Optical computing element and multi-neural network
An optical operational element which enables a multilayered optical neural network to be constructed without using an optical amplifier is provided. The optical operational element includes: a photothermal conversion unit 30 which converts light energy of input light A into thermal energy; a light intensity variation unit 20 which is in contact with the photothermal conversion unit 30 and which varies, in accordance with a temperature variation accompanying heat generation or heat absorption by the photothermal conversion unit 30 , intensity of external light B that is introduced from the outside; and a housing unit 10 which houses the light intensity variation unit 20 and which introduces the external light B from one side and outputs output light C obtained by attenuating intensity of the external light B to the outside on an opposite side to the one side.
1. An optical operational element, comprising:
a photothermal conversion unit which converts light energy of input light into thermal energy; a light intensity variation unit which is in contact with the photothermal conversion unit and which varies, in accordance with a temperature variation accompanying heat generation or heat absorption by the photothermal conversion unit, intensity of external light that is introduced from the outside; and
a housing unit which houses the light intensity variation unit and which introduces the external light from one side and outputs output light obtained by attenuating intensity of the external light to the outside on an opposite side to the one side.
2. The optical operational element according to claim 1 , wherein the photothermal conversion unit includes a protruding portion in a direction perpendicular to a propagation direction of the external light, the protruding portion is provided with an opening which is long in a transmission direction of temperature of the protruding portion, and a beam of the external light is irradiated into the opening.
3. The optical operational element according to claim 1 , wherein the photothermal conversion unit includes a protruding portion in a direction perpendicular to a propagation direction of the external light, the protruding portion is provided with an opening which is short in a transmission direction of temperature of the protruding portion, and a beam of the external light is irradiated into the opening.
4. A multilayer neural network in which N (N≥2)-number of the optical operational element according to claim 1 are cascade-connected, wherein the input light of an optical operational element of an n (n=2, 3, . . . , N)-th layer includes output light of an optical operational element of an n−1-th layer.
5. A multilayer neural network in which N (N≥2)-number of the optical operational element according to claim 2 are cascade-connected, wherein the input light of an optical operational element of an n (n=2, 3, . . . , N)-th layer includes output light of an optical operational element of an n−1-th layer.
6. A multilayer neural network in which N (N≥2)-number of the optical operational element according to claim 3 are cascade-connected, wherein the input light of an optical operational element of an n (n=2, 3, . . . , N)-th layer includes output light of an optical operational element of an n−1-th layer.