Research

Modelling natural processes requires submission to the fact that every model is wrong, but a good model is useful . This means that we never reach – and never should try to reach – full repetition or of the level of details of the natural process itself. Instead, we need to capture the key parameters for the target system which are necessary to explain the process we are interested in.

We try to understand the computational architecture of cerebral cortex by following the structure and function of a well-known model organism of vision, the Macaque monkey. In parallel, we are including the contemporary knowledge of cortical computations into the models. This approach is complementary to Human Brain Project where large-scale models aim at detailed repetition of physiology of local networks. We, instead, simplify the neurons, thus gaining computational efficiency, which we use for searching model structure and parameters, as well as for learning.

Our primary tool is CxSystem2, in-house programmed software, currently available in Github. CxSystem2 is intended to reduce the complexity of modeling by hiding the code and allowing efficient search of biological parameters and model structure.

We are developing a macaque retina simulator with the aim of providing phenomenological retina simulator which transforms videos into retinal spike trains.