The architecture of the thin-film transistor is based on a floating gate and an algorithm which weighs multiple input signals. Only above a certain threshold, the transistor opens its gate, thus permitting neural network applications.
CAS press release, March 26, 2013