rameziophobia / travelling_salesman_optimization

Python implementation for TSP using Genetic Algorithms, Simulated Annealing, PSO (Particle Swarm Optimization), Dynamic Programming, Brute Force, Greedy and Divide and Conquer

Solving the Travelling Salesman Problem in Python

Implemented techniques

• Genetic Algorithm
• Simulated Annealing
• PSO: Particle Swarm Optimization
• Divide and conquer
• Dynamic Programming
• Greedy
• Brute Force

When the solution is found it is plotted using Matplotlib and for some algorithms you can see the intermediate results

Example visualizations

The following visualizations are all applied on the EIL51 dataset available through the TSP online library

Genetic Algorithm

Divide and Conquer

Greedy

PSO: Particle Swarm Optimization

travelling_salesman_optimization's Issues

Inquiry about PSO code

Hi, I have a question about how you implemented particle swarm optimization. In pso.py, the section of code that updates the new routes has a bit of code that I'm not sure of the function of.

`

            for i in range(len(self.cities)):
                if new_route[i] != particle.pbest[i]:
                    swap = (i, particle.pbest.index(new_route[i]), self.pbest_probability)
                    temp_velocity.append(swap)
                    new_route[swap[0]], new_route[swap[1]] = \
                        new_route[swap[1]], new_route[swap[0]]

            for i in range(len(self.cities)):
                if new_route[i] != gbest[i]:
                    swap = (i, gbest.index(new_route[i]), self.gbest_probability)
                    temp_velocity.append(swap)
                    gbest[swap[0]], gbest[swap[1]] = gbest[swap[1]], gbest[swap[0]]

            particle.velocity = temp_velocity

            for swap in temp_velocity:
                if random.random() <= swap[2]:
                    new_route[swap[0]], new_route[swap[1]] = \
                        new_route[swap[1]], new_route[swap[0]]

            particle.route = new_route
            particle.update_costs_and_pbest()`

In the second for loop, what does gbest[swap[0]], gbest[swap[1]] = gbest[swap[1]], gbest[swap[0]] do? gbest doesn't seem to get used as it just gets replaced. It seems like you implemented pso by first making the new route the same as the personal best, then made the global best the same as the new route, and then returned the new route back to what it was originally based on a given probability. The first and third for loops make sense, but the second one is a bit confusing. I ran the code after commenting out gbest[swap[0]], gbest[swap[1]] = gbest[swap[1]], gbest[swap[0]] and it seemed to run the same. Could you clarify what that line of code is for? Thank you for your time.