Difference between revisions of "NeuronConnectionsResearch"

Revision as of 01:29, 16 June 2015

Neural Connections Research

@@Home -> NeuralNetworksResearch -> NeuronConnectionsResearch

What are connections

• connection is made between two neurons

* connection is uni-directional
* source of connection is always axon branch of first neuron
* target of connection can be dendrite, soma or initial axon segment of second neuron
* axon-axon connections are usually ignored in neural network models

• axon-dendrite connection elements are:

* neuron1 soma
* neuron1 initial axon segment
* neuron1 axon trunk (myelinated)
* neuron1 axon branchX
* neuron1 axon branchX terminal
* inter-neuron space (synaptic cleft)
* receptors of post-synaptic cell membrane on spine of neuron2 dendrite
* neuron2 dendrite trunk
* neuron2 soma

• there are two types of synapses, electrical and chemical

* *electrical synapse* - protein junction forms hole between axon terminal and post-synaptic neuron membranes, allowing the electrical signal to pass directly from one cell to another;
* electrical synapse is much faster than chemical synapse, but unlike chemical synapse, cannot be regulated or controlled
* *chemical synapses* may be regulated and are affected by methamphetamine, signals always travel from presynaptic membrane, through synaptic cleft, and to postsynaptic membrane

Connection dynamics

• connection can be stronger or weaker, thus having connectivity factor (see Jeff Hawkins)
• above certain connectivity factor threshold connection allows signal propagation, when presynaptic signal (action potential) produces post-synaptic signal
• below threshold connection still exists, because activity in both neurons affects connectivity factor
• connectivity factor differs from connection weight of classical neural networks, as weight always produces output which depends on weight value; connectivity factor is continuous, but its effect is binary

When connection is enforced

• Options under consideration*:
• when signal propagates through connection
• when action potential encounters fire state (Hebb's learning), e.g.:

* just after firing there is negative potential in all dendrites
* which electically attracts axon terminal having positive action potential
* while firing is impossible (refractory period) and action potential energy is spent for increasing connectivity factor

• using interneurons and non-neuron interneuron matter (glia)
• using complex structures as controllers (bump attractors and so on)