approach. To appear in Conference of the Spanish Association for Artificial Intelligence, CAEPIA, 2018.
• Camero A, Toutouh J, Alba E. A specialized evolutionary strategy using mean absolute error random sampling to design recurrent neural networks. arXiv preprint arXiv:1909.02425. 2019 Sep 4.
• Camero A, Toutouh J, Ferrer J, Alba E. Waste generation prediction in smart cities through deep neuroevolution. InIbero-American Congress on Information Management and Big Data 2018 Sep 26 (pp. 192-204). Springer, Cham.
• Camero A, Wang H, Alba E, Bäck T. Bayesian Neural Architecture Search using A Training-Free Performance Metric. arXiv preprint arXiv:2001.10726. 2020 Jan 29.
• Camero A, Toutouh J, Ferrer J, Alba E. Waste generation prediction under uncertainty in smart cities through deep neuroevolution. Revista Facultad de Ingeniería Universidad de Antioquia. 2019 Dec(93):128-38.

Contributing

Please read CONTRIBUTING.md for details.

Authors

• Andrés Camero - Initial work - dlopt

Please see the list of scientific publications for more information about people who has participated in this project or visit NEO research webpage.

License

This project is licensed under the GNU GPL v3 license - see the LICENSE.md file for details.

Acknowledgments

This research was partially funded by Ministerio de Economı́a, Industria y Competitividad, Gobierno de España, and European Regional Development Fund grant numbers:

• MoveOn TIN2014-57341-R
• Cirti TIN2016-81766-REDT
• 6City TIN2017-88213-R
• PRECOG UMA18-FEDERJA-003