Order-Based Representation in Random Networks of Cortical Neurons
Figure 7
Sensitivity of classification accuracy to the number of sampled electrodes.
Two examples from two different networks are shown (left and right panels). In both cases the networks were challenged with a two-stimulation sites classification task (black horizontal lines depict chance accuracy level). Support vector machine (SVM) algorithm with a Gaussian Radial Based function kernel was applied to vectors of training sets (see Materials and Methods). The resulting classifiers, one for first spike absolute latency and the other for recruitment order, were then validated using test sets vectors. Mean test accuracy (thick lines) and confidence intervals (interrupted lines) are calculated from 200 random combinations of electrodes per given sample size from all analyzed electrodes. Classification accuracy of first spike latency is depicted Red; that of recruitment order is depicted Blue. Thin lines depict the best classifier for each group size.