Paddy: Evolutionary Optimization Algorithm for Chemical Systems and Spaces

Paddy is an evolutionary optimization platform, writen in Python. Our preprint with benchmarking and cheminformatic experimentation can be found on arXiv.

The GitHub repo for computational experiments used in the preprint can be found at: https://github.com/chopralab/Paddy_Manuscript_Repo

Installation

Paddy can be installed from PyPI with the command:

pip install pip

Example

To run paddy, one simply needs to define their evaluation function, and parameter space.

>>> import paddy
>>> def parabola(input):
...     x = input[0][0]
...     y = input[1][0]
...     return(((x**2)/7)-((y**2)/2)+1)# The maximum is when x and y are 0
...
>>> # now we need to set our parameter space for x and y
>>> x_param = paddy.PaddyParameter(param_range=[-5,5,.2],
...                                param_type='continuous',
...                                limits=None, gaussian='scaled',
...                                normalization = False)
...
>>> y_param = paddy.PaddyParameter(param_range=[-7,3,.2],
...                                param_type='continuous',
...                                limits=None, gaussian='scaled',
...                                normalization = False)
...
>>> # now we make a class with the parameter spaces we defined
>>> class paraboloid_space(object):
...     def __init__(self):
...         self.xp = x_param
...         self.yp = y_param
...
>>> # now we need to initialize a `PFARunner`
>>> example_space = paraboloid_space() #the space parameter
>>> example_runner = paddy.PFARunner(space=example_space,
...                                  eval_func=parabola,
...                                  paddy_type='population',
...                                  rand_seed_number = 20,
...                                  yt = 10,
...                                  Qmax = 5,
...                                  r=.2,
...                                  iterations = 5)
...
>>> example_runner.run_paddy(file_name='paddy_example')
paddy is done!

Saved instances of Paddy can be extended, even once the initial code has be run.

>>> import paddy
>>> # we need to define the dependent evaluation function
>>> def parabola(input):
...     x = input[0][0]
...     y = input[1][0]
...     return(((-x**2)/7)-((y**2)/2)+1)# The maximum is when x and y are 0
...
>>> recovered_example = paddy.paddy_recover('paddy_example')
>>> recovered_example.extend_paddy(5)
paddy is done!

Visualizing results is made simple with built in plotting capabilities.

>>> example_runner.paddy_plot_and_print(('scatter',
...                                      'average_gen',
...                                      'average_population'))
...

Full documentation of the code is available at: https://chopralab.github.io/paddy/index.html