USP Lecture | The cerebral cortex, a delay coupled oscillator network: Computations in high dimensional dynamic space.

The cerebral cortex can be considered as a delay coupled recurrent network whose nodes are feature selective and have a propensity to oscillate. Such networks exhibit high dimensional non-linear dynamics that can be exploited for computations. Results obtained with parallel recordings of neuronal responses in the visual cortex suggest that the cerebral cortex uses this high dimensional dynamic space for the flexible encoding of relations among features (feature binding), for the acquisition and storage of information about statistical contingencies of features in the environment (priors), for the ultra-fast matching of priors with sensory evidence (predictive coding) and the segregation of stimulus specific activity vectors in high dimensional space (classification). In addition, the network dynamics allow for the generation of stimulus specific response sequences (temporal codes) and the superposition of information provided by sequentially presented stimuli (Fading memory). Simulations of such networks demonstrate the functional significance of the rich dynamics emerging from reciprocally coupled oscillators such as synchronisation, resonance, entrainment, phase shifts and reverberation. These computational principles differ from those realized in the multilayer feed forward architectures that characterize the deep learning networks currently used in AI systems. It is proposed that the remarkable differences between the performance of natural and artificial systems are mainly due to the fact that the former rely on analogue computation and exploit the temporal dimension as coding space. Saiba mais...