Kevin, I just came back from a workshop in Washington D.C. and in one talk someone mentioned that they were surprised to find an axon making multiple synapses onto the same dendrite contrary to common belief. I didn’t get the detail on this so I don’t know if it is true or not, but just want to let you know that there are often people reporting to find something that breaks a well-established rule.
Jeff From: nupic [mailto:nupic-boun...@lists.numenta.org] On Behalf Of Kevin Martin Jose Sent: Wednesday, February 12, 2014 3:12 AM To: NuPIC general mailing list. Subject: Re: [nupic-discuss] Quick clarification Thank you, Jeff. _____ From: Jeff Hawkins <mailto:jhawk...@numenta.org> Sent: 10-02-2014 00:13 To: 'NuPIC general mailing list.' <mailto:nupic@lists.numenta.org> Subject: Re: [nupic-discuss] Quick clarification The short answer to your questions are No for the first question and Yes for the second. In CLA theory and implementation we allow only one connection between an input bit and a column’s proximal dendrite. We don’t want to allow multiple connections from the same input bit. We want the column to detect a spatial pattern in the input and if we allowed 15 of the synapses to come from the same input bit then the column wouldn’t recognize a pattern, just the activity of a single input bit. In biology it is rare, and maybe unheard of, for an axon from a relay cell in the thalamus to make multiple synapses on the same proximal dendrite of cortical cell. The axon splits and makes connections to many different cells but not multiple synapses on the same cell. This is a general rule. For example, if you take two cells near each other in a layer of cortex, cell A and Cell B, it is not uncommon to find multiple connections from cell A onto the distal dendrites of cell B, but not multiple synapses near each other, not to the same dendrite segment. This is the same property applied to distal dendrites. I read a paper that showed there was a signaling mechanism within the branching ends of an axon that discouraged the formation of two or more synapses onto the same dendrite segment. The authors were surprised because two synapses on the same axon had to know that there was coincident post-synaptic activity and then transmit this knowledge through the axon (down one branch an up the next) for this to happen. Anyway, there appears to be a mechanism that prevents axons from forming multiple synapses onto the same dendrite segment. BTW, this isn’t true for inhibitory cells (they usually make multiple synapses onto the same cell) and for other parts of the brain. Jeff From: nupic [mailto:nupic-boun...@lists.numenta.org] On Behalf Of Kevin Martin Sent: Sunday, February 09, 2014 2:54 AM To: NuPIC general mailing list. Subject: [nupic-discuss] Quick clarification A small doubt. Can multiple synapses from the same column (proximal dendrite) connect to the same input bit? eg: 20 input bits, one column. 15 synapses. Each synapse should connect to a unique bit?
_______________________________________________ nupic mailing list nupic@lists.numenta.org http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org
