Differential Evolution Optimization Save

A simple, bare bones, implementation of differential evolution optimization.

Project README

Differential Evolution (DE) Optimization with Python

A simple, bare bones, implementation of differential evolution optimization that accompanies a tutorial I made which can be found here: https://nathanrooy.github.io/posts/2017-08-27/simple-differential-evolution-with-python/

Installation

You can either download/clone this repo and use as is, or you can pip install it with the following command:

pip install git+https://github.com/nathanrooy/differential-evolution-optimization

Usage

Once you have completed the installation, usage is similar to that of other common optimization frameworks.

>>> from diffevo import de_simple

Next, you need to specify a cost fucntion. I included the sphere function for example purposes, but you'll probably end up using your own.

>>> from diffevo.cost_functions import sphere

Lastly we need to specify some optimizer specific constants. For more information on what these values are and how best to use them, see the linked blog post at the top of this page.

>>> bounds = [(-1,1),(-1,1)]            # bounds [(x1_min, x1_max), (x2_min, x2_max),...]
>>> popsize = 10                        # population size, must be >= 4
>>> mutate = 0.5                        # mutation factor [0,2]
>>> recombination = 0.7                 # recombination rate [0,1]
>>> maxiter = 20                        # max number of generations

Now, let's minimize this!

>>> de_simple.minimize(sphere, bounds, popsize, mutate, recombination, maxiter)

The output of which should look close to this:

...
GENERATION: 20
   > 2.6635635326983712e-05 [0.0018208872335391882, 0.004829079105763097]
   > 4.9580606970726694e-05 [0.0070407589280894346, 9.121780119916148e-05]
   > 2.2162273539063642e-05 [-0.003015611999457198, -0.003615018368942736]
   > 1.959970961017459e-05 [0.002787324055676138, 0.0034395543634057764]
   > 7.025341756207273e-05 [-0.0046121121033863045, -0.006998702701777687]
   > 3.5972219029065346e-05 [-0.0004056761521820749, -0.005983949021224704]
   > 2.4987227314433647e-05 [-0.002532393588388723, -0.0043097807401213965]
   > 3.358213318176877e-05 [0.0029016104781936065, -0.0050162525668676055]
   > 7.198803745771137e-06 [-0.00018143147319893155, 0.002676917325265015]
   > 1.577366384694402e-05 [0.0035989612612941646, -0.0016796254602285399]
      > GENERATION AVERAGE: 3.0574569012700424e-05
      > GENERATION BEST: 7.198803745771137e-06
         > BEST SOLUTION: [-0.00018143147319893155, 0.002676917325265015] 

Open Source Agenda is not affiliated with "Differential Evolution Optimization" Project. README Source: nathanrooy/differential-evolution-optimization

Open Source Agenda Badge

Open Source Agenda Rating