We propose meta-control of social learning strategies, referred as meta-social learning, to provide effective and sample-efficient learning as a population in multi-agent systems. We investigate models that can arbitrate between individual, success-based and conformist strategies depending on the environment.
04.2020 - 06.2021
We investigate the effect of local information exchange on distributed embodied evolution. More specifically, we consider a network of agents distributed at fixed locations within a certain environment. We assume that each agent is required to perform a certain local behavior, that is described by some agent-local parameters and optimized using distributed embodied evolution approach.
01.2020 - 04.2020
The plasticity property is one of the fundamental property of Biological Neural Networks which allows them to change their configuration over lifetime. In this project, we aim to produce plasticity property in Artificial Neural Networks to allow lifetime leaning inspired by the Biological Neural Networks.
10.2015 - 12.2019
The parameters of evolutionary algorithms (i.e. type of evolutionary operator and their parameters) influence the behavior of the search process. We investigate ways of optimizing the parameters of the algorithms to perform an efficient search.
01.2016 - 10.2016