Imagining Data. Part 2: Trials
spikes_noosc_200s.audio-visual texture 1
The cell populations exhibiting correlated activities (self excitation) are audified in the MUMUTH concert hall through auralisation with binaural room impulse respones. The 9 areas of activities are mapped to 9 loudspeakers around the central listening position. The 200s of simulation data are compressed to 4,53s (sped up by a factor of 44.1) and repeated 22 times, resulting in an experience of about 100s.
The sound is presented with a time-line visualisation of the cells grouped according to their activities. Both image and sound form this transposition trial. The sound has to be listened to via headphones while looking at the image.
Two unique events in the data (a plop and a beep) allow for some temporal orientation. The two types of irregularly recurring events (stong and weak excitations) form the main elements of the texture, which is set against a background of noise.
A reading of the image as a score is suggested but close to impossible. A space of uncertainly is opened in a completely determined situation. How does this relate to the simulation which created the data?
Is the rhythmical pattern one starts to perceive after having listened to several iterations in any way significant to the phenomena simulated or is it just an extract from a random sequence which gains significance through repetition alone?
spikes_noosc_200s_9x3_16.attractor_search
Parameters: ./neuro_bremen /Users/msc/ownCloud/TP/data/neuro/txt/spikes_noosc_200s_9x3_16.txt 0 0 200000 300 36 13 0.375667 0.000000 -5.107941 3.086567 1631.541748 0.0 4.724410 1.468504 1.791435 1.496465 1.772441 0.457471 0.206519 2.310104 1 1 0 1 1
The first 10 minutes of this piece was presented during a conference in Bremen. This piece attempts what may be described as 'balanced approach', pushing against artistic formal solutions while remaining legible as data-diagram. For example, attractors are clearly visible - and so are other clusters with much less clear meaning.
For aesthetic reasons - mainly to contain levels of information - this trial shows clusters as sets of minimally three points (in a 9x9 network); any single point represents an average of three cells in the same minicolumn of the network.
The 4-channel sound track is based on a granulation of the spikes_noosc_200s data set. From the 9 cell populations showing common activity, one (nr. 5) was excluded because it is active only once at the beginning of the simulation and quiet for the rest. From the 9 minicolumns in each population only 8 were used, each of them containing 30 cells. The 240 cells of each population were divided in 4 channels of 60 cells each, which were binned at 1ms. This resulted in 8 4-channel signals, one for each activity region.
synchAlpha_100s.folding/unfolding 1
Parameters: ./neuro_3 /Users/msc/ownCloud/TP/data/neuro/txt/c2_data.txt 0 0 100000 503 20 2 0.202809 -1.039371 -1.496063 1.411811 3369.452148 0.000000 100.000000 3.272441 1.007874 0.001000 0.100000 0.457471 0.206519 2.310104 1 0 0 0 1
The data entails the responses of the network to a set of patters. 'Response' here is understood as the correlated behaviour of cells belonging to the same hypercolumn (and less correlated behaviour across hypercolumns). In neuro_3, this is expressed by small triangular shapes, which tend not to overlap.
Two aspects of this can be highlighted:
(1) While it is not clear which pattern was applied (and how), it is interesting that the order by which hypercolumns become correlated is different for the different events. We ask ourselves if this has something to do with the stimulation or, alternatvely, if it is due to the modes by which the networks recalls the pattern?
(2) synchAlpha_100s_spikes1 also contain data for non-learned patterns. Here we can clearly see limited responses, although some response is still visible. We ask, may there be a way to discuss how far the folding/unfolding process goes (i.e. involve all hypercolumns or only some) and what the status of those 'semi-events' is, in particular when we look at the other data set, which contains events not triggered by an external stimulus?
spikes_noosc_200s_16.attractor_search
Parameters: 3306 0 200000 300 36 13 0.375667 0.000000 -7.070866 3.086567 185.232315 0.000000 5.511811 1.468504 1.791435 1.496465 1.772441 0.307744 0.206519 2.310104 1 1 0 1 1
Compared to spikes_noosc_200s_9x3_16, this version of the trial uses the full 9x9 data of the network. This results in more complex drawings presenting more detail.