Evolving to be flexible – optimizing task-dependent information processing in the visual system

Udo Ernst – Universität Bremen
Andreas Kreiter – Universität Bremen

Highly evolved brains typically develop in complex environments that provide ecological niches for species with a wide range of cognitive abilities and behaviours. The high complexity of the environment, as well as the need to function consistently and robustly in the face of high environmental variability, places a variety of demands on neural processing. The optimisation of specialised neural pathways and networks for each possible task must quickly have reached a limit due to the available brain mass alone. An evolutionary principle and way out of the dilemma would be the development of ever greater flexibility in reconfiguring existing networks into functionally varying circuits. This implies a co-evolution of function and flexibility. Our project goal is to investigate the hypothesis that neural systems continuously optimise their ability to operate flexibly while processing the current task robustly against random variability and irrelevant competing stimuli in sensory signals. Since information processing in the brain is realized by the interaction of different network structures, the coordination of flexibility must also be done in parallel and a task must be decomposed into appropriate control signals for the individual players. Which network configurations are optimal for flexible processing, and how do they arise under given physiological constraints? Do multiple solutions exist for given flexibility problems, and can these explain the variability observed in the experiment? What is the functional significance of routing mechanisms based on the synchronisation of neuronal oscillations compared to mechanisms based on neuronal "avalanches" of spontaneous synchronisation - and what is the possible evolutionary relationship between them? To answer these fundamental questions, we will study optimisation of flexibility in theory and experiment with a focus on the visual system. Theoretical approaches will formally describe central processing aspects such as parallel coordination and the evolution of flexible circuits, considering co-evolution of flexibility and function as an optimisation problem with constraints. Modelling and experimentation will investigate selective attention as a central aspect of flexibility in the visual system, involving task-dependent coordination of multiple visual areas. Our experiments will characterise and compare potential mechanisms of neural flexibility and their robustness to sensory variability. Biophysically realistic modelling will accompany the experiments to critically test formal theories and identify mechanisms of flexibility and optimisation, as well as their control and coordination.