Open tclose opened 10 years ago
PyNN allows multiple layouts for a given population (e.g. x,y,z position in cortical space plus theta,phi position of receptive field within the visual field). The layout should in any case be defined as a Component outside the population. I am worried that allowing the layout to be provided to the connection generator would increase the likelihood of writing inconsistent or incorrect models. Layouts could also potentially be used for neuron parameterization.
The reason I thought it would be neater to use provide the layouts to the components that required them directly is that you then wouldn't have to specify a means to reference which layout property of the population you want to use (I couldn't think of a nice way of doing this). I could be wrong, but I don't envisage it increasing the likelihood of mistakenly using the wrong layout any more than having to specify which layout property of the population.
Layouts defined at the document level could be used to initialise neuron parameterisations in much the same way as I was proposing for projections.
In the 2.0dev spec, I have proposed that such layout information would be simply provided by properties that sit under the top-level Population
container. For layouts derived from components a DerivedProperty
element could be used instead of a Property
element. (see #73)
The committee has addressed this issue by #63, with the additional information that layout and other non-dynamical cell properties should be included in the Component by not in the Prototype.
I believe that the layout of cells is a property of the population in both PyNN and SpineCreator. However, it may be better if the cell layouts are provided to the connection generator as required.