Dask-PatternSearch

This project is in alpha phase. Please join the discussion and experiment with it, but be aware that it may eat your cat. Better documentation coming soon.

The Pattern

Installation Instructions

I recommend conda, and right now dask-patternsearch depends on the development version of distributed. In your conda environment, run:

conda install dask distributed -c conda-forge
pip install git+https://github.com/dask/dask.git --upgrade
pip install git+https://github.com/dask/distributed.git --upgrade
pip install git+https://github.com/eriknw/dask-patternsearch.git --upgrade

Python 2.7 and 3.4, 3.5, and 3.6 are tested.

Getting Started

Here is a simple example optimizing a trivial sphere function with minimum at 0.

>>> import numpy as np
>>> import distributed
>>> from dask_patternsearch import search

>>> def sphere(x):
...     return x.dot(x)

>>> client = distributed.Client()
>>> x0 = np.array([12.3, 4.5])
>>> stepsize = np.array([1, 1])
>>> best, results = search(sphere, x0, stepsize, client=client, stopratio=1e-4)
>>> best.point
array([ 0.,  0.])

Useful Links

http://www.cs.wm.edu/~va/research/sirev.pdf