3 ms·

I didn't find the jand-waving analogies persuasive, but it hadn't really struck me before how extraordinary it is that DNA uses a code for proteins (3-base).

DoctorOetker 6y ago
So I read the article on the ferret experiment "Memory trace and timing mechanism localized to cerebellar Purkinje cells" , and very simplified: theres this big neuron called a Purkinje cell with a large dendritic tree (input side). Purkinje cells have parallel fibers (axons) and climbing fibers (CF also axons) that synapse with their dendritic tree. To make a long story short one type carries a warning signal which I will call the Start signal, and the other carries unconditional reflex signal (induced such that it elicits a blink) which I will call the Stop signal. The training involves a warning start stimulus, then a fixed pause, and then the annoying stop stimulus. After training, it is observed that the neuron statistically responds after a pause commensurate with the trained pause. Okay perhaps a single neuron programmable variable delay is not easy to explain with a conventional synaptically weighted neuron. (at this point I remark to myself that a conventional synaptic weighted neuron might still do that in theory if the input axon is long and windy to use it as a delay line memory, but then it would need to have lots of synapses with the same Purkinje cell, so I reject this brainfart as not biologically plausible) Then training was resumed with a second interval, and after this secondary training the following was observed: after receiving a start signal, the purkinje cell statistically responds with 2 (!) responses corresponding to the trained pauses! To them this is the nail in the coffin for the combinational synaptic weighted neuron model. But This just makes me remember my original tap filter layout: the same single input axon could function as a delay line with multiple synaptic taps activated at different times. So I look up the terms parallel fiber and climbing fiber... and guess what: unless the brain is still gestating or suffered local damage each Purkinje cell has exactly 1 climbing fiber axon associated with it. And this climbing fiber is long and windy and makes loads of synaptic connections. So the initial remaining explanation delay line tap filter brainfart (which I viewed as myself clasping at straws to save the synaptically weighted neuron model), turns out to be nearly exactly the layout of climbing fiber on purkinje cell. you can view this as a programmable tap filter, or as an impulse response convolver. So instead of deposing the synaptically weighted neuron perspective, I view this as strengthening the usual interpretation! Evolution forced synaptically weighted neuron model to contort itself into an awkward tapped delay line filter!

DoctorOetker 6y ago

So I read the article on the ferret experiment "Memory trace and timing mechanism localized to cerebellar Purkinje cells" , and very simplified: theres this big neuron called a Purkinje cell with a large dendritic tree (input side). Purkinje cells have parallel fibers (axons) and climbing fibers (CF also axons) that synapse with their dendritic tree. To make a long story short one type carries a warning signal which I will call the Start signal, and the other carries unconditional reflex signal (induced such that it elicits a blink) which I will call the Stop signal. The training involves a warning start stimulus, then a fixed pause, and then the annoying stop stimulus. After training, it is observed that the neuron statistically responds after a pause commensurate with the trained pause. Okay perhaps a single neuron programmable variable delay is not easy to explain with a conventional synaptically weighted neuron. (at this point I remark to myself that a conventional synaptic weighted neuron might still do that in theory if the input axon is long and windy to use it as a delay line memory, but then it would need to have lots of synapses with the same Purkinje cell, so I reject this brainfart as not biologically plausible) Then training was resumed with a second interval, and after this secondary training the following was observed: after receiving a start signal, the purkinje cell statistically responds with 2 (!) responses corresponding to the trained pauses! To them this is the nail in the coffin for the combinational synaptic weighted neuron model. But This just makes me remember my original tap filter layout: the same single input axon could function as a delay line with multiple synaptic taps activated at different times. So I look up the terms parallel fiber and climbing fiber... and guess what: unless the brain is still gestating or suffered local damage each Purkinje cell has exactly 1 climbing fiber axon associated with it. And this climbing fiber is long and windy and makes loads of synaptic connections. So the initial remaining explanation delay line tap filter brainfart (which I viewed as myself clasping at straws to save the synaptically weighted neuron model), turns out to be nearly exactly the layout of climbing fiber on purkinje cell. you can view this as a programmable tap filter, or as an impulse response convolver. So instead of deposing the synaptically weighted neuron perspective, I view this as strengthening the usual interpretation! Evolution forced synaptically weighted neuron model to contort itself into an awkward tapped delay line filter!