Galaxy-based Search Algorithm
- Source: Software project as part of Natural Computing lecture
- Type: Group student project (of 2 students)
- Language(s): Python
In this lecture project for the Natural Computing lecture at the University of Edinburgh, we implemented the Galaxy-based Search Algorithm (GbSA) and Particle Swarm Optimisation (PSO) as a baseline for PCA approximation. The goal of the project was to evaluate and compare the GbSA algorithm to the metaheuristic optimisation algorithm of PSO.
Therefore, we evaluted the PCA approximation of both algorithms for three datasets, compared them to the exact PCA solution and analysed the performance. We found severe limitations of the GbSA algorithm as well as inconsistencies in its foundational research paper. Following, we proposed adjustments to the algorithm and showed their positive impact on performance in an evaluation.