Parallel Genetic Algorithms' Implementation Using a Scalable Concurrent Operation in Python.

Area of Science:

• Computer Science
• Artificial Intelligence
• Computational Optimization

Background:

• Genetic algorithms (GAs) are powerful optimization tools but can be computationally intensive.
• Parallelization offers a way to accelerate GA execution by distributing computational tasks.
• Evaluating different parallelization models and Python libraries is crucial for efficient implementation.

Purpose of the Study:

• To implement and compare three parallel genetic algorithm (GA) models: Master-Slave, Coarse-Grained, and Fine-Grained.
• To investigate the suitability of Python parallelization modules, selecting Scalable Concurrent Operation in Python (SCOOP).
• To analyze the hardware utilization and performance of these parallel GA models.

Main Methods:

• Implementation of Master-Slave, Coarse-Grained, and Fine-Grained GA models using Python.
• Selection and integration of the Scalable Concurrent Operation in Python (SCOOP) library.
• Comparison of parallel GA models against a serial GA model.
• Performance evaluation focusing on hardware utilization and communication within the GA topology.

Main Results:

• The Scalable Concurrent Operation in Python (SCOOP) was identified as a favorable module for parallel GA implementation.
• The study provides a comparative analysis of hardware utilization across the implemented parallel GA models.
• Performance variations were observed based on the chosen parallelization strategy and underlying hardware.

Conclusions:

• Parallel genetic algorithms can significantly enhance computational efficiency.
• The choice of parallelization model (Master-Slave, Coarse-Grained, Fine-Grained) impacts hardware utilization and performance.
• SCOOP offers a viable framework for developing efficient parallel genetic algorithms in Python.