Optical implementation of neural networks with wavelength-encoded bipolar weight using bacteriorhodopsin

A unique biological molecule is presented that has an intrinsic optoelectrical characteristic which can simulate both excitation and inhibition. This distinctive material is a relative of the visual pigment rhodopsin and is called bacteriorhodopsin (BR). The polarity of the optoelectrical signal induced by BR by light depends on both its photochromic state and the excitation wavelength. This unique characteristic can be used to perform the subtraction operation using incoherent light. An optical implementation of a neural network with wavelength-encoded bipolar interconnections using BR is introduced. In this scheme, the polarity of the interconnection matrix is represented by the wavelength of the light