Tansaku

Introduction

Tansaku allows the user to implement population-based optimizers.

Modules

Tansaku: Tako consists of several types of functioanlity for implementing optimization algorithms

• perturbation - Functionality for changing the values in an individual or population

• reduction - Functionality for converting a population to an individual

• selection - Functionality for selecting individuals to use in breeding

• division - Functionality for dividing a population into multiple populations

• assessment - Functionality for assessing a population or Individual

• … and so on.

Key Features and Functions

Tansaku can be used to implement a variety of metaheuristics and optimization algorithms. Below are just a couple of the possibilities.

• Tansaku Example 1: The below is an example of how to use a Tako in hill climbing .

  from zenkai.tansaku import Populator, Individual
  from zenkai.utils import set_model_parameters, get_model_parameters
  
  # create an individual and then repeat the values k times to construct a population of clones
  population = Individual(x=get_model_parameters(self.model)).populate(k=4)
  population = self.perturb(populator)
  population = self.assessor(population)
  # choose the best candidate
  individual = self.reducer(population)
  # update the parameters of the model with the new parameters
  set_model_parameters(self.model, individual['x'])
  

• Tansaku Example 2: It can also be used as below for genetic algorithms.

  # You can use Tansaku to do hill climbing and test different hypotheses
  from zenkai.tansaku import Populator, Individual
  
  # mutate the population
  population = self.mapper(population)
  children1, children2 = self.divider(populator)
  population = self.mixer(children1, children2)
  # generate the next generation
  population = self.assessor(population)