EDRS

Evolutionary Computation (EC) is a family of machine-learning methods for solving decision-making problems. Loosely inspired by natural evolution, EC methods have the distinctive capability of producing solutions to non-differentiable and ill-defined problems in a wide variety of areas, such as robotics, architecture, electronic circuit design, civil engineering, or te reverse engineering of biological gene regulatory networks, to mention a few. In addition, EC methods can be used as a meta-learning technique to define properties of neural networks, such as their architecture, learning meta-parameters, or new learning algorithms.

The course will give an introduction to the theory and inspiration behind evolutionary computation, provide foundations of the most popular methods, and describe specific algorithms for each method.

The course will be complemented by exercises that students can perform on their own to explore how the algorithms work. In addition, students will apply the methods to a research project either on a problem related to their own PhD research, or on a problem related to the design and control of embodied AI systems.

The exam will consist of the assessment of the project report. The research projects will be formulated so that the best ones could be submitted for publication to a conference or journal.

• Eiben, A. E. and Smith, J. E. (2003, 2015) Introduction to Evolutionary Computing. Berlin: Springer Verlag 
• Floreano, D. and Mattiussi, C. (2008, 2023) Bio-inspired Artificial Intelligence. Cambridge, MA: MIT Press. 
• Risi, S., Tang, Y., Ha, D. and Miikkulainen, R. (2025) Neuroevolution. Cambridge, MA: MIT Press. 
• Recent research articles will be provided during the lectures