Positioning problems

[muellerbjoern]

The current implementation reads position data from a file and positions the vertices accordingly. By default these would directly correspond to spatial positions in the recording or the simulation. In some cases, however, this is not trivial. Eg. Utah arrays often record 2 or 3 neurons per electrode, thus these have the same position. In simulations it's even possible that there are no positions, because the simulation does not have any notion of locality. These cases need to be investigated and some kind of alternative positioning needs to be found. For non-positional cases some kind of fallback mechanism would be required, e.g. a circle or some precalculated arrangement already based on correlation data (e.g. different clustering algorithms).

[muellerbjoern]

Maybe: Starting position in circle that results with same correlation between all pairs, e.g,weights == np.zeros(...) or siimilar. This however needs a starting point, too, e.g. a rectangle like the Utah array positioning, maybe random distribution on that rectangle.

[muellerbjoern]

Initial positioning other idea: https://doi.ieeecomputersociety.org/10.1109/TC.1984.1676443; https://ieeexplore.ieee.org/ielx5/12/35237/01676443.pdf?tp=&arnumber=1676443&isnumber=35237&ref=aHR0cHM6Ly9pZWVleHBsb3JlLmllZWUub3JnL3N0YW1wL3N0YW1wLmpzcD90cD0mYXJudW1iZXI9MTY3NjQ0Mw==&tag=1

Direct positioning algorithm based on eigenvectors of a matrix. Intended for weighted hypergraphs but may be applicable for lower orders