It is exciting to think about how these seemingly small units of LA can potentially form parts of larger, more intricate systems in the future, and maybe change the way we interact with technology in various aspects of life and which problems we can solve by using AI.
Rebekka Olsson Omslandseter
Rebekka Olsson Omslandseter is defending her Ph.D. thesis: "On the Theory and Applications of Hierarchical Learning Automata and Object Migration Automata" 11 December 2023.
In the intriguing world of Artificial Intelligence (AI), particularly in the subset of Machine Learning (ML), there's a classic yet ever-evolving concept known as Learning Automata (LA). Imagine teaching a computer program to solve problems like a gardener cultivates a garden. Just as a gardener learns
which plants thrive under certain conditions and which do not, LA involves teaching non-human agents, like computers, to solve problems by learning from experience. They try different actions, much like a gardener experiment with different planting strategies, and they adopt the actions that yield the best results, akin to a gardener sticking with the most effective gardening techniques.
Rebekka Olsson Omslandseter, in her Ph.D. research, has made significant strides in this area. Her work focuses on enhancing the capabilities of LA, particularly in hierarchical structures and in the domain of Object Migration Automata (OMA), a specific type of LA algorithm.
OMA is particularly interesting because it is like a smart system that can organize and reorganize objects, even under highly changing conditions. These objects could be anything from users in a mobile network, to files in a database, to animals on a farm, or even shopping items in an online store. Imagine a system that can efficiently manage resources in a mobile network, ensuring optimal connectivity for all users by constantly adjusting to their movements and needs.
One of Rebekka and her team’s notable contributions is the development of the Hierarchical Discrete Pursuit Automaton (HDPA). This innovation significantly boosts the LA effectiveness in solving complex problems. During the Ph.D. work, they have also introduced a method called the Action Distribution Enhanced (ADE) approach, which makes these LA systems learn faster, reaching effective solutions in even fewer steps.
Moreover, Rebekka has expanded the application of LA to solve real-world problems, like optimizing groupings and power allocation in mobile radio communication systems using a novel OMA-based approach. This demonstrates the practical utility of the field of LA in handling different scenarios.
The dissertation summarizes the contributions to the field, but also lays out a roadmap for future researchers. It is exciting to think about how these seemingly small units of LA can potentially form parts of larger, more intricate systems in the future, and maybe change the way we interact with technology in various aspects of life and which problems we can solve by using AI.