Ant Colony Optimization in CUDA and Python

This repository contains two implementations of the Ant Colony Optimization algorithm

Visualization

This image represents a graph adjacency matrix. We can cleary see that the pheromone concentrates in a specific path (one square per line)

Implementations

Python

Install dependencies: pip3 install -r python_implementation/requirements.txt --user

Run: python3 python_implementation/main.py

Usage

python3 python_implementation/main.py -h

usage: Ant Colony Optimization - Longest Path [-h] [-i INPUT] [-p POP_SIZE] [-a ALPHA] [-b BETA] [-e EVAPORATION] [-m MAX_ITERATIONS]

optional arguments:
  -h, --help            show this help message and exit
  -i INPUT, --input INPUT
                        Input graph
  -p POP_SIZE, --pop-size POP_SIZE
                        Number of ants
  -a ALPHA, --alpha ALPHA
                        Pheromone weight
  -b BETA, --beta BETA  Desirability weight
  -e EVAPORATION, --evaporation EVAPORATION
                        Pheromone evaporation
  -m MAX_ITERATIONS, --max-iterations MAX_ITERATIONS
                        Max Iterations

CUDA

Install dependencies: CUDA

Compile: make

Run:

• mkdir -p results/
• ./ACO_red bases_grafos/entrada1.txt 100 50 0.2 1 2 test_exp

Usage

./ACO <input database> <N_ITER> <N_ANTS> <EVAPORATION RATE> <ALPHA> <BETA> [exp_name]

Positional Arguments:
  input database    - Input graph
  N_ITER            - Max Iterations
  N_ANTS            - Number of ants
  EVAPORATION RATE  - Pheromone evaporation
  ALPHA             - Pheromone weight
  BETA              - Desirability weight
  exp_name          - [OPTIONAL] Save results in file results/[exp_name].txt