GaryGaulin
Posts: 5385
Joined: Oct. 2012
|
There is a new neural model that comes in three varieties getting around in neuroscience that works great with the ID Lab. It's more or less what I have been assuming to be happening, and provides the solution to a problem caused by there sometimes being more than one directional path contained in surrounding incoming signals. It's biological detail I hoped a breakthrough like this one would eventually explain happening in real cells.
At the moment the program code sums multiple possibilities to a single angle that is often the wrong way to go when near a place to avoid. With so much else needing work it was good enough the way it was. Now it's worth adding what's described in this paper:
New Types of Experiments Reveal that a Neuron Functions as Multiple Independent Threshold Units
www.nature.com/articles/s41598-017-18363-1
| Quote | | More precisely, the neuron contains many independent excitable sites, each functioning as an independent threshold unit which sums up the incoming signals from a given limited spatial direction, most probably by a dendrite or a bunch of dendrites (Fig. 1C, model III). These anisotropic excitable sites are not identical and are characterized by different spike waveforms and different summation specifications. The neuron is a more complex and structured computational element than expected, and the implications on the functionality of neural networks are stimulating. |
It makes sense that at least some neurons need these multiple "anisotropic excitable sites". The implications are indeed very stimulating.
Below is a HD sized peek at what the ID Lab critter now looks like. Spike train signals are shown at left. The spatial network that needs neurons with properties as described in the paper is at the upper right. New visual code for the RGB colors made it easy enough to model back from its new vestibular-like motion system with what the 12+ video MIT course Motion perception and pursuit eye movements describes for reducing a large data stream to a small number of directional spike trains that no longer need color information in them. At some point I wanted to include all that applies to 2D vision into the model. So before deciding what to do with the RGB signals I forced myself to not miss a minute of the one semister class and kept replaying parts that didn't at first make perfect sense. It included all the information I needed to make simple to code, and applies to other types of sensors where many need to be reduced down to a small stream of directional signals before addressing memory that recalls appropriate motor routines. I previously had an idea of what was needed but not being absolutely sure what was slowing me down enough to make a few days of classroom work worth the time in the long run. Right after in program comments drawing out the circuit(s) using text characters and summing up what the code has to do I found the paper at the Reddit neuroscience forum that pertains to the navigational reasoning network code. I then had to write that into program comments that now contains a link to the paper too. Without it there is nothing "in the literature" to support the reasons why the simulated neurons must work the way they do. It's like the perfect ending to the issue over the way I have been modeling neurons. My program code need for what will by virture of having been published in Nature otherwise endlessly puzzle neuroscientists is scientifically useful for explaining why it actually makes more sense than all the models these researchers are attempting to obsolete. It's much more than my being able to find evidence that neurons work as I expected there is a two way thing going on where combined they are like scientifically invincible. Buzzword based AI expectations are no longer even in the neuroscience arena. What I have became useful to what's winning for the wet-labs that like to grow neurons to experiment with, and will now explain where the authors of the paper would likely like to see me online briefly explain what I know. My having something like this to say is able to speak for itself in regards to how well things are going. Much of the above message was practice for what I need to say, at this time, even though doing so also slows down the programming work a little. There has been so much making sense all at once that keeping up is beyond what I have for free time. But now at least I have what neuroscience expects to see on the screen for signals and new way to explain what's novel about how it works:
sites.google.com/site/intelligencedesignlab/home/ScreenFor7.png
--------------
The theory of intelligent design holds that certain features of the universe and of living things are best explained by an intelligent cause, not an undirected process such as natural selection.
|
