Wesley R. Elsberry
Posts: 4991
Joined: May 2002
|
| Quote (GaryGaulin @ April 13 2015,09:14) | | Quote (Wesley R. Elsberry @ April 13 2015,01:18) | | Quote (GaryGaulin @ April 12 2015,14:12) | | Quote (Wesley R. Elsberry @ April 12 2015,07:02) | | Quote (GaryGaulin @ April 11 2015,10:11) | | Quote (Wesley R. Elsberry @ April 11 2015,08:04) | | I wonder if Gary can point to a place in any of the neuroscience articles where they state that an organism is born with its place and grid cells already mapped to an existing environment, at the correct scale, its current location represented there, and its orientation set. |
It looks like Wesley is going to try impressing us with their sciency sounding Evo-Devo vocabulary that explains absolutely nothing, but at least makes them look smart to peers and to funding agencies who regularly flush money down a toilet. |
Gary is confused. I was making a critique of systems that claim to model place and grid cells but which start with such cells initialized to map to a particular environment, at a particular scale, with the state set to reflect a particular organism location and orientation. I don't think such systems can be justified by reference to the literature, and by Gary's non-response, neither does he. |
It is not a static grid anymore. Last week I added what was needed to "remap with subtle changes to the spatial geometry of the environment" and that led to it being easy to add articulated proprioception, which I expect will in turn cause remapping the whole thing to new scales, which in turn accounts for grid field behavior I read about elsewhere but until now I did not know what could be causing it.
The model already does what is highlighted below, and all the rest just became easier:
| Quote | FINAL NOTE: HIPPOCAMPAL PLACE CELLS AS AN EXAMPLE OF MODULAR CODING OF THE ALLOCENTRIC REPRESENTATION?
Thus far, we have focused on behavioral, lesion, and fMRI studies, which argue against allocentric navigational strategies depending on a single brain region (Figure 3A) and as decomposable into contributions from individual brain regions (Figure3B). One might argue, as others have (O’Keefe and Nadel, 1978; Redish, 1999), however, that place cells, present in the rodent, monkey, and human hippocampus (O’Keefe and Dostrovsky, 1971; Ekstrom et al., 2003; Hori et al., 2003), are the neural instantiation of an allocentric representation, or cognitive map. While place cells do have many features similar to what one might expect in a neural systems that code spatial environments in a map-like fashion, there are other important features of place cells that are decidedly not map-like. Place cells in the rodent and human hippocampus remap based on egocentric direction (Markus et al., 1995; Miller et al., 2013), are sensitive to goal and other temporal variables (Gothard et al., 2001; Hollup et al., 2001; Ekstrom et al., 2003; Bahar et al., 2011), and remap with subtle changes to the spatial geometry of the environment (Leutgeb et al., 2005; Wills et al., 2005). Indeed, recent theoretical models of the cognitive map now suggest that time and geometry less variant spatial coding mechanisms possibly resides outside of the hippocampus (Buzsaki, 2006; Buzsaki and Moser, 2013). Grid cells, neurons in enthorhinal cortex that fire in a regularly spaced fashion as the rat explores a spatial environment (Fyhn et al., 2004; Jacobs et al., 2013), may be a better candidate for the neural basis of allocentric representation (Buzsaki and Moser, 2013). Yet lesions of entorhinal cortex, at least in rodents, do not abolish place cell firing in the CA3 subfield of the hippocampus (Lu et al., 2013) and impair, but do not abolish, the place code in CA1 (Brun et al., 2008). While many details of entorhinal–hippocampal neural interactions remain to be established, grid cells do not contribute in a clear or modular fashion to place coding in the hippocampus, at least based on what the above-mentioned studies have determined so far in the rat. Furthermore, in addition to grid cells, entorhinal cortex cells also respond to egocentric direction (Sargolini et al., 2006), suggesting this area may not be specialized for allocentric computations alone. In addition, consistent with what we have argued here, it is clear that other areas, like prefrontal and retrosplenial cortex, also contribute critically, via oscillatory synchrony, to spatial coding in the hippocampus (Benchenane et al., 2010; Battaglia et al., 2011; Fujisawa and Buzsaki, 2011). Thus, although many aspects of the hippocampal neural code would appear sufficient to support an allocentric representation, the neural code itself is not map-like and depends, at least in part, on coordinated input and activity from other brain structures. |
http://www.ncbi.nlm.nih.gov/pmc........4193251 |
"Static grid" is not among the topics I broached. |
You did not raise (a sensitive or difficult subject) for discussion. You just still seem to be expecting neuron behavior to be modeled another way that you're used to instead of new way that does away with needing an "ANN" or Hopfield network type thing, while at the same time giving the behavior what makes places on the grid map depend on what it wants, without having to add code to make it want to go where it should, towards goals it inherently sets based on immediate needs.
In this modeling method timed signals of all cells/neurons add up to waves that get crowd surfed around by them. Exactly how cells easily manage this using proper timing of multiplexed waves is unique to cells, while how a personal computer can accomplish the same is unique to how digital circuitry works. It's therefore proper for this model to show how that is most easily accomplished in code, not what is only applicable to cells. The neuron on up models that universities are working on would in this case only reinvent the wheel while violating the Occam's Razor simple requirement that the ID Lab has upon it.
Using another method to model the neural behavior will not make the critter more representative of biological systems. What I'm programming is a benchmark for models that try the same thing another way would behave if is working just right. |
Gary is confused yet again. I didn't say a word about "NEURONS"* in the criticism in question.
*Meaning the absence of any such from Gary's code. Given that the whole exercise is premised on modeling neural assemblages, that is a major oversight, but not the particular criticism of the moment.
Edited by Wesley R. Elsberry on April 13 2015,09:42
--------------
"You can't teach an old dogma new tricks." - Dorothy Parker
|
