High level: the net of extracellular goo around neurons inside which synapses sit (called peri-neuronal nets/PNNs) that keeps them stable day to day and sets thresholds for their “learning” based remodeling, appears to be structured by astrocytes, the third cell in many synapses, that sends out “tentacles” (not the official term, just a good visualization) to sheathe many synapses.
When the circadian clock of just these astrocyes is disrupted, after birth, the PNNs got thinner and didn’t have rhythmic structural remodeling.
Synaptic strength in these knockouts went up, but long term potentiation, a major mechanism for memory in the synapse, gets weaker. The mice had impaired learning and memory performance also.
tldr: it’s not all about the neurons. The surroundings also remodel with time, and shape the neurons themselves. Also, biological synapses are ridiculously more complicated and dynamically very different from the “weights” of a model.
High level: the net of extracellular goo around neurons inside which synapses sit (called peri-neuronal nets/PNNs) that keeps them stable day to day and sets thresholds for their “learning” based remodeling, appears to be structured by astrocytes, the third cell in many synapses, that sends out “tentacles” (not the official term, just a good visualization) to sheathe many synapses.
When the circadian clock of just these astrocyes is disrupted, after birth, the PNNs got thinner and didn’t have rhythmic structural remodeling.
Synaptic strength in these knockouts went up, but long term potentiation, a major mechanism for memory in the synapse, gets weaker. The mice had impaired learning and memory performance also.
tldr: it’s not all about the neurons. The surroundings also remodel with time, and shape the neurons themselves. Also, biological synapses are ridiculously more complicated and dynamically very different from the “weights” of a model.