Carlotta Langer

1. 

   Information Flows in Embodied Agents

   The interaction between body, brain and environment of an embodied agent can be modeled by the sensorimotor loop. Using information-theoretic measures we are able to quantify the information flow among the different parts of this system, including measures realted to Morphological Computation and Integrated Information.
   
   
   

   Complexity Measures

   Complexity in itself is not easy to define. The information-geometric complexity measures I am interested in consider a system to be complex if the whole is more than the sum of its parts. This approach makes use of concepts such as stochastic processes, graphical models and conditional independence relations.
   
   
   

   Information-Geometric Algorithms

   The information-geometric algorithms called iterative scaling and em-algorithm can be used to approximate the maximum likelihood estimation in different settings. The latter can also be applied as a learning algorithm for embodied agents. I am interested in studying the mathematical properties of these algorithms.

1. 

   How morphological computation shapes integrated information in embodied agents

   Presented at the Models of Consciousness Conference 2
   (online, 27. September 2021)

   YouTube


2. 

   Learning to predict requires Integrated Information

   Presented at the Models of Consciousness Conference 3
   (Stanford, 6. September 2022)

   YouTube

1. Outsourcing Control requires Control Complexity

   Langer, C. and N. Ay

   ArXiv Preprint, 2023.

   ArXiv Code

Published