Demian Battaglia

Aix-Marseille University, France

Neuronal dynamics and dynamics of information

Abstract

The dynamics of neural circuits give rise to rich patterns of coordinated activity, such as changing spatiotemporal patterns of synchronisation. This self-organized formation and reconfiguration of coordinated activity patterns give necessarily rise to emergent information processing, with specific processing operations coming into act and being applied to the information carried by the activity of system’s units. Information theory provides tools able to decompose any ongoing, unknown, information processing into elementary operations such as “storage”, “copy”, “transfer” or “modification”. These primitive information manipulation operations could be seen as part of a neuronal assembly language out of which more complex, arbitrary computations could emerge. Ideally, one could thus hope to establish a mapping between, on one side, patterns of coordinated activity (building blocks of high-dimensional neural dynamics) and, on the other, specific blends of primitive processing operations (building blocks of computations). We will illustrate here various examples of the “algorithmic effects” of complex neural network dynamics. First, we will analyze the simulated activity of model circuits involving coupled neuronal populations measuring quantitatively the dynamic occurrence of storage, transfer and modification informational operations and showing how changes of dynamical state result in changes of information processing and how the emulation of different cognitive functions give rise to different combinations of primitive processing operations. Second, we will analyze actual rodent electrophysiological recordings in rodents during anesthesia and sleep, chasing for information processing states (IPSs) and a structured dynamics of them. We will find not only that well-defined IPSs exist and can be extracted, but also that they form rich sequences whose degree of organization is modulated by brain state or pathology, with structured complexity being replaced, in part, by increased randomness in epilepsy. Although preliminary, our results suggest the feasibility of analyses at an intermediate, algorithmic level, between structure and actual functions –that may well be unknown–, in which the plainness and possible universality of the considered processing operations make possible to tightly map dynamic patterns and the primitive computations they produce.

References

  • Clawson, W., Madec, T., Ghestem, A., Quilichini, P.P., Battaglia, D., Bernard, C. (2021). Disordered information processing dynamics in experimental epilepsy. bioRxiv.
  • Pedreschi N, Bernard C, Clawson W, Quilichini P, Barrat A, Battaglia D (2020). Dynamic core-periphery structure of information sharing networks in entorhinal cortex and hippocampus. Netw Neurosci 4: 946–975.
  • Clawson W, Vicente AF, Ferraris M, Bernard C, Battaglia D, Quilichini PP (2019). Computing hubs in the hippocampus and cortex. Sci Adv 5: eaax4843.
  • Palmigiano A, Geisel T, Wolf F, Battaglia D (2017). Flexible information routing by transient synchrony. Nat Neurosci. 20: 1014–1022.

Find Demian Battaglia