Evolutionary optmisation algorithms
Devising evolutionary algorithms to solve optimisation problems with many-objectives.
Project description:
Many real world problems involve multiple objectives to be satisfied concurrently subject to some constraints. Solution to such problems involve a set of solutions each of which can’t be improved in one objective without deteriorating in another. We say these solutions are non-dominated with respect to one another. Evolutionary algorithms are particularly suitable for solving such problems due to the property of achieving the entire set of solution in one single run.
Problems involving more than three objectives are commonly known as Many-objective Optimization Problems. Since the number of non-dominated solutions increases exponentially with increase in the number of objectives, many-objective optimisation problems are challenging to solve. The objective is to devise an algorithm for solving many-objective optimisation problems.
Contributions:
In my undergraduate thesis, in collaboration with Siddhartha Shankar Das, I adopted the notion of fuzzy-dominance to solve many-objective optimisation problems. Later Siddhartha adopted the notion of reference points to develop a new algorithm that significantly improves the performance of our algorithm.
Publication/Preprints:
If you are interested in the fuzzy-dominance based algorithm, please refer to my undergraduate thesis and poster-
- Undergraduate Thesis
- Undergraduate Poster presentation, BUET: Many-Objective Evolutionary Approach using Fuzzy Dominance with Bidirectional Bias ( Distinguished Poster award, 2014)
If you are interested in the fuzzy-dominance and reference point based algorithm, please refer to the following paper-