Skip to main content
Join the official 2020 Python Developers SurveyStart the survey!

A tree-based parallel surrogate optimization algorithm for optimizing noisy expensive functions

Project description

ProSRS algorithm

Progressive Stochastic Response Surface (ProSRS) is a parallel surrogate-based optimization algorithm for optimizing noisy expensive functions. This algorithm utilizes a radial basis function (RBF) as the surrogate, and adopts stochastic response surface (SRS) framework to balance exploitation and exploration. Compared to the original parallel SRS work, the novelties of this algorithm include

  • Introducing a new tree-based technique, known as the "zoom strategy", for efficiency improvement.
  • Extending the original work to the noisy setting (i.e., an objective function corrupted with random noise) through the development of a radial basis regression procedure.
  • Introducing weighting to the regression for exploitation enhancement.
  • Implementing a new SRS that combines the two types of candidate points that were originally proposed in the SRS work.

ProSRS algorithm is configured in a master-worker structure, where in each optimization iteration, the algorithm (master) constructs a RBF surrogate using the available evaluations, then proposes new points based on the constructed RBF, and finally distributes the tasks of evaluating these points to parallel processes (workers).

Compared to the popular Bayesian optimization algorithms, ProSRS is able to achieve faster convergence on some difficult benchmark problems, and is orders of magnitude cheaper to run. Moreover, ProSRS enjoys asymptotic convergence gaurantees. The common applications of this algorithm include efficient hyperparamter tuning of machine learning models and characterizing expensive simulation models.

version license language platform


Python dependencies:

To install prosrs package, type and run the following:

pip install prosrs

Note: The above pip method should work for most users. If a user encounters any installation problems including import errors or warnings, please refer to the Wiki page for possible solutions.

Getting started

After having successfully installed the prosrs package, users are ready to use ProSRS algorithm for solving optimization problems. The easiest way of getting started is to read the tutorials in the examples directory, where different usages, from the basic level to the advanced level, are demonstrated through code examples. Of course, users are also encouraged to check out the source codes of the algorithm in the prosrs directory.

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 prosrs, version 1.1.0
Filename, size File type Python version Upload date Hashes
Filename, size prosrs-1.1.0-py3-none-any.whl (37.2 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size prosrs-1.1.0.tar.gz (33.4 kB) File type Source Python version None Upload date Hashes View

Supported by

Pingdom Pingdom Monitoring Google Google Object Storage and Download Analytics Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page