2014 12 23, Gerhard
In our last two Skype meetings Dec 18 (D, G, M) and Dec 22 (G, M), Michael suggested a way of looking at the spikes of the synchAlpha_100s data set. A space shall be created by correlating the spike sequences of each cell with each other cell. This results in n! / ((n - 2)! * 2) distance values between 0 and 1, where n is the number of cells. Each of the two connected networks (c1 and c2) has 15.360 cells (16 hyper-colums with 32 mini-columns of 30 cells each), resulting into 117.957.120 correlations each. The 15.360 cells should now be arranged in a space such that the distances between them are represented in the best possible way.
2014 01 17, Gerhard
Today David and I made a first test of the ideas described above by computing the correlations between all the 80 potentials of the c1 network in the synchAlpha_100s data set. This resulted into 3.160 correlation values (representing a data reduction by a factor of more than 25.000 with respect to the 80 million samples in the original data set). A grey scale matrix representation of the 80 x 80 correlation values can be seen in the image to the right (white = max, black = 0). The grey values are wrongly represented in the RC. The original image can be found here.
David arranged 40 of tha data points and their 780 corresponding correlation values in a model where the correlations between pairs of points are expressed as forces. This results into constellations of point in a plane representing a compromise of the forces (degrees of correlations) between all points.