* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Topology - UCSB Physics
Biology of depression wikipedia , lookup
Executive functions wikipedia , lookup
Neural coding wikipedia , lookup
Embodied language processing wikipedia , lookup
Neurophilosophy wikipedia , lookup
Neuropsychology wikipedia , lookup
Neural modeling fields wikipedia , lookup
Emotional lateralization wikipedia , lookup
Neurogenomics wikipedia , lookup
Central pattern generator wikipedia , lookup
Donald O. Hebb wikipedia , lookup
Mirror neuron wikipedia , lookup
Artificial general intelligence wikipedia , lookup
Cognitive neuroscience wikipedia , lookup
Single-unit recording wikipedia , lookup
Clinical neurochemistry wikipedia , lookup
Affective neuroscience wikipedia , lookup
Brain Rules wikipedia , lookup
Development of the nervous system wikipedia , lookup
Biological neuron model wikipedia , lookup
Activity-dependent plasticity wikipedia , lookup
Optogenetics wikipedia , lookup
Cognitive neuroscience of music wikipedia , lookup
Convolutional neural network wikipedia , lookup
Recurrent neural network wikipedia , lookup
Time perception wikipedia , lookup
Neuroesthetics wikipedia , lookup
Embodied cognitive science wikipedia , lookup
Neuroplasticity wikipedia , lookup
Premovement neuronal activity wikipedia , lookup
Environmental enrichment wikipedia , lookup
Cortical cooling wikipedia , lookup
Orbitofrontal cortex wikipedia , lookup
Eyeblink conditioning wikipedia , lookup
Human brain wikipedia , lookup
Anatomy of the cerebellum wikipedia , lookup
Types of artificial neural networks wikipedia , lookup
Holonomic brain theory wikipedia , lookup
Metastability in the brain wikipedia , lookup
Aging brain wikipedia , lookup
Neural correlates of consciousness wikipedia , lookup
Neuropsychopharmacology wikipedia , lookup
Neuroanatomy wikipedia , lookup
Motor cortex wikipedia , lookup
Inferior temporal gyrus wikipedia , lookup
Feature detection (nervous system) wikipedia , lookup
Neuroeconomics wikipedia , lookup
Synaptic gating wikipedia , lookup
The topology of the central nervous system has been, and remains today a topic of considerable study. It is known that for humans, the central nervous system starts in the embryo as a plate, eventually deforming into a tube, one end of which thickens to become the brain (the remainder being the spinal chord). The cerebral cortex, which houses much of the machinery of intelligence, is a sheet with six layers, and numerous small columns running perpendicular to the layers. Neurons in these columns often seem to fire together in distinct patterns (Tsodyks et al 1999). Certain regions of the cortex have the same, identifiable function in all humans (e.g. the area “V1” does very basic visual processing, and is located in the same place in all humans). Thus structure in the cortex, or in the wiring to the cortex seems to be at least partially responsible for function. Visible and large-scale topology in the brain can be misleading, as the network topology of the wiring is more important than physical location. The exact wiring in the cortex is not known, because there are far too many connections (thousands per neuron) and the connections themselves are small, but may follow a convoluted path over long distance. Fortunately, it may be unnecessary to follow the exact plan of the cortex: birds lack a cortex, yet have displayed some intelligent behavior, such as tool-making (Weir et al 2002). Instead, some general design principles, may be more important. The human brain uses hierarchy (e.g. V1 sends its output to a region called V2) and feedback (V2 sends predictions back down to V1) in sensory systems, and even planning (Dehaene et al 1997). Any theoretical approach to understanding cognition must incorporate those basic principles, and any realistic topology should be consistent with them. Thus it is necessary to move beyond simple feed-forward networks with a few layers. The necessity of moving to more complex network topologies also places restrictions on neuron models. Static neurons that are either activated for all time, or never at all, will not work with feedback. Similarly, learning cannot be implemented via back-propagation to update synapse weights, as there is no logical starting or stopping point. Instead, more biologically plausible techniques must be used, like Hebbian learning, where neurons that fire together with appropriate timing will have their connections strengthened. Thus network topology and neuron models are connected and must be implemented in a complimentary manner.