Weighting of synaptic responses

Matt Jones jonesmat at ohsu.edu
Mon Aug 26 14:12:16 EST 1996


In article <4vofut$s8f at reader1.reader.news.ozemail.net> Rose Reich,
rreich at shell02.ozemail.com.au writes:
>Q: In a modeled neural network the input into a neuron is regulated via 
>   a weighting mechanism. In "real" brains is this performed by
regulating
>   Na+ concentration within the dendrite thus changing ion gradient,
having
>   a nett effect on the input level at the hillock ? or is it done some
other
>   way ?
>
>Q: How is this weighting controlled? In a modeled neural net one simply 
>   adjusts the weighting vector, but how is this achieved in real brains
?

The most obvious modern candidate mechanisms for changing connection
weight in neurons are Long Term Potentiation (LTP) for increasing weights
and Long Term Depression (LTD) for decreasing them. Rather than changing
the intrinsic firing properties of the affected neurons, LTP and LTD
operate on the excitatory synaptic connection between one neuron and
another. That is, certain patterns of activity at a particular synapse
will cause a persistent increase or decrease in the effectiveness of that
synapse at driving postsynaptic depolarization (while spike threshold
etc. remain unaltered). This is commonly thought to be involved in memory
formation. A medline search for LTP or LTD will uncover more literature
than you ever wanted to know about. These areas are actively under study,
and some aspects of the field remain quite controversial and contentious. 

Now a question for you: In standard neural net modelling, are there any
such things as inhibitory synapses? 

Cheers,
Matt



More information about the Neur-sci mailing list