WO2020105459A1 - Optical arithmetic operation element and multi-layered neural network - Google Patents

Optical arithmetic operation element and multi-layered neural network

Info

Publication number
WO2020105459A1
WO2020105459A1 PCT/JP2019/043736 JP2019043736W WO2020105459A1 WO 2020105459 A1 WO2020105459 A1 WO 2020105459A1 JP 2019043736 W JP2019043736 W JP 2019043736W WO 2020105459 A1 WO2020105459 A1 WO 2020105459A1
Authority
WO
WIPO (PCT)
Prior art keywords
optical
light
refractive index
path
optical waveguide
Prior art date
Application number
PCT/JP2019/043736
Other languages
French (fr)
Japanese (ja)
Inventor
大塚　卓哉
Original Assignee
日本電信電話株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 日本電信電話株式会社 filed Critical 日本電信電話株式会社
Publication of WO2020105459A1 publication Critical patent/WO2020105459A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G02—OPTICS
    • G02B—OPTICAL ELEMENTS, SYSTEMS OR APPARATUS
    • G02B26/00—Optical devices or arrangements for the control of light using movable or deformable optical elements
    • G02B26/08—Optical devices or arrangements for the control of light using movable or deformable optical elements for controlling the direction of light
    • G—PHYSICS
    • G02—OPTICS
    • G02B—OPTICAL ELEMENTS, SYSTEMS OR APPARATUS
    • G02B6/00—Light guides; Structural details of arrangements comprising light guides and other optical elements, e.g. couplings
    • G02B6/24—Coupling light guides
    • G02B6/26—Optical coupling means
    • G02B6/35—Optical coupling means having switching means
    • G—PHYSICS
    • G02—OPTICS
    • G02F—OPTICAL DEVICES OR ARRANGEMENTS FOR THE CONTROL OF LIGHT BY MODIFICATION OF THE OPTICAL PROPERTIES OF THE MEDIA OF THE ELEMENTS INVOLVED THEREIN; NON-LINEAR OPTICS; FREQUENCY-CHANGING OF LIGHT; OPTICAL LOGIC ELEMENTS; OPTICAL ANALOGUE/DIGITAL CONVERTERS
    • G02F3/00—Optical logic elements; Optical bistable devices
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
    • G06N3/067—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using optical means

Definitions

  • the present invention relates to an optical operation element and a multi-layer neural network forming an optical neural network.
  • An optical neural network models a neural network in the human brain with a unit consisting of two neurons, an input layer neuron and an output layer neuron, and a synapse connecting each neuron, and networked using optical signals. It is a thing.
  • An optical neural network is generally configured by connecting neuron elements that perform sum-of-products calculation and non-linear calculation and forming a multilayer (for example, Non-Patent Document 1).
  • the conventional optical neural network needs to perform photoelectric conversion to amplify the light attenuated by making it multi-layered. There is a problem that the power loss and the speed loss associated with the photoelectric conversion are large.
  • the present invention has been made in view of this problem, and an object thereof is to provide an optical operation element and a multilayer neural network that can amplify light without performing photoelectric conversion.
  • An optical arithmetic element has a refractive index that matches a light deforming member that deforms according to the intensity of input light, a deforming lid that is connected to the light deforming member and that is deformed by the light deforming member.
  • Refractive index matching agent that transmits external light introduced from the outside, a matching agent reservoir covered with the deformable lid and filled with the refractive index matching agent, and the refractive index matching agent from the matching agent reservoir ,
  • a first optical waveguide that is inclined with respect to the path and that propagates the external light, and a path that is disposed on an extension line of the first optical waveguide with the path interposed therebetween and that passes through the path.
  • the gist of the present invention is to include a second optical waveguide that outputs the external light that has come.
  • the gist is that the input light of the optical operation element of the layer is the output light of the optical operation element of the (n-1) th layer.
  • an optical operation element and a multilayer neural network that can amplify light without performing photoelectric conversion.
  • FIG. 1 is a perspective view schematically showing a configuration example of an optical operation element according to the first embodiment of the present invention. It is sectional drawing which follows the AA line shown in FIG. It is a figure which shows typically the relationship between the light beam which propagates the 1st optical waveguide shown in FIG. 1, and reaches
  • FIG. 2B is a schematic diagram showing an example of the relationship between input light and output light. It is a figure which shows typically the example which cascaded two optical operation elements shown in FIG. It is a figure which shows typically the example which connected the optical operation element shown in FIG.
  • FIG. 1 is a perspective view schematically showing a configuration example of an optical operation element according to the first embodiment of the present invention.
  • the optical arithmetic element 1 shown in FIG. 1 is an optical arithmetic element that amplifies light without using photoelectric conversion.
  • FIG. 2 is a sectional view taken along the line AA shown in FIG. The configuration of the optical operation element 1 will be described with reference to FIGS. 1 and 2.
  • the optical arithmetic element 1 includes a light deforming member 30, a deforming lid 40, a refractive index matching agent 50, a matching agent reservoir 60, a path 70, a first optical waveguide 80a, and a second optical waveguide 80b.
  • the inside of the matching agent reservoir 60 is filled with the refractive index matching agent 50.
  • Each component of the optical arithmetic element 1 is housed in, for example, a rectangular parallelepiped casing 10.
  • the housing 10 is made of, for example, an organic molecular polymer or quartz.
  • the housing 10 may be made of another material or metal.
  • the shape of the housing 10 is not limited to a rectangular parallelepiped as long as the optical operation element 1 is configured by combining the respective components as shown in FIG. Further, the housing 10 may be composed of a frame. That is, it is not necessary to hold each component in a solid such as a rectangular parallelepiped.
  • An opening 20 is provided on one end side of the housing 10.
  • the opening 20 of this example has a quadrangular plane and penetrates in the height direction of the housing 10.
  • the input light A of the optical signal is input to the opening 20.
  • a direction is defined for explanation.
  • the side of the opening 20 of the housing 10 is rear, and the opposite side is front.
  • the central portion of the inner wall on the front side of the opening 20 is hollowed out in a cylindrical shape to form a matching agent reservoir 60.
  • the deforming lid 40 is fitted on the opening 20 side (rear side) of the matching agent reservoir 60.
  • the deformable lid 40 is made of a flexible material and deforms when a force is input.
  • the deformable lid 40 is made of rubber, for example.
  • a flat U-shaped locking portion 41 is formed in the center of the deformable lid 40.
  • the locking portion 11 having the same shape is also formed on the inner wall on the rear side of the opening 20 facing the locking portion 41.
  • the optical deformation member 30 is stretched between the locking portion 11 and the locking portion 41, and both ends of the optical deformation member 30 are fixed to the locking portion 11 and the locking portion 41, respectively.
  • the light deforming member 30 connects the deforming lid 40 and the inner wall (rear side) of the opening 20 while maintaining a predetermined tension.
  • the light deforming member 30 deforms according to the intensity of the input light A.
  • a crosslinked polymer having diarylethene, cyclodextrin, and azobenzene can be used.
  • the inside of the matching agent reservoir 60 is filled with the refractive index matching agent 50 and sealed with the deformation lid 40.
  • a path 70 having a rectangular cross section is formed from the central portion of the front end surface of the matching agent reservoir 60.
  • the refractive index matching agent 50 for example, silicone oil can be used.
  • the refractive index of the refractive index matching agent 50 is, for example, 1.485 (25 ° C.), and has substantially the same refractive index as the first optical waveguide 80a and the second optical waveguide 80b.
  • the path 70 has, for example, a rectangular cross section, and extends horizontally from the front end surface of the matching agent reservoir 60 to the front end surface of the housing 10, and the tip is open.
  • the inside of the path 70 is led out of the refractive index matching agent 50 from the matching agent reservoir 60, and is filled with the refractive index matching agent 50 up to about a half position in the front-rear direction of the path 70.
  • the tip portion of the refractive index matching agent 50 moves in the front-back direction in response to the deformation of the light deformable member 30 according to the intensity of the input light A.
  • the derivation direction of the route 70 is not limited to the horizontal direction.
  • the index matching agent 50 in the path 70 is hardly affected by gravity due to its surface tension. Therefore, the derivation direction of the path 70 may be any of the vertical up and down direction, the oblique up and down direction, and the like.
  • the position of the tip portion of the refractive index matching agent 50 in the path 70 is mainly determined by the deformation amount of the deformation lid 40.
  • the first optical waveguide 80a is arranged so as to be inclined with respect to the path 70 of the portion in which the tip portion of the refractive index matching agent 50 moves in the front-rear direction, and the tip portion is in contact with the path 70.
  • External light B is input from the side of the first optical waveguide 80a opposite to the path 70.
  • the external light B is ambient light of the environment in which the optical arithmetic element 1 is arranged.
  • the external light B is kept at a constant intensity (illuminance). The intensity of the external light B may vary to some extent.
  • the second optical waveguide 80b is arranged on the extension line of the first optical waveguide 80a with the path 70 interposed therebetween.
  • the second optical waveguide 80b outputs the external light B transmitted through the path 70 to the outside.
  • the first optical waveguide 80a and the second optical waveguide 80b are made of a material having a higher refractive index than the other housing 10 when the housing 10 is made of an organic molecular polymer.
  • the refractive index of the first optical waveguide 80a and the second optical waveguide 80b is, for example, 1.49.
  • the housing 10, the matching agent reservoir 60, and the path 70 are formed by processing, for example, rectangular parallelepiped quartz by a well-known semiconductor process and micromachining technology.
  • the optical arithmetic element 1 includes the light deformable member 30 that deforms according to the intensity of input light, and the deformable lid 40 that is connected to the light deformable member 30 and deformed by the light deformable member 30.
  • a refractive index matching agent 50 that matches the refractive index and transmits external light B introduced from the outside, a matching agent reservoir 60 that is covered with the deformation lid 40 and is filled with the refractive index matching agent 50,
  • the path 70 is sandwiched between the path 70 having an opening at the tip for leading out the refractive index matching agent 50 from the agent reservoir 60 in the horizontal direction, the first optical waveguide 80a inclined to the path 70 and propagating the external light B.
  • the second optical waveguide 80b is provided on the extension line of the first optical waveguide 80a and outputs the external light C transmitted through the path 70 (output light C).
  • FIG. 3 is a diagram schematically showing the relationship between the light beam propagating through the first optical waveguide 80 a and reaching the path 70 and the refractive index matching agent 50 moving in the path 70.
  • FIG. 3A is a schematic diagram showing the relationship between the light beam and the refractive index matching agent 50
  • FIG. 3B shows an example of the relationship between the input light and the output light.
  • the ellipse 81 shown in FIG. 3A schematically represents a light beam (hereinafter, light beam 81) propagating through the first optical waveguide 80a and reaching the path 70. Since the first optical waveguide 80a is in contact with the path 70 with an inclination, the shape of the light beam 81 is elliptical.
  • the tip of the refractive index matching agent 50 that moves in the path 70 due to the change in the intensity of the input light A is positioned at the rear end of the light beam 81 when the intensity of the input light A is maximum.
  • the input light A is adjusted to be positioned at the front end portion of the light beam 81 when the intensity of the input light A is minimum.
  • the tip of the refractive index matching agent 50 is adjusted to be located at the front end of the light beam 81 when the intensity of the input light A is maximum, for example.
  • the intensity of the input light A is minimum, the light beam 81 is adjusted to be located at the rear end portion. That is, the logic for the intensity of the input light A can be inverted depending on the direction in which the light deformable member 30 deforms.
  • the tip of the refractive index matching agent 50 is located at ⁇ shown in FIG. 3 (a). Therefore, since the refractive index matching agent 50 occupies most of the area of the light beam 81, the refractive index of the first optical waveguide 80a and the refractive index of the path 70 match within the range occupied by the refractive index matching agent 50. Therefore, the external light A is transmitted through the path 70 in the range occupied by the refractive index matching agent 50 in the light beam 81.
  • the tip portion of the refractive index matching agent 50 is located at ⁇ shown in FIG. 3 (a). Therefore, the refractive index of the optical waveguide 80 a and the refractive index of the path 70 match only in a part of the light beam 81, and most of the external light A is reflected by the path 70. Therefore, since only part of the external light B is transmitted through the path 70, the intensity of the output light C decreases ( ⁇ in FIG. 3B).
  • the tip portion of the refractive index matching agent 50 is located at ⁇ shown in FIG. 3 (a). Therefore, since the refractive index matching agent 50 occupies half the area of the light beam 81, half of the external light B is transmitted through the path 70 and becomes the output light C.
  • the intensity of the external light B can be changed (inversely proportional in this example) in response to the change in the intensity of the input light A.
  • the intensity of the external light B is constant. Therefore, by setting the intensity of the external light B to be greater than the maximum intensity of the input light A, the output light B obtained by amplifying the input light A can be output.
  • the light intensity of the input light A can be amplified without using photoelectric conversion. Further, since the optical operation element 1 does not perform photoelectric conversion, it does not cause power loss and speed loss accompanying it.
  • a constant external light B is given (irradiated) to each of the optical operation elements 1 connected in multiple layers (a plurality of cascades) to form a multilayer neural network that does not cause a loss in the intensity of the optical signal.
  • FIG. 4 is a diagram schematically showing a configuration in which two optical operation elements 1 according to the present embodiment are connected in cascade.
  • the optical signal output from the second optical waveguide 80b of the first optical operation element 1 1 is input to the opening 20 2 of the second optical operation element 1 2 .
  • the optical processing element 1 1 and the optical operation elements 1 first optical waveguide 80b 1 of 2 and the first optical waveguide 80b 2, the external light B constant light intensity are input.
  • -A multilayer neural network can be configured by connecting two or more optical operation elements 1 in cascade.
  • FIG. 5 is a diagram schematically showing a configuration example of the multilayer neural network according to the present embodiment.
  • the multi-layer neural network 100 shown in FIG. 5 has the above-described optical operation elements 1 connected in cascade in two or more layers.
  • First layer of output light Z 1 of the optical processing element 1 1 is input to the multiplier 101 2 to generate a second layer of the input light of the optical operation elements 1 2.
  • the multiplier 101 2 is multiplied by the weight w 3 outputs to one input of the adder 103 2 on the output light Z 1.
  • the adder 103 2 adds the output of the multiplier 101 2 and the output of the multiplier 102 2 to generate the input light of the optical arithmetic element 1 2 .
  • the output of the multiplier 102 2 is obtained by multiplying the output light Z 2 of the optical calculation element (not shown) by the weight w 4 .
  • the output light Z 5 to 3 and subsequent layers of the optical operation element 1 3 is also generated, configuration for generating the output light Z 5 is the same as the optical operation elements 1 2 in the second layer.
  • the reference numbers in the figure are updated and shown, and the description thereof is omitted.
  • the input light A n of the optical operation element in the) -th layer includes the output light Z n ⁇ 1 of the optical operation element in the (n ⁇ 1) th layer.
  • the external light B having a constant intensity is input to each of the optical operation elements 1 n of each layer, and the external light B is output light Z n ⁇ of the optical operation element 1 n-1 of the previous layer n ⁇ 1.
  • the output light Z n converted by 1 is generated. Therefore, the intensity of the output light Z n of the rear optical operation element 1 n connected in cascade in multiple layers is not attenuated. As a result, photoelectric conversion is unnecessary, and the multilayer neural network can be made to have no power.
  • optical components may be arranged between the optical operation elements 1 n of each layer.
  • the optical component may be, for example, an optical filter or a coupler.
  • Figure 6 is a diagram schematically showing an example in which for example the light filter 90 between the optical processing element 1 1 and the optical operation element 1 2 shown in FIG.
  • the optical filter 90 may be replaced with a coupler that branches an optical signal.
  • FIG. 7 is a perspective view schematically showing a configuration example of the optical operation element according to the second embodiment of the present invention.
  • FIG. 7 is a diagram corresponding to FIG.
  • the optical arithmetic element 2 shown in FIG. 7 is different from the optical arithmetic element 1 (FIG. 1) in that it has a third optical waveguide 80c.
  • the third optical waveguide 80c guides the reflected light reflected by the path 70 to the outside.
  • reflected light reflected by the path 70 there are two types of reflected light reflected by the path 70: reflected light that propagates through the first optical waveguide 80a and returns to the input side, and reflected waves that repeat reflection until they disappear near the path 70.
  • the latter reflected wave may deteriorate the SN ratio of the calculation of the optical signal.
  • the optical arithmetic element 2 has a reflection angle that is the same as the incident angle of the first optical waveguide 80a with respect to the path 70, and makes the tip end contact the tip end of the first optical waveguide 80a. Equipped with 80c.
  • the external light B reflected by the path 70 is guided to the outside by the third optical waveguide 80c. Therefore, the number of reflected waves that are repeatedly reflected in the vicinity of the path 70 is reduced, so that the SN ratio of the optical signal calculation can be improved.
  • the reflected light derived from the third optical waveguide 80c for the calculation.
  • the degree of freedom in designing the multilayer neural network can be improved.
  • the optical operation elements 1 and 2 of this embodiment it is possible to construct a multilayer optical neural network without performing photoelectric conversion. Further, according to the multilayer neural network 100 of the present embodiment, the external light B is introduced into each layer from the outside, so that it is not necessary to alternately perform the calculation using the optical signal and the electric signal. As a result, it is possible to reduce the power consumption of the calculation.
  • the light deformable member 30 of the present invention has been described in the example of a so-called string shape, but the invention is not limited to this example.
  • the light deformable member 30 may be, for example, a belt-shaped member or a braided member.
  • the light deformable member 30 may be anything as long as it undergoes deformation such as expansion, bending, and expansion.
  • the present invention is not limited to the above-described embodiment, but can be modified within the scope of the gist thereof.
  • Optical operation element 10 Housing 20: Opening 30: Optical deformation member 40: Deformation lid 50: Refractive index matching agent 60: Matching agent reservoir 70: Path 80a: First light Waveguide 80b: Second optical waveguide 80c: Third optical waveguide 90: Optical filter or coupler 100: Multilayer neural network A: Input light B: External light C: Output light

Abstract

Provided is an optical arithmetic operation element that is capable of amplifying input light without using an optical amplifier. This element is provided with: a photo-deformable member 30 that is deformed according to the intensity of input light A; a deformable lid 40 that is coupled to the photo-deformable member 30 and that is deformed by the photo-deformable member 30; a refractive index matching material 50 that is for matching refractive indices and allows transmission of external light B introduced thereto from outside; a matching material reservoir 60 that is covered with the deformable lid 40 and filled with the refractive index matching material 50; a passage way 70 that has an open leading end through which the refractive index matching material 50 is drawn out from the matching material reservoir 60; a first optical waveguide 80a that is inclined with respect to the passage way 70 and propagates the external light B; and a second optical waveguide 80b that is disposed on an extension line of the first optical waveguide 80a across the passage way 70 and outputs the external light B that has transmitted through the passage way 70.

Description

光演算素子と多層ニューラルネットワークOptical operation element and multilayer neural network
　本発明は、光ニューラルネットワークを構成する光演算素子と多層ニューラルネットワークに関する。 The present invention relates to an optical operation element and a multi-layer neural network forming an optical neural network.
　光ニューラルネットワークは、人間の脳内にある神経細胞網を入力層ニューロンと出力層ニューロンの二つのニューロンと、それぞれのニューロンを連結するシナプスから成る単位でモデル化し、光信号を用いてネットワーク化したものである。 An optical neural network models a neural network in the human brain with a unit consisting of two neurons, an input layer neuron and an output layer neuron, and a synapse connecting each neuron, and networked using optical signals. It is a thing.
　光ニューラルネットワークは、一般的に積和演算と非線形演算を実行するニューロン素子を結合し、多層化されて構成される（例えば非特許文献１）。 An optical neural network is generally configured by connecting neuron elements that perform sum-of-products calculation and non-linear calculation and forming a multilayer (for example, Non-Patent Document 1).
　従来の光ニューラルネットワークは、多層化することで減衰した光を増幅する為に光電変換を行う必要がある。その光電変換に伴う電力損失と速度損失が大きいという課題がある。 The conventional optical neural network needs to perform photoelectric conversion to amplify the light attenuated by making it multi-layered. There is a problem that the power loss and the speed loss associated with the photoelectric conversion are large.
　本発明は、この課題に鑑みてなされたものであり、光電変換を行うことなく光を増幅できる光演算素子と多層ニューラルネットワークを提供することを目的とする。 The present invention has been made in view of this problem, and an object thereof is to provide an optical operation element and a multilayer neural network that can amplify light without performing photoelectric conversion.
　本発明の一態様に係る光演算素子は、入力光の強度に応じて変形する光変形部材と、前記光変形部材と連結され該光変形部材によって変形させられる変形蓋と、屈折率を整合させて外部から導入される外部光を透過させる屈折率整合剤と、前記変形蓋で蓋をされ前記屈折率整合剤で満たされた整合剤溜まり部と、前記整合剤溜まり部から前記屈折率整合剤を導出する先端が開口された経路と、前記経路に対して傾斜し前記外部光を伝搬させる第１光導波路と、前記経路を挟んで前記第１光導波路の延長線上に配置され前記経路を透過して来た前記外部光を出力する第２光導波路とを備えることを要旨とする。 An optical arithmetic element according to an aspect of the present invention has a refractive index that matches a light deforming member that deforms according to the intensity of input light, a deforming lid that is connected to the light deforming member and that is deformed by the light deforming member. Refractive index matching agent that transmits external light introduced from the outside, a matching agent reservoir covered with the deformable lid and filled with the refractive index matching agent, and the refractive index matching agent from the matching agent reservoir , A first optical waveguide that is inclined with respect to the path and that propagates the external light, and a path that is disposed on an extension line of the first optical waveguide with the path interposed therebetween and that passes through the path. The gist of the present invention is to include a second optical waveguide that outputs the external light that has come.
　また、本発明の一態様に係る多層ニューラルネットワークは、上記の光演算素子をＮ（Ｎ≧2）個縦続接続させた多層ニューラルネットワークであって、ｎ（ｎ＝2,3,…,Ｎ）層目の光演算素子の前記入力光は、ｎ－１層目の光演算素子の出力光を含むことを要旨とする。 A multi-layer neural network according to an aspect of the present invention is a multi-layer neural network in which N (N ≧ 2) optical operation elements are cascade-connected, and n (n = 2, 3, ..., N). The gist is that the input light of the optical operation element of the layer is the output light of the optical operation element of the (n-1) th layer.
　本発明によれば、光電変換を行うことなく光を増幅できる光演算素子と多層ニューラルネットワークを提供することができる。 According to the present invention, it is possible to provide an optical operation element and a multilayer neural network that can amplify light without performing photoelectric conversion.
本発明の第1実施形態に係る光演算素子の構成例を模式的に示す斜視図である。FIG. 1 is a perspective view schematically showing a configuration example of an optical operation element according to the first embodiment of the present invention. 図１に示すＡ－Ａ線に沿う断面図である。It is sectional drawing which follows the AA line shown in FIG. 図１に示す第１光導波路を伝搬して経路に到達する光ビームと経路内を移動する屈折率整合剤との関係を模式的に示す図であり、（ａ）光ビームと屈折の関係を示す模式図、（ｂ）は入力光と出力光の関係の一例を示す図である。It is a figure which shows typically the relationship between the light beam which propagates the 1st optical waveguide shown in FIG. 1, and reaches | attains a path | route, and the refractive index matching agent which moves in a path | route, (a) shows the relationship between a light beam and refraction. FIG. 2B is a schematic diagram showing an example of the relationship between input light and output light. 図１に示す光演算素子を2個縦続させた例を模式的に示す図である。It is a figure which shows typically the example which cascaded two optical operation elements shown in FIG. 図１に示す光演算素子を多層に接続して多層ニューラルネットワークを構成した例を模式的に示す図である。It is a figure which shows typically the example which connected the optical operation element shown in FIG. 1 in multiple layers, and comprised the multilayer neural network. 図４に示す2個の光演算素子の間に光学部品を配置した例を模式的に示す図である。It is a figure which shows typically the example which has arrange | positioned the optical component between the two optical arithmetic elements shown in FIG. 本発明の第２実施形態に係る光演算素子の構成例を模式的に示す斜視図である。It is a perspective view which shows typically the structural example of the optical arithmetic element which concerns on 2nd Embodiment of this invention.
　以下、本発明の実施形態について図面を用いて説明する。複数の図面中同一のものには同じ参照符号を付し、説明は繰り返さない。 Embodiments of the present invention will be described below with reference to the drawings. The same parts in the drawings are designated by the same reference numerals, and the description thereof will not be repeated.
　〔第１実施形態〕
　図１は、本発明の第１実施形態に係る光演算素子の構成例を模式的に示す斜視図である。図１に示す光演算素子１は、光電変換を用いずに光を増幅する光演算素子である。 [First Embodiment]
FIG. 1 is a perspective view schematically showing a configuration example of an optical operation element according to the first embodiment of the present invention. The optical arithmetic element 1 shown in FIG. 1 is an optical arithmetic element that amplifies light without using photoelectric conversion.
　（光演算素子の構成）
　図２は、図１に示すＡ－Ａ線に沿う断面図である。図１と図２を参照して光演算素子１の構成について説明する。 (Structure of optical operation element)
FIG. 2 is a sectional view taken along the line AA shown in FIG. The configuration of the optical operation element 1 will be described with reference to FIGS. 1 and 2.
　光演算素子１は、光変形部材３０、変形蓋４０、屈折率整合剤５０、整合剤溜まり部６０、経路７０、第１光導波路８０ａ、及び第２光導波路８０ｂを備える。整合剤溜まり部６０の内部は、屈折率整合剤５０で満たされている。 The optical arithmetic element 1 includes a light deforming member 30, a deforming lid 40, a refractive index matching agent 50, a matching agent reservoir 60, a path 70, a first optical waveguide 80a, and a second optical waveguide 80b. The inside of the matching agent reservoir 60 is filled with the refractive index matching agent 50.
　光演算素子１の各構成部は、例えば直方体の筐体１０に収められている。筐体１０は、例えば有機分子ポリマーあるいは石英で構成される。なお、筐体１０は他の材料、金属で構成してもよい。 Each component of the optical arithmetic element 1 is housed in, for example, a rectangular parallelepiped casing 10. The housing 10 is made of, for example, an organic molecular polymer or quartz. The housing 10 may be made of another material or metal.
　図１に示すように各構成部が組み合わされて光演算素子１が構成されれば、筐体１０の形状は直方体に限られない。また、筐体１０はフレームで構成してもよい。つまり、各構成部を、直方体等の立体で保持する必要もない。 The shape of the housing 10 is not limited to a rectangular parallelepiped as long as the optical operation element 1 is configured by combining the respective components as shown in FIG. Further, the housing 10 may be composed of a frame. That is, it is not necessary to hold each component in a solid such as a rectangular parallelepiped.
　筐体１０の一方の端部側には、開口部２０が設けられる。この例の開口部２０は、平面が四角形であり、筐体１０の高さ方向を貫通している。開口部２０は、光信号の入力光Ａが入力される。ここで、説明のために方向を定義する。筐体１０の開口部２０側を後、その反対側を前とする。 An opening 20 is provided on one end side of the housing 10. The opening 20 of this example has a quadrangular plane and penetrates in the height direction of the housing 10. The input light A of the optical signal is input to the opening 20. Here, a direction is defined for explanation. The side of the opening 20 of the housing 10 is rear, and the opposite side is front.
　開口部２０の前側の内壁の中央部分は、円柱状に刳り抜かれ、整合剤溜まり部６０が形成される。整合剤溜まり部６０の開口部２０側（後側）は、変形蓋４０が嵌められている。 The central portion of the inner wall on the front side of the opening 20 is hollowed out in a cylindrical shape to form a matching agent reservoir 60. The deforming lid 40 is fitted on the opening 20 side (rear side) of the matching agent reservoir 60.
　変形蓋４０は、柔軟性を持つ素材で構成され、力が入力されることで変形する。変形蓋４０は、例えばゴムで構成される。 The deformable lid 40 is made of a flexible material and deforms when a force is input. The deformable lid 40 is made of rubber, for example.
　変形蓋４０の中央部分には、平面がＵ字状の係止部４１が形成されている。係止部４１が対向する開口部２０の後側の内壁にも、同形状の係止部１１が形成されている。 A flat U-shaped locking portion 41 is formed in the center of the deformable lid 40. The locking portion 11 having the same shape is also formed on the inner wall on the rear side of the opening 20 facing the locking portion 41.
　係止部１１と係止部４１の間に光変形部材３０が掛け渡され、光変形部材３０の両端部は、係止部１１と係止部４１にそれぞれ固定される。光変形部材３０は所定の張力を保持した状態で、変形蓋４０と開口部２０の内壁（後側）を接続させる。 The optical deformation member 30 is stretched between the locking portion 11 and the locking portion 41, and both ends of the optical deformation member 30 are fixed to the locking portion 11 and the locking portion 41, respectively. The light deforming member 30 connects the deforming lid 40 and the inner wall (rear side) of the opening 20 while maintaining a predetermined tension.
　光変形部材３０は、入力光Ａの強度に応じて変形する。光変形部材３０は、例えばジアリールエテン、シクロデキストリン、及びアゾベンゼンを有する架橋ポリマー等を用いることができる。 The light deforming member 30 deforms according to the intensity of the input light A. For the light deformable member 30, for example, a crosslinked polymer having diarylethene, cyclodextrin, and azobenzene can be used.
　整合剤溜まり部６０の内部は、屈折率整合剤５０で満たされ変形蓋４０で密閉される。整合剤溜まり部６０の前側の端面の中央部分から、断面が長方形の経路７０が形成される。 The inside of the matching agent reservoir 60 is filled with the refractive index matching agent 50 and sealed with the deformation lid 40. A path 70 having a rectangular cross section is formed from the central portion of the front end surface of the matching agent reservoir 60.
　屈折率整合剤５０は、例えば、シリコンオイルを用いることができる。屈折率整合剤５０の屈折率は、例えば1.485（25℃）であり、第１光導波路８０ａ及び第２光導波路８０ｂとほぼ同じ屈折率を持つ。 As the refractive index matching agent 50, for example, silicone oil can be used. The refractive index of the refractive index matching agent 50 is, for example, 1.485 (25 ° C.), and has substantially the same refractive index as the first optical waveguide 80a and the second optical waveguide 80b.
　経路７０は、その断面が例えば長方形であり、整合剤溜まり部６０の前側の端面から、筐体１０の前側の端面まで水平に貫通し先端は開放されている。経路７０の内部は、整合剤溜まり部６０から屈折率整合剤５０が導出され、経路７０の前後方向の半分程度の位置まで屈折率整合剤５０で満たされている。屈折率整合剤５０の先端部分は、光変形部材３０が入力光Ａの強度に応じて変形するのに対応して前後方向に移動する。 The path 70 has, for example, a rectangular cross section, and extends horizontally from the front end surface of the matching agent reservoir 60 to the front end surface of the housing 10, and the tip is open. The inside of the path 70 is led out of the refractive index matching agent 50 from the matching agent reservoir 60, and is filled with the refractive index matching agent 50 up to about a half position in the front-rear direction of the path 70. The tip portion of the refractive index matching agent 50 moves in the front-back direction in response to the deformation of the light deformable member 30 according to the intensity of the input light A.
　なお、経路７０の導出方向は、水平方向に限られない。経路７０内の屈折率整合剤５０は、その表面張力によって重力の影響をほとんど受けない。したがって、経路７０の導出方向は、鉛直方向の上下、又は斜め上下方向等の何れで有ってもよい。経路７０内の屈折率整合剤５０の先端部分の位置は、主に変形蓋４０の変形量によって決定される。 Note that the derivation direction of the route 70 is not limited to the horizontal direction. The index matching agent 50 in the path 70 is hardly affected by gravity due to its surface tension. Therefore, the derivation direction of the path 70 may be any of the vertical up and down direction, the oblique up and down direction, and the like. The position of the tip portion of the refractive index matching agent 50 in the path 70 is mainly determined by the deformation amount of the deformation lid 40.
　第１光導波路８０ａは、屈折率整合剤５０の先端部分が前後方向に移動する部分の経路７０に対して傾斜して配置され、その先端部分は経路７０に当接されている。第１光導波路８０ａの経路７０と反対側からは外部光Ｂが入力される。外部光Ｂは、光演算素子１が配置された環境の環境光のことである。外部光Ｂは一定の強度（照度）に保たれている。外部光Ｂの強度は、ある程度変動しても構わない。 The first optical waveguide 80a is arranged so as to be inclined with respect to the path 70 of the portion in which the tip portion of the refractive index matching agent 50 moves in the front-rear direction, and the tip portion is in contact with the path 70. External light B is input from the side of the first optical waveguide 80a opposite to the path 70. The external light B is ambient light of the environment in which the optical arithmetic element 1 is arranged. The external light B is kept at a constant intensity (illuminance). The intensity of the external light B may vary to some extent.
　第２光導波路８０ｂは、経路７０を挟んで第１光導波路８０ａの延長線上に配置される。第２光導波路８０ｂは、経路７０を透過して来た外部光Ｂを外部に出力する。 The second optical waveguide 80b is arranged on the extension line of the first optical waveguide 80a with the path 70 interposed therebetween. The second optical waveguide 80b outputs the external light B transmitted through the path 70 to the outside.
　第１光導波路８０ａと第２光導波路８０ｂは、筐体１０を有機分子ポリマーで構成した場合、他の筐体１０の部分よりも屈折率が高い材料で構成される。第１光導波路８０ａと第２光導波路８０ｂの屈折率は、例えば1.49である。 The first optical waveguide 80a and the second optical waveguide 80b are made of a material having a higher refractive index than the other housing 10 when the housing 10 is made of an organic molecular polymer. The refractive index of the first optical waveguide 80a and the second optical waveguide 80b is, for example, 1.49.
　筐体１０、整合剤溜まり部６０、及び経路７０は、周知の半導体プロセス及びマイクロマシン加工技術によって、例えば直方体の石英を加工することで形成される。 The housing 10, the matching agent reservoir 60, and the path 70 are formed by processing, for example, rectangular parallelepiped quartz by a well-known semiconductor process and micromachining technology.
　以上述べたように本実施形態に係る光演算素子１は、入力光の強度に応じて変形する光変形部材３０と、光変形部材３０と連結され該光変形部材３０によって変形させられる変形蓋４０と、屈折率を整合させて外部から導入される外部光Ｂを透過させる屈折率整合剤５０と、変形蓋４０で蓋をされ屈折率整合剤５０で満たされた整合剤溜まり部６０と、整合剤溜まり部６０から屈折率整合剤５０を水平方向に導出する先端が開口された経路７０と、経路７０に対して傾斜し外部光Ｂを伝搬させる第１光導波路８０ａと、経路７０を挟んで第１光導波路８０ａの延長線上に配置され経路７０を透過して来た外部光Ｃを出力（出力光Ｃ）する第２光導波路８０ｂとを備える。 As described above, the optical arithmetic element 1 according to the present embodiment includes the light deformable member 30 that deforms according to the intensity of input light, and the deformable lid 40 that is connected to the light deformable member 30 and deformed by the light deformable member 30. A refractive index matching agent 50 that matches the refractive index and transmits external light B introduced from the outside, a matching agent reservoir 60 that is covered with the deformation lid 40 and is filled with the refractive index matching agent 50, The path 70 is sandwiched between the path 70 having an opening at the tip for leading out the refractive index matching agent 50 from the agent reservoir 60 in the horizontal direction, the first optical waveguide 80a inclined to the path 70 and propagating the external light B. The second optical waveguide 80b is provided on the extension line of the first optical waveguide 80a and outputs the external light C transmitted through the path 70 (output light C).
　（光演算素子の作用）
　図３は、第１光導波路８０ａを伝搬して経路７０に到達する光ビームと経路７０内を移動する屈折率整合剤５０との関係を模式的に示す図である。図３（ａ）は光ビームと屈折率整合剤５０の関係を示す模式図、図３（ｂ）は入力光と出力光の関係の一例を示す。 (Function of optical operation element)
FIG. 3 is a diagram schematically showing the relationship between the light beam propagating through the first optical waveguide 80 a and reaching the path 70 and the refractive index matching agent 50 moving in the path 70. FIG. 3A is a schematic diagram showing the relationship between the light beam and the refractive index matching agent 50, and FIG. 3B shows an example of the relationship between the input light and the output light.
　図３（ａ）に示す楕円８１は、第１光導波路８０ａを伝搬して経路７０に到達する光ビーム（以降、光ビーム８１）を模式的に表す。第１光導波路８０ａは、経路７０に対して傾斜して接するので光ビーム８１の形状は楕円形になる。 The ellipse 81 shown in FIG. 3A schematically represents a light beam (hereinafter, light beam 81) propagating through the first optical waveguide 80a and reaching the path 70. Since the first optical waveguide 80a is in contact with the path 70 with an inclination, the shape of the light beam 81 is elliptical.
　ここで、入力光Ａの強度の変化によって経路７０内を移動する屈折率整合剤５０の先端は、入力光Ａの強度が最大の場合に光ビーム８１の後側の端部に位置するように調整されていると仮定する。また、入力光Ａの強度が最小の場合に光ビーム８１の前側の端部に位置するように調整されていると仮定する。 Here, the tip of the refractive index matching agent 50 that moves in the path 70 due to the change in the intensity of the input light A is positioned at the rear end of the light beam 81 when the intensity of the input light A is maximum. Suppose it has been adjusted. Further, it is assumed that the input light A is adjusted to be positioned at the front end portion of the light beam 81 when the intensity of the input light A is minimum.
　この仮定は、光変形部材３０が入力光Ａの強度が大きいと収縮する場合に成立する。光変形部材３０が変形する方向が逆であれば、屈折率整合剤５０の先端は、例えば入力光Ａの強度が最大の場合に光ビーム８１の前側の端部に位置するように調整され、入力光Ａの強度が最小の場合に光ビーム８１の後側の端部に位置するように調整される。つまり、光変形部材３０の変形する方向によって、入力光Ａの強度に対する論理を反転させることができる。 This assumption holds when the light deformable member 30 contracts when the intensity of the input light A is large. If the direction in which the light deformable member 30 deforms is opposite, the tip of the refractive index matching agent 50 is adjusted to be located at the front end of the light beam 81 when the intensity of the input light A is maximum, for example. When the intensity of the input light A is minimum, the light beam 81 is adjusted to be located at the rear end portion. That is, the logic for the intensity of the input light A can be inverted depending on the direction in which the light deformable member 30 deforms.
　上記の仮定において、入力光Ａの強度が最小の強度よりも少し大きい場合は、屈折率整合剤５０の先端部分が図３（ａ）に示すαに位置する。よって、光ビーム８１の面積のほとんどを屈折率整合剤５０が占めるので、屈折率整合剤５０が占める範囲内で第１光導波路８０ａの屈折率と経路７０の屈折率が整合する。よって、外部光Ａは、光ビーム８１内の屈折率整合剤５０が占める範囲で経路７０を透過する。 Under the above assumption, when the intensity of the input light A is slightly higher than the minimum intensity, the tip of the refractive index matching agent 50 is located at α shown in FIG. 3 (a). Therefore, since the refractive index matching agent 50 occupies most of the area of the light beam 81, the refractive index of the first optical waveguide 80a and the refractive index of the path 70 match within the range occupied by the refractive index matching agent 50. Therefore, the external light A is transmitted through the path 70 in the range occupied by the refractive index matching agent 50 in the light beam 81.
　この状態を図３（ｂ）にαで示す。図３（ｂ）に示すように入力光Ａの強度が最小の強度よりも少し大きい場合、外部光Ｂのほとんどが経路７０を透過し出力光Ｃとなる。つまり、この例では、入力光Ａの強度が最小の場合に外部光Ｂは遮光される。 This state is indicated by α in Fig. 3 (b). As shown in FIG. 3B, when the intensity of the input light A is slightly higher than the minimum intensity, most of the external light B passes through the path 70 and becomes the output light C. That is, in this example, the external light B is blocked when the intensity of the input light A is minimum.
　また、上記の仮定において、入力光Ａの強度が最大の強度よりも少し小さい場合は、屈折率整合剤５０の先端部分が図３（ａ）に示すγに位置する。したがって、光ビーム８１内の一部分でしか光導波路８０ａの屈折率と経路７０の屈折率が整合せず、外部光Ａのほとんとが経路７０で反射する。よって、外部光Ｂの一部しか経路７０を透過しないので出力光Ｃの強度は低下する（図３（ｂ）のγ）。 Further, under the above assumption, when the intensity of the input light A is slightly smaller than the maximum intensity, the tip portion of the refractive index matching agent 50 is located at γ shown in FIG. 3 (a). Therefore, the refractive index of the optical waveguide 80 a and the refractive index of the path 70 match only in a part of the light beam 81, and most of the external light A is reflected by the path 70. Therefore, since only part of the external light B is transmitted through the path 70, the intensity of the output light C decreases (γ in FIG. 3B).
　また、上記の仮定において、入力光Ａの強度が最小－最大の範囲の中間の強度の場合は、屈折率整合剤５０の先端部分が図３（ａ）に示すβに位置する。よって、光ビーム８１の面積の半分を屈折率整合剤５０が占めるので、外部光Ｂの半分が経路７０を透過して出力光Ｃとなる。 Further, under the above assumption, when the intensity of the input light A is an intermediate intensity between the minimum and maximum ranges, the tip portion of the refractive index matching agent 50 is located at β shown in FIG. 3 (a). Therefore, since the refractive index matching agent 50 occupies half the area of the light beam 81, half of the external light B is transmitted through the path 70 and becomes the output light C.
　以上説明したように、図３に示す例では外部光Ｂの強度を、入力光Ａの強度の変化に対応させて変化（この例では反比例）させることができる。外部光Ｂの強度は一定である。よって、外部光Ｂの強度を入力光Ａの最大強度よりも大きな強度に設定しておくことで、入力光Ａを増幅した出力光Ｂを出力することができる。 As described above, in the example shown in FIG. 3, the intensity of the external light B can be changed (inversely proportional in this example) in response to the change in the intensity of the input light A. The intensity of the external light B is constant. Therefore, by setting the intensity of the external light B to be greater than the maximum intensity of the input light A, the output light B obtained by amplifying the input light A can be output.
　つまり、本実施形態に係る光演算素子１によれば、光電変換を用いずに入力光Ａの光強度を増幅することができる。また、光演算素子１は、光電変換を行わないのでそれに伴う電力損失と速度損失を生じさせない。 That is, according to the optical arithmetic element 1 according to the present embodiment, the light intensity of the input light A can be amplified without using photoelectric conversion. Further, since the optical operation element 1 does not perform photoelectric conversion, it does not cause power loss and speed loss accompanying it.
　また、一定の外部光Ｂを、多層（複数を縦続）に接続された光演算素子１のそれぞれに与える（照射する）ことで、光信号の強度の損失が生じさせない多層ニューラルネットワークを構成することができる。 In addition, a constant external light B is given (irradiated) to each of the optical operation elements 1 connected in multiple layers (a plurality of cascades) to form a multilayer neural network that does not cause a loss in the intensity of the optical signal. You can
　（多層ニューラルネットワーク）
　図４は、本実施形態に係る光演算素子１を2個縦続接続させた構成を模式的に示す図である。1個目の光演算素子１１の第２光導波路８０ｂから出力された光信号を、2個目の光演算素子１２の開口部２０２に入力させる。光演算素子１１と光演算素子１２の第１光導波路８０ｂ１と第１光導波路８０ｂ２には、一定の光強度の外部光Ｂがそれぞれ入力される。 (Multilayer neural network)
FIG. 4 is a diagram schematically showing a configuration in which two optical operation elements 1 according to the present embodiment are connected in cascade. The optical signal output from the second optical waveguide 80b of the first optical operation element 1 1 is input to the opening 20 2 of the second optical operation element 1 2 . The optical processing element 1 1 and the optical operation elements 1 first optical waveguide 80b 1 of 2 and the first optical waveguide 80b 2, the external light B constant light intensity are input.
　2個以上の光演算素子１を縦続接続することで多層ニューラルネットワークを構成することができる。 -A multilayer neural network can be configured by connecting two or more optical operation elements 1 in cascade.
　図５は、本実施形態に係る多層ニューラルネットワークの構成例を模式的に示す図である。図５に示す多層ニューラルネットワーク１００は、上記の光演算素子１を２層以上、縦続に接続したものである。 FIG. 5 is a diagram schematically showing a configuration example of the multilayer neural network according to the present embodiment. The multi-layer neural network 100 shown in FIG. 5 has the above-described optical operation elements 1 connected in cascade in two or more layers.
　１層目の光演算素子１１の出力光Ｚ１は、２層目の光演算素子１２の入力光を生成する乗算器１０１２に入力される。乗算器１０１２は、出力光Ｚ１に重みｗ３を乗じて加算器１０３２の一方の入力に出力する。 First layer of output light Z 1 of the optical processing element 1 1 is input to the multiplier 101 2 to generate a second layer of the input light of the optical operation elements 1 2. The multiplier 101 2 is multiplied by the weight w 3 outputs to one input of the adder 103 2 on the output light Z 1.
　加算器１０３２は、乗算器１０１２の出力と乗算器１０２２の出力を加算して光演算素子１２の入力光を生成する。乗算器１０２２の出力は、図示しない光演算素子の出力光Ｚ２に重みｗ４を乗じたものである。 The adder 103 2 adds the output of the multiplier 101 2 and the output of the multiplier 102 2 to generate the input light of the optical arithmetic element 1 2 . The output of the multiplier 102 2 is obtained by multiplying the output light Z 2 of the optical calculation element (not shown) by the weight w 4 .
　２層目の光演算素子１２は、外部から取り込む外部光Ｂを、入力光Ａに相当する加算器１０３２が出力する積和信号で変換した出力光Ｚ３を生成する。３層目以降の光演算素子１３が生成する出力光Ｚ５についても、当該出力光Ｚ５を生成するための構成は、２層目の光演算素子１２と同じである。図中の参照符号の番号を更新して表記し、その説明は省略する。 The second layer of the optical operation element 1 2, the external light B to be imported from the outside, to generate an output light Z 3 the corresponding adder 103 2 is converted by the product-sum signal to be output to the input light A. The output light Z 5 to 3 and subsequent layers of the optical operation element 1 3 is also generated, configuration for generating the output light Z 5 is the same as the optical operation elements 1 2 in the second layer. The reference numbers in the figure are updated and shown, and the description thereof is omitted.
　以上説明したように本実施形態に係る多層ニューラルネットワーク１００は、光演算素子１をＮ（Ｎ≧2）個縦続接続させた多層ニューラルネットワークであって、ｎ（ｎ＝2,3,…,Ｎ）層目の光演算素子の入力光Ａｎは、ｎ－１層目の光演算素子の出力光Ｚｎ－１を含む。 As described above, the multilayer neural network 100 according to the present embodiment is a multilayer neural network in which N (N ≧ 2) optical operation elements 1 are cascade-connected, and n (n = 2, 3, ..., N). The input light A n of the optical operation element in the) -th layer includes the output light Z n−1 of the optical operation element in the (n−1) th layer.
　この構成によれば、各層の光演算素子１ｎのそれぞれに一定強度の外部光Ｂが入力され、該外部光Ｂが前層ｎ－１の光演算素子１ｎ－１の出力光Ｚｎ－１で変換された出力光Ｚｎを生成する。したがって、多層に縦続接続された後方の光演算素子１ｎの出力光Ｚｎの強度は減衰しない。その結果、光電変換が不要であり、多層ニューラルネットワークを無電力化することができる。 According to this configuration, the external light B having a constant intensity is input to each of the optical operation elements 1 n of each layer, and the external light B is output light Z n− of the optical operation element 1 n-1 of the previous layer n−1. The output light Z n converted by 1 is generated. Therefore, the intensity of the output light Z n of the rear optical operation element 1 n connected in cascade in multiple layers is not attenuated. As a result, photoelectric conversion is unnecessary, and the multilayer neural network can be made to have no power.
　各層の光演算素子１ｎの間に他の光学部品を配置するようにしてもよい。光学部品は、例えば光フィルター及びカプラー等が考えられる。 Other optical components may be arranged between the optical operation elements 1 n of each layer. The optical component may be, for example, an optical filter or a coupler.
　図６は、図４に示した光演算素子１１と光演算素子１２の間に例えば光フィルター９０を配置した例を模式的に示す図である。光フィルター９０は、光信号を分岐させるカプラーに置き換えてもよい。 Figure 6 is a diagram schematically showing an example in which for example the light filter 90 between the optical processing element 1 1 and the optical operation element 1 2 shown in FIG. The optical filter 90 may be replaced with a coupler that branches an optical signal.
　このようにｎ（ｎ＝2,3,…,Ｎ）層目の光演算素子とｎ＋１層目の光演算素子の間に、n個目の光演算素子の出力光を、分岐させるカプラー又は変調させる光フルターが配置されるように多層ニューラルネットワークを構成してもよい。これによれば多層ニューラルネットワークの設計の自由度を向上させることができる。 Thus, a coupler or a modulator for branching the output light of the n-th optical operation element between the n-th (n = 2, 3, ..., N) -th layer optical operation element and the (n + 1) -th layer optical operation element. The multi-layer neural network may be configured such that the optical filters that cause it are arranged. According to this, the degree of freedom in designing the multilayer neural network can be improved.
　〔第２実施形態〕
　図７は、本発明の第２実施形態に係る光演算素子の構成例を模式的に示す斜視図である。図７は図１に対応する図である。 [Second Embodiment]
FIG. 7 is a perspective view schematically showing a configuration example of the optical operation element according to the second embodiment of the present invention. FIG. 7 is a diagram corresponding to FIG.
　図７に示す光演算素子２は、第３光導波路８０ｃを備える点で光演算素子１（図１）と異なる。第３光導波路８０ｃは、経路７０で反射した反射光を外部に導出させるものである。 The optical arithmetic element 2 shown in FIG. 7 is different from the optical arithmetic element 1 (FIG. 1) in that it has a third optical waveguide 80c. The third optical waveguide 80c guides the reflected light reflected by the path 70 to the outside.
　経路７０で反射した反射光は、第１光導波路８０ａを伝搬して入力側に戻る反射光と、経路７０付近で消失するまで反射を繰り返す反射波の2つが存在すると考えられる。後者の反射波は、光信号の演算のＳＮ比を劣化させる場合がある。 It is considered that there are two types of reflected light reflected by the path 70: reflected light that propagates through the first optical waveguide 80a and returns to the input side, and reflected waves that repeat reflection until they disappear near the path 70. The latter reflected wave may deteriorate the SN ratio of the calculation of the optical signal.
　そこで本実施形態に係る光演算素子２は、第１光導波路８０ａの経路７０に対する入射角度と同じ角度の反射角度を持ち、先端部を第１光導波路８０ａの先端部に接触させる第３光導波路８０ｃを備える。これにより、経路７０で反射された外部光Ｂは、第３光導波路８０ｃによって外部に導出される。したがって、経路７０付近で反射を繰り返す反射波が減少するので、光信号の演算のＳＮ比を向上させることができる。 Therefore, the optical arithmetic element 2 according to the present embodiment has a reflection angle that is the same as the incident angle of the first optical waveguide 80a with respect to the path 70, and makes the tip end contact the tip end of the first optical waveguide 80a. Equipped with 80c. As a result, the external light B reflected by the path 70 is guided to the outside by the third optical waveguide 80c. Therefore, the number of reflected waves that are repeatedly reflected in the vicinity of the path 70 is reduced, so that the SN ratio of the optical signal calculation can be improved.
　また、第３光導波路８０ｃから導出される反射光を演算に用いることも可能である。第３光導波路８０ｃから出力される反射光を演算に用いることで多層ニューラルネットワークの設計の自由度を向上させることができる。 Also, it is possible to use the reflected light derived from the third optical waveguide 80c for the calculation. By using the reflected light output from the third optical waveguide 80c for the calculation, the degree of freedom in designing the multilayer neural network can be improved.
　以上説明したように本実施形態の光演算素子１，２によれば、光電変換を行うことなく、多層化した光ニューラルネットワークを構築できる。また、本実施形態の多層ニューラルネットワーク１００によれば、各層のそれぞれに外部から外部光Ｂが導入されるので、光信号と電気信号を用いた演算を交互に行う必要がない。その結果、演算を無電力化することができる。 As described above, according to the optical operation elements 1 and 2 of this embodiment, it is possible to construct a multilayer optical neural network without performing photoelectric conversion. Further, according to the multilayer neural network 100 of the present embodiment, the external light B is introduced into each layer from the outside, so that it is not necessary to alternately perform the calculation using the optical signal and the electric signal. As a result, it is possible to reduce the power consumption of the calculation.
　また、非金属製の材料で構成することが可能なので無電力であることも含め、従来の半導体チップを利用したＩｏＴデバイスの適用が困難な利用場面での光演算素子の活用を可能にする。また、光電変換を行わないことによる部品点数の削減によるコストダウン及び故障リスクを減少させるという効果も奏する。 Also, since it can be composed of non-metallic materials, it can be used for optical computing elements in situations where it is difficult to apply conventional IoT devices using semiconductor chips, including no power consumption. Further, the effect of reducing cost and failure risk by reducing the number of parts by not performing photoelectric conversion is also obtained.
　なお、本発明の光変形部材３０は、いわゆる紐状の形態の例で説明したが、この例に限定されない。光変形部材３０は、例えば帯状であっても紐を編んだものであってもよい。要するに光変形部材３０は、膨張、屈曲、及び伸張等の変形が生じるものであれば何でも構わない。このように本発明は、上記の実施形態に限定されるものではなく、その要旨の範囲内で変形が可能である。 Note that the light deformable member 30 of the present invention has been described in the example of a so-called string shape, but the invention is not limited to this example. The light deformable member 30 may be, for example, a belt-shaped member or a braided member. In short, the light deformable member 30 may be anything as long as it undergoes deformation such as expansion, bending, and expansion. As described above, the present invention is not limited to the above-described embodiment, but can be modified within the scope of the gist thereof.
　なお、光演算素子１，２の加工方法について具体例を示した説明を行わなかったが、当該加工については既存の半導体プロセス及びマイクロマシン加工技術の全てを用いることができる。 Note that, although a specific example of the method of processing the optical operation elements 1 and 2 has not been described, all existing semiconductor processes and micromachine processing techniques can be used for the processing.
１、１１～１３、２：光演算素子
１０：筐体
２０：開口部
３０：光変形部材
４０：変形蓋
５０：屈折率整合剤
６０：整合剤溜まり部
７０：経路
８０ａ：第１光導波路
８０ｂ：第２光導波路
８０ｃ：第３光導波路
９０：光フィルター又はカプラー
１００：多層ニューラルネットワーク
Ａ：入力光
Ｂ：外部光
Ｃ：出力光
  1, 1 1 to 1 3 2: Optical operation element 10: Housing 20: Opening 30: Optical deformation member 40: Deformation lid 50: Refractive index matching agent 60: Matching agent reservoir 70: Path 80a: First light Waveguide 80b: Second optical waveguide 80c: Third optical waveguide 90: Optical filter or coupler 100: Multilayer neural network A: Input light B: External light C: Output light

Claims (5)

  1. 　入力光の強度に応じて変形する光変形部材と、
    　前記光変形部材と連結され該光変形部材によって変形させられる変形蓋と、
    　屈折率を整合させて外部から導入される外部光を透過させる屈折率整合剤と、
    　前記変形蓋で蓋をされ前記屈折率整合剤で満たされた整合剤溜まり部と、
    　前記整合剤溜まり部から前記屈折率整合剤を導出する先端が開口された経路と、
    　前記経路に対して傾斜し前記外部光を伝搬させる第１光導波路と、
    　前記経路を挟んで前記第１光導波路の延長線上に配置され前記経路を透過して来た前記外部光を出力する第２光導波路と
    　を備えることを特徴とする光演算素子。     A light deforming member that deforms according to the intensity of input light,
    A deformable lid connected to the light deformable member and deformed by the light deformable member;
    A refractive index matching agent that matches the refractive index and transmits external light introduced from the outside,
    A matching agent reservoir covered with the deformable lid and filled with the refractive index matching agent;
    A path in which a tip for leading out the refractive index matching agent from the matching agent reservoir is opened,
    A first optical waveguide that is inclined with respect to the path and propagates the external light;
    A second optical waveguide which is arranged on an extension of the first optical waveguide with the path interposed therebetween and outputs the external light transmitted through the path.
  2. 　前記第１光導波路の前記経路に対する入射角度と同じ角度の反射角度を持ち、先端部を前記第１光導波路の先端部に接触させる第３光導波路を
    　備えることを特徴とする請求項１に記載の光演算素子。     The third optical waveguide having a reflection angle that is the same as the incident angle of the first optical waveguide with respect to the path and having a tip portion in contact with the tip portion of the first optical waveguide. Optical computing element.
  3. 　入力光の強度が最大の場合に前記外部光は遮光される
    　ことを特徴とする請求項１又は２に記載の光演算素子。     The optical operation element according to claim 1, wherein the external light is blocked when the intensity of the input light is maximum.
  4. 　請求項１乃至３の何れかに記載した光演算素子をＮ（Ｎ≧2）個縦続接続させた多層ニューラルネットワークであって、
    　ｎ（ｎ＝2,3,…,Ｎ）層目の光演算素子の前記入力光は、ｎ－１層目の光演算素子の出力光を含む
    　ことを特徴とする多層ニューラルネットワーク。     A multilayer neural network in which N (N ≧ 2) optical operation elements according to any one of claims 1 to 3 are cascade-connected,
    The multi-layer neural network according to claim 1, wherein the input light of the nth (n = 2, 3, ..., N) -th layer optical arithmetic element includes the output light of the n−1-th layer optical arithmetic element.
  5. 　ｎ（ｎ＝2,3,…,Ｎ）層目の光演算素子とｎ＋１層目の光演算素子の間に、n個目の光演算素子の出力光を、分岐させるカプラー又は変調させる光フルターが配置される
    　ことを特徴とする請求項４に記載の多層ニューラルネットワーク。
          An optical filter that branches or modulates the output light of the n-th optical operation element between the n (n = 2, 3, ..., N) -th optical operation element and the n + 1-th optical operation element. Are arranged. The multilayer neural network according to claim 4, wherein
PCT/JP2019/043736 2018-11-21 2019-11-07 Optical arithmetic operation element and multi-layered neural network WO2020105459A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2018217931A JP2020086049A (en) 2018-11-21 2018-11-21 Optical computing element and multilayer neural network
JP2018-217931 2018-11-21

Publications (1)

Publication Number Publication Date
WO2020105459A1 true WO2020105459A1 (en) 2020-05-28

Family

ID=70773417

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2019/043736 WO2020105459A1 (en) 2018-11-21 2019-11-07 Optical arithmetic operation element and multi-layered neural network

Country Status (2)

Country Link
JP (1) JP2020086049A (en)
WO (1) WO2020105459A1 (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH05303048A (en) * 1992-04-27 1993-11-16 Fujikura Ltd Optical path switching device
US5978527A (en) * 1997-10-20 1999-11-02 Hewlett-Packard Company Thermal optical switches for light
JP2003161895A (en) * 2001-09-17 2003-06-06 Agilent Technol Inc Total internal reflection optical switch
US6999221B1 (en) * 2003-11-17 2006-02-14 Alabama A&M University Bimorphic polymeric photomechanical actuator
US20060088268A1 (en) * 2002-09-23 2006-04-27 Doron Nevo Optical micro-actuator
JP2012099431A (en) * 2010-11-05 2012-05-24 Konica Minolta Holdings Inc Lighting system and light emitting panel

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH05303048A (en) * 1992-04-27 1993-11-16 Fujikura Ltd Optical path switching device
US5978527A (en) * 1997-10-20 1999-11-02 Hewlett-Packard Company Thermal optical switches for light
JP2003161895A (en) * 2001-09-17 2003-06-06 Agilent Technol Inc Total internal reflection optical switch
US20060088268A1 (en) * 2002-09-23 2006-04-27 Doron Nevo Optical micro-actuator
US6999221B1 (en) * 2003-11-17 2006-02-14 Alabama A&M University Bimorphic polymeric photomechanical actuator
JP2012099431A (en) * 2010-11-05 2012-05-24 Konica Minolta Holdings Inc Lighting system and light emitting panel

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
OTSUKA, TAKUYA ET AL.: "Non-official translation: Neuron Element for a Deep Neural Network Passively Driven by External Light", PROCEEDINGS OF THE 2018 IEICE GENERAL CONFERENCE (2) ; THE INSTITUTE OF ELECTRONICS, INFORMATION AND COMMUNICATION ENGINEERS (IEICE), 6 March 2018 (2018-03-06), pages S-76 - S-77 *

Also Published As

Publication number Publication date
JP2020086049A (en) 2020-06-04

Similar Documents

Publication Publication Date Title
Kitayama et al. Novel frontier of photonics for data processing—Photonic accelerator
JP2019523932A (en) Apparatus and method for optical neural networks
US3814498A (en) Integrated optical circuit devices employing optical gratings
TWI287917B (en) Optical device and circuit using phase modulation and related methods
Huang et al. Programmable matrix operation with reconfigurable time-wavelength plane manipulation and dispersed time delay
JP6882686B2 (en) Optical computing elements and multi-layer neural networks
Ishihara et al. An optical neural network architecture based on highly parallelized WDM-multiplier-accumulator
Schmidt et al. Suspended mid-infrared waveguides for Stimulated Brillouin Scattering
WO2020105459A1 (en) Optical arithmetic operation element and multi-layered neural network
EP4042332A2 (en) Optical neuron unit and network of the same
WO2020225873A1 (en) Optical operation element and multilayer neural network
Zhang et al. Folded Digital Meta-Lenses for on-Chip Spectrometer
WO2020240623A1 (en) Optical computing element and multi-layer neural network
CN113392965B (en) Hadamard product realization method, device and storage medium
JP6534987B2 (en) Perturbation learning device and its method
Verber et al. Pipelined polynomial processors implemented with integrated optical components
Magnusson et al. Dispersion engineering with leaky-mode resonant photonic lattices
Winnie et al. Stable all-optical limiting in nonlinear periodic structures. III. Nonsolitonic pulse propagation
Wang et al. Thin metal superlens imaging in nanolithography
US5121249A (en) Fiber optic computational network utilizing optical and/or gate employing non-linear organic polymers and photovoltaic/piezoelectric optical interfaces
Estruch et al. Mason’s rule and Signal Flow Graphs applied to subwavelength resonant structures
Hattori et al. Optical-electronic implementation of artificial neural network for ultrafast and accurate inference processing
MacFarlane et al. Extended lattice filters enabled by four-directional couplers
WO2021205547A1 (en) Optical signal processing device
Kojima et al. Application of deep learning for nanophotonic device design

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19887045

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19887045

Country of ref document: EP

Kind code of ref document: A1