Supplementary MaterialsS1 Text: Additional information concerning the analysis procedures. appear randomly and (probably) collectively. In the demonstrated example, the overlap can be 32.25%. Middle sections display example raster plots. Each raster storyline represents one trial. Bottom level panels display discretization from the related spike raster plots. The real amount of spikes in each bin is counted to create the gray value blocks. (B) Patterns of overlapping blocks are mixed to generate 6 different stimulus circumstances. Blocks from overlap 3 (Fig 4A) are demonstrated in the remaining column. In each stimulus condition (correct column), precisely two blocks can be found. Selecting blocks can be constant for confirmed stimulus and for that reason characterizes the stimulus condition.(TIF) pcbi.1005189.s003.tif (297K) GUID:?4EF73F41-08FE-4B15-AF9C-2CEA175EDBE1 S3 Fig: Orthogonal Tucker-2 centered stimulus decoding of simulated data. (A) Stimulus decoding efficiency on simulated data built like in Fig 4A FG-4592 tyrosianse inhibitor and 4B with differing amount of tests per stimulus FG-4592 tyrosianse inhibitor and signal-to-noise percentage (SNR) acquired using orthogonal Tucker-2. (B) Percentage of right selection of the amount of modules like a function of the amount of tests per stimulus for orthogonal Tucker-2. We chosen the smallest amounts of modules with the utmost test arranged decoding performance and compared the selected numbers to the ground truth numbers (2 temporal and 2 spatial modules).(TIF) pcbi.1005189.s004.tif (485K) GUID:?39FB035F-6849-4493-A4D7-9E4C78C7F67A S4 Fig: Selection of the optimal number of spatial and temporal modules. (A) Average leave-one-out validation set decoding performance with SNR FG-4592 tyrosianse inhibitor = 20 and number of trials (training+test) per stimulus = 30 for spatiotemporal NMF (top row) and for space-by-time NMF (bottom row) is shown. Overlap of the patterns is increasing from left to right (overlap 0 to 3). We select the smallest numbers of modules with the maximum validation decoding performance (white squaresalso corresponding to the ground truth: 4 spatiotemporal modules, 2 temporal and 2 spatial modules). (B) Average decoding performance for an example experimental session on the training set averaged over leave-one-out cross-validation sub-samples for different numbers of temporal modules (x-axis) and FG-4592 tyrosianse inhibitor spatial modules (y-axis). (C) Like B but for the validation set. The smallest numbers of modules with the maximum average validation set performance are selected (white square: 8 spatial modules, 3 temporal modules).(TIF) pcbi.1005189.s005.tif (527K) GUID:?375D4B45-BC56-436C-A5F9-BDE40BC30C1B S5 Fig: Movie autocorrelations. (Top remaining) Trial still structures from both films. (Top ideal) Autocorrelation from the salamander film (dashed blue) and tiger film (dashed dark) and related salamander and tiger film autocorrelation fits (solid blue and black, respectively). (Bottom) Equations of the autocorrelation fits.(TIF) pcbi.1005189.s006.tif (449K) GUID:?134F906B-FC2C-4CFB-AEBD-554CE44A81C8 S6 Fig: Eight spatial modules that were identified by decomposition methods. The modules are represented as receptive fields. Cyan represents positive module amplitude and magenta represents negative module amplitude. The more saturated the color the stronger the absolute amplitude of the neuron in the module. (A) Modules identified by PCA. (B) Modules identified by ICA. (C) Modules identified by FA. (D) Modules identified by NMF.(TIF) pcbi.1005189.s007.tif (1.1M) GUID:?D03C5FD0-99F8-403F-ACCC-76AB58D88E1B S7 Fig: Spatial and temporal modules that were identified by orthogonal Tucker-2. Representation of the modules as in S6 Fig and for the same image dataset. (A) Modules that were identified by orthogonal Tucker-2 for numbers of modules that were optimal for space-by-time NMF to facilitate comparisons with Fig 6 and S6 Fig. (B) Modules that were identified by Rabbit polyclonal to IFIT5 orthogonal Tucker-2 for numbers of modules that were optimal for orthogonal Tucker-2. The optimal numbers of modules are higher than for space-by-time NMF.(TIF) pcbi.1005189.s008.tif (1.0M) GUID:?08F9B76A-5CE0-4F7F-83D3-C8A882C81414 S8 Fig: Module recovery similarity of modules recovered by orthogonal Tucker-2 and space-by-time NMF. Geodesic similarity between your modules retrieved for the entire FG-4592 tyrosianse inhibitor amount of tests per stimulus as well as the modules retrieved for a lesser amount of tests for orthogonal Tucker-2 (magenta) and space-by-time NMF (blue) like a function of the amount of tests per stimulus averaged total picture datasets (A, B) or all film datasets (C, D) for temporal modules (A, C) or spatial modules (B, D).(TIF) pcbi.1005189.s009.tif (202K) GUID:?0BD47935-6F70-4610-9A65-5E4FAAB4ED97 S9 Fig: Stability of modules more than stimulus sets for just one representative session with image stimuli. (A) Temporal (best) and spatial (bottom level) modules that were obtained by training on data in response to 1 1 (left), 5 (middle), and 10 (right) training images, drawn randomly from the complete set of 60 stimuli and repeated 10 times to calculate averages and standard deviations. (B) Types of first trial recordings and reconstructions with different amounts of stimuli for schooling the space-by-time.