A python3 package implementing randomized algorithms for Travelling Salesman Problem. The implementations are based off of A first course in Artificial Intelligence: Deepak Khemani.
The algorithms implemented include:
You can install this using pip
. Just run
pip install randomized_tsp
See example here.
If you have n
cities in your problem then you need to define the distances between cities using a two dimensional n X n
list.
distances[i][j]
should give the distance between city i and city j.
from randomized_tsp.tsp import tsp
# If you have 3 cities then you need to define this list
# which gives the distances between two cities
distances = [[0, 4, 2], [4, 0, 3], [2, 3, 0]]
tsp_obj = tsp(distances)
# To run genetic algorithm
tour, cost = tsp_obj.genetic_algorithm()
# To run simulated annealing
tour, cost = tsp_obj.simulated_annealing()
# To run ant colony optimization
tour, cost = tsp_obj.ant_colony()
All the three algorithms return that best tour found and the cost of that tour. Tour is represented using path representation.
Some optional parameters are also available for the above algorithms.
def genetic_algorithm(self,
population_size=50,
mutation_prob=0.1,
crossover='order'):
"""
:param population_size: Defines the size of the population used in the algorithm
:param mutation_prob: Probability that a offspring will mutate
:param crossover: Defines the crossover operator, currently two options are available
`order` and `cycle`.
:return: Returns the best tour found and the cost of that tour
A tour is represented using path representation
"""
def simulated_annealing(self):
"""
:return: Returns the best tour found and the cost of that tour
A tour is represented using path representation
"""
def ant_colony(self,
num_of_ants=20,
pheromone_evapouration=0.2):
"""
:param num_of_ants: Number of ants in the colony
:param pheromone_evapouration: Evapouration rate of the pheromone deposited by ants
:return: Returns the best tour found and the cost of that tour
A tour is represented using path representation
"""
Many of these algorithms need improvements and optimizations. If you want to improve this project or fix a bug then all contributions are welcome.
If you find any bugs then open a issue on this repository. You can also contact me on gitter.