Skip to main content

Random search optimization and experiment logging. Support async, visualization, distributed execution.

Project description


Build Status PyPI version

randopt is a Python package for machine learning experiment management, hyper-parameter optimization, and results visualization. Some of its features include:

  • result logging and management,
  • human-readable format,
  • support for parallelism / distributed / asynchronous experiments,
  • command-line and programmatic API,
  • shareable, flexible Web visualization,
  • automatic hyper-parameter search, and
  • pure Python - no dependencies !

Installation

pip install randopt

Usage

import randopt as ro

def loss(x):
    return x**2

e = ro.Experiment('myexp', {
        'alpha': ro.Gaussian(mean=0.0, std=1.0, dtype='float'),
    })

# Sampling parameters
for i in xrange(100):
    e.sample('alpha')
    res = loss(e.alpha)
    print('Result: ', res)
    e.add_result(res)

# Manually setting parameters
e.alpha = 0.00001
res = loss(e.alpha)
e.add_result(res)

# Search over all experiments results, including ones from previous runs
opt = e.minimum()
print('Best result: ', opt.result, ' with params: ', opt.params)

Results Visualization

Once you obtained some results, run roviz.py path/to/experiment/folder to visualize them in your web browser.

For more info on visualization and roviz.py, refer to the Visualizing Results tutorial.

Hyper-Parameter Optimization

To generate results and search for good hyper-parameters you can either user ropt.py or write your own optimizaiton script using the Evolutionary and GridSearch classes.

For more info on hyper-parameter optimization, refer to the Optimizing Hyperparams tutorial.

Documentation

For more examples, tutorials, and documentation refer to the wiki.

Contributing

To contribute to Randopt, it is recommended to follow the contribution guidelines.

Acknowledgements

Randopt is maintained by Séb Arnold, with numerous contributions from the following persons.

  • Noel Trivedi
  • Cyrus Jia
  • Daler Asrorov

License

Randopt is released under the Apache 2 License. For more information, refer to the LICENSE file.

I would love to hear how your use Randopt. Feel free to drop me a line !

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for randopt, version 0.2.4
Filename, size File type Python version Upload date Hashes
Filename, size randopt-0.2.4.tar.gz (21.7 kB) File type Source Python version None Upload date Hashes View hashes

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN SignalFx SignalFx Supporter DigiCert DigiCert EV certificate StatusPage StatusPage Status page