A Synapse

At a presynaptic terminal (top) small vesicles twenty or thirty nanometers in diameter filled with neurotransmitter molecules are waiting. Arrival of an action potential (or spike) induces a fusion of the membrane with some of the vesicles so that a neurotransmitter can diffuse into the synaptic cleft and reach receptors (not shown) at the other side, which then open ion channels they are attached to. A synapse becomes more or less efficient, like when vesicles get bigger or smaller, or more or fewer release sites become available, while postsynaptically the ion channels may increase or decrease iun number and stay open during a longer or shorter period of time. So most, though not all, of the active processes are happening in the pre- and postsynaptic membrane. The result is called learning.

(Image courtesy of Synaptic Corporation Aurora, Colorado, United States; www.synapticuse.com)

Comments

Dinesh ^club

Every step in the process of neuronal activity -- from the most distant dentdrie to the farthest axon terminal -- is fine-tuned by a slew of control mechanism. There are estimated to be at least thirty separate neurotransmitters with enzymatic steps in the creation and destruction of each transmitter affected by everything from genetics to disease. Feedback loops alter the availability and receptivity of postsynaptic receptors sites and even how cells signal and adhere to one another.

Despite a veritable symphony of interacting mechanisms, the neuron ultimately only has two options -- it either fires or it doesn’t. at this most basic level, the brain might appear like a massive compilation of on-and-off switches. But the connections between neurons are not fixed entitles. Rather they are in constant flux -- being strengthened or diminished by ongoing stimuli. Connections are enhanced with use, weakened with neglect, and are themselves affected by other connections to the same neurons. Once we leave the individual synapse between two neurons, the complexity skyrockets -- from individual neurons to a hundred billion brain cells each with thousands of connections. Although unraveling how individual neurons collectively create thought remains the Holy Grail of neuroscience, the artificial intelligence (AI) community has given us some intriguing clues as to how this might occur.

Using the biological neuron and its connections as a model, AI scientists have been able to Build artificial neural networks (ANN) that can play chess and poker, read faces, recognize speech, and recommend books on Amazon.Com. while standard computer programs work line on line, yest or no, all eventualities programmed in advance, the ANN takes an entirely different approach. The ANN is based upon mathematical programs that are initially devoid of any specific values. The programmers only provide the equations; incoming information determines how connections are formed and how strong each connection will be in relationship to all the other connections (or weighting). There is no predictable solution to a problem -- rather as one connection changes, so do all the others. These shifting interrelationships are the basis for “learning.”

With an ANN, the ‘hidden layer’ is conceptually localte within the complex interrelationships between all acquired (incoming) information and the mathematical code used to process this information. In the human brain, the ‘hidden layer’ doesn’t exist as a discrete interface or specific anatomic structure; rather it resides within the connections between all neurons involved in any neural network. A network can be relatively localized (as in a specialized visual module confined to a small area of occipital cortex), or can be widely distributed throughout the brain. …. Page 44

4 years ago. Edited 4 years ago.

Dinesh ^club

4 years ago.

		Page color
		Comments Panel
		Screen Margins
▼▲
		Widescreen

← Among Dinesh's photos

See also...

Book Images Book Images

Keywords

Authorizations, license

A Synapse

Ξ	Frame Options
3 - +