notes-science-neuro-miscNeuro

fly's sensilla (sp) smell organ has a set of odor receptors; each one has two odor receptors, a big one and a small one; if the big one activates, it suppresses the small one. This is used to hardcodes instinctual overrides into odor detection; eg flies typically ewant to stay away from high CO2 concentrations, but fruits emit CO2, and flys want to be near fruit even more; so a sensilla has a fruit odor receptor and a CO2 odor receptor, and if the fruit odor is detected, it overrides the CO2 odor receptor.

Textbook/intro links:

List of some important neuro questions:

memory idea: conceptual short term memory with a mechanism similar to auditory loop

memory idea (i think someone already had this one decades ago): 'radio' like memory with AM-like modulation

---

nicotine and aging, dementia, Alzheimer's:

http://cochranelibrary-wiley.com/doi/10.1002/14651858.CD001749/abstract

" Data collection and analysis: The one included trial did not present results suitable for inclusion in the review.

Main results: There were no results available from the one included study.

Authors' conclusions: This review is not able to provide any evidence that nicotine is or is not a useful treatment for Alzheimer´s disease. "

toread:

https://onlinelibrary.wiley.com/doi/full/10.1002/neu.10102

Nicotinic receptors in aging and dementia Marina R. Picciotto Michele Zoli J Neurobiol 53: 641–655, 2002 First published: 15 November 2002 https://doi.org/10.1002/neu.10102

Activation of neuronal nicotinic acetylcholine receptors (nAChRs) has been shown to maintain cognitive function following aging or the development of dementia. Nicotine and nicotinic agonists have been shown to improve cognitive function in aged or impaired subjects. Smoking has also been shown in some epidemiological studies to be protective against the development of neurodegenerative diseases. This is supported by animal studies that have shown nicotine to be neuroprotective both in vivo and in vitro. Treatment with nicotinic agonists may therefore be useful in both slowing the progression of neurodegenerative illnesses, and improving function in patients with the disease. While increased nicotinic function has been shown to be beneficial, loss of cholinergic markers is often seen in patients with dementia, suggesting that decreased cholinergic function could contribute to both the cognitive deficits, and perhaps the neuronal degeneration, associated with dementia. In this article we will review the literature on each of these areas. We will also present hypotheses that might address the mechanisms underlying the ability of nAChR function to protect against neurodegeneration or improve cognition, two potentially distinct actions of nicotine.

---

recall Connection Machine-style 'active data structures' (example: to sort a list you put each list element in its own CPU node, with the CPU nodes arranged in a line, then you literally have the list elements spatially migrate between CPU nodes to swap with each other until the list is sorted. Another example: to find the max of a list you again put each list element into a separate CPU node arranged as a line, then, assuming the available CPU nodes have the geometry of a grid, you literally/spatially form a pyramidal tree structure in that grid where the base of the pyramid is the input line, and then for each level above the base you do local max() operations to determine which data item goes into that CPU node, until when you reach the apex of the pyramid you have the max of the whole list)

what kinds of active data structure algorithms you can (efficiently) implement depends on the actual geometry of the nodes. For example, if the nodes have a 1D linear geometry, then that idea to use a tree to find the max isn't very efficient/feasible. So, if the API that your computing platform exposed just allowed you to reference any node by node ID, but didn't tell you the geometry of local connectivity, you might be reluctant to use it in the active data structure style.

But, in the brain, you have a dense 3D thicket of connections with each neuron having many connections. This suggests that you could probably find a subset of neurons whose connectivity pattern fits many different desired shapes, including even some higher-than-3D ones (e.g. a 4D hypercube), provided that the total number of nodes in the shape is small. So, the brain may be able to efficiently accomodate any sort of active data structure style computation for small numbers of nodes, although it may have to first search within itself to find a subset of neurons with the desired connectivity.

Furthermore, for these purposes the brain doesn't actually know the actual spatial layout of its neurons -- but it doesn't have to care -- because all that really matters for computations is (a) the time delay when sending signals between nodes, and (b) whether those signal paths are interfering with other signal paths being used in the same computation.

So, perhaps the brain scans itself for the desired subset of neurons by asking subsets of neurons to send signals to each other and seeing how long it takes for the signals to arrive/converge at various target neurons.

I feel like there is some connection here with the ideas of polychronization and with short-term plasticity (Hebbian and/or STDP), as well as the idea of the brain modeling external dynamical systems by scanning within itself for subsets of neurons that happen to be displaying similar dynamics.

---