Modeling somatic computation with non-neural bioelectric networks

The set of input patterns used for training the pattern detector consists of two categories: (1) French-flag (FF) and (2) Not-French-flag (NFF). The subset FF consists of the French-flag pattern itself and its slightly noisy variants. The latter is generated by perturbing each pixel of the French-flag approximately within the range [−24 mV, +24 mV]. Any perturbation that takes a pixel below the bounds of [−80 mV, +80 mV] are reset to the closest boundary values. The subset NFF consists of the more distorted variants of the French-flag. It is generated in the same way as FF except the perturbation range is [−40 mV, +40 mV]. A total of 20 samples in each set were generated during each iteration of the GA-BP search.