Analogue optoelectronic implementation of a neural network with in-situ learning

We present an implementation of a neural network using specially designed ferroelectric liquid crystal on silicon (FLC/Si) spatial light modulators. The spatial light modulators act as a somatic and a synaptic plane, and provide analogue weight multiplication. A learning algorithm is implemented locally in every synapse. The optoelectronic system therefore realises a fully interconnected neural network with a truly in-situ learning capability.