JPH077315B2 - Optical euro computer - Google Patents
Optical euro computerInfo
- Publication number
- JPH077315B2 JPH077315B2 JP19754788A JP19754788A JPH077315B2 JP H077315 B2 JPH077315 B2 JP H077315B2 JP 19754788 A JP19754788 A JP 19754788A JP 19754788 A JP19754788 A JP 19754788A JP H077315 B2 JPH077315 B2 JP H077315B2
- Authority
- JP
- Japan
- Prior art keywords
- light
- element array
- positive
- optical
- computer
- Prior art date
- Legal status (The legal status 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 status listed.)
- Expired - Lifetime
Links
Description
【発明の詳細な説明】 〔産業上の利用分野〕 この発明は,生物の神経回路網を模倣し,連想機能,パ
ターン認識機能などを有するコンピユータ(ニユーロコ
ンピユータ)を光技術を用いて達成する光ニユーロコン
ピユータに関するものである。DETAILED DESCRIPTION OF THE INVENTION [Industrial field of application] The present invention achieves a computer (nieuro computer) having an associative function, a pattern recognition function, and the like by imitating a neural network of a living being by using optical technology. It is related to the optical euro computer.
第4図は例えば刊行物(電子情報通信学会光・量子エレ
クトロニクス研究会資料OQE87−174(1988))に示され
た従来の光連想ニユーロコンピユータの構成図である。
図において,(1a),(1b)は発光ダイオードアレイ,
(2a),(2b)は光学マスク,(3a),(3b)は受光素
子アレイ,(4)は差動増幅器,(5)は比較器,
(6)は不完全情報入力,(7)は完全情報出力であ
る。FIG. 4 is a block diagram of a conventional opto-associative neutron computer shown in, for example, a publication (Material OQE87-174 (1988) of the Institute of Optics and Quantum Electronics of the Institute of Electronics, Information and Communication Engineers).
In the figure, (1a) and (1b) are light emitting diode arrays,
(2a) and (2b) are optical masks, (3a) and (3b) are light receiving element arrays, (4) is a differential amplifier, (5) is a comparator,
(6) is incomplete information input, and (7) is complete information output.
次に動作について説明する。第4図は光連想ニユーロコ
ンピユータの基本構成を示したものである。発光ダイオ
ードアレイ(1a),(1b)から扇状の光を光学マスク
(2a),(2b)へ照射する。発光ダイオードアレイ(1
a),(1b)はそれぞれの要素が点燈するか消燈するか
に対応して,1か0の状態を表わす一種のベクトル を表わしている。光学マスク(2a),(2b)はn×nの
要素に分けられており、その各要素の光透過率が異なつ
ている一種のマトリクス を表わし,受光素子アレイ(3b),(3b)は一種のベク
トル を表わしている。j番目の発光ダイオードで光学マスク
のj行目を照射し,光マスクのi列目の光をi番目を受
光素子で受けると, なるベクトル,マトリクス演算が行なえる。Next, the operation will be described. Fig. 4 shows the basic configuration of the optical association model computer. The optical masks (2a) and (2b) are irradiated with fan-shaped light from the light emitting diode arrays (1a) and (1b). Light-emitting diode array (1
a) and (1b) are a kind of vector representing the state of 1 or 0 depending on whether each element is turned on or off. Is represented. The optical masks (2a) and (2b) are divided into n × n elements, and each element has a different light transmittance. , And the photo detector array (3b), (3b) is a kind of vector Is represented. When the j-th light emitting diode irradiates the j-th row of the optical mask and receives the light of the i-th column of the photomask at the i-th light receiving element, Vector and matrix operations can be performed.
ニユーロコンピユータでは各ニユーロン間の結合強度に
情報を蓄積している。この光連想ニユーロコンピユータ
では光学マスクの透過率 に情報を蓄積している。その情報蓄積規則はHop field
モデルに基いたものであり、 で与えられる。Tijは正の値も負の値もとり得るが,光
学的に負の値は扱えないため,第3図に示すようにTij
の成分の正値Tij(+)と負値Tij(-)に対応して光学系を2
系統(Ui(+),Ui(-))作製し,その差 Ui=Ui(+)−Ui(-) を差動増幅器(4)で得ている。この出力信号を比較器
(5)でしきい値処理 Vi=θ(Ui) 但し, した後,発光ダイオードアレイ(1a),(1b)にフイー
ドバツクしている。In the euro computer, information is stored in the bond strength between each euro. In this photo-associative neutronics computer, the transmittance of the optical mask Is accumulating information in. The information storage rule is Hop field
Based on the model, Given in. Tij can be either positive or negative, but it is not possible to handle negative values optically.
2 of the optical system corresponding to the positive value Tij (+) and negative value Tij (-) of the component of
A system (Ui (+) , Ui (-) ) was produced and the difference Ui = Ui (+) -Ui (-) was obtained by the differential amplifier (4). This output signal is thresholded by the comparator (5) Vi = θ (Ui) After that, the light emitting diode arrays (1a) and (1b) are fed back.
このようにすると,例えば光学マスク(2a),(2b)に
アルフアベツトA,J,Eに相当する3種類の情報を蓄積し
ておけば,最初に発光ダイオード(1a),(1b)に不完
全な情報,例えばA′を入力しても,フイードバツクを
繰り返すうちに,その出力は入力情報A′に最も近い情
報Aを完全出力として表示する。By doing this, for example, if three types of information corresponding to the alpha-vegets A, J, and E are stored in the optical masks (2a) and (2b), the imperfections in the light-emitting diodes (1a) and (1b) will occur first. Inputting such information, for example, A ', the output displays the information A closest to the input information A'as a complete output while the feedback back is repeated.
このことを別の言葉で言えば,系のエネルギが蓄積情報
A,J,Eで極小値を取り,不完全情報が与えられると,そ
の近くのエネルギ極小値を取るように系全体が変化して
行く(発光ダイオードアレイの点滅状態を変化させるこ
とによる)という事である。In other words, the energy of the system is the stored information.
When the minimum value is taken at A, J, and E, and the incomplete information is given, the entire system changes so as to take the local minimum value of the energy (by changing the blinking state of the light emitting diode array). It is a thing.
従来の光ニユーロコンピユータは以上のように構成され
ているので、Tijマトリクスの正値,負値に対応して光
学系を2系統必要とし,系が複雑であるという問題があ
つた。Since the conventional optical neuro-computer is configured as described above, there is a problem that the system is complicated because it requires two optical systems corresponding to the positive and negative values of the Tij matrix.
この発明は上記のような問題点を解消するためになされ
たもので,光学系を1系統で構成できる光ニユーロコン
ピユータを得ることを目的とする。The present invention has been made in order to solve the above problems, and an object thereof is to obtain an optical euro-computer which can be configured with one optical system.
この発明に係る光ニユーロコンピユータは,ニユーロン
の状態ベクトルに応じて点滅する発光素子アレイ,記憶
情報の正値,負値に対してそれぞれ異なつた周波数もし
くは位相,または上記正値,負値に対して交互の時分割
により多重化して駆動することにより相関行列を実現
し,上記発光素子アレイからの光を変調させて上記ニユ
ーロンの状態ベクトルと相関行列との積信号光を透過す
る空間光変調器,この空間光変調器からの透過光を受信
し上記積信号の和を得る受光素子アレイ,この受光素子
アレイの出力信号を弁別して上記正値,負値に相当する
信号を得る信号処理器,この信号処理器で得られた正
値,負値に相当する信号の差を得る差動増幅器,および
この差動増幅器で得られた差の値をしきい値処理して上
記発光素子アレイにフイールドバツクする比較器を備え
たものである。The optical neuro-computer according to the present invention includes a light-emitting element array that blinks according to the state vector of the neuron, different frequencies or phases for positive and negative values of stored information, or for the positive and negative values. A spatial light modulator that realizes a correlation matrix by multiplexing and driving by alternating time division, modulates the light from the light emitting element array, and transmits the product signal light of the state vector of the Nyron and the correlation matrix. A light receiving element array for receiving the transmitted light from the spatial light modulator to obtain the sum of the product signals, a signal processor for discriminating output signals of the light receiving element array to obtain signals corresponding to the positive and negative values, A differential amplifier that obtains the difference between the signals corresponding to the positive and negative values obtained by this signal processor, and the difference value obtained by this differential amplifier are threshold-processed to the light emitting element array. Those having a comparator for Rudobatsuku.
この発明における空間光変調器は,Tijマトリクスの正
値,負値が多重化されて表現されているので,空間光変
調器の透過光を電気信号に変換した後,分離することに
より,光学系は一系統でよい。The spatial light modulator according to the present invention is represented by multiplexing positive and negative values of the Tij matrix. Therefore, after converting the transmitted light of the spatial light modulator into an electric signal, it is separated into an optical system. One line is enough.
以下,この発明の一実施例を図について説明する。第1
図において,(1)は発光素子アレイすなわち発光ダイ
オードアレイ,(3)は受光素子アレイ,(4)は差動
増幅器,(5)は比較器,(10)は空間光変調器,(1
1)は信号処理器すなわちバンドパスフイルタ,(12)
は整流器,(13)は空間光変調器のドライバである。An embodiment of the present invention will be described below with reference to the drawings. First
In the figure, (1) is a light emitting element array, that is, a light emitting diode array, (3) is a light receiving element array, (4) is a differential amplifier, (5) is a comparator, (10) is a spatial light modulator, and (1)
1) is a signal processor or bandpass filter, (12)
Is a rectifier and (13) is a spatial light modulator driver.
次に動作について説明する。基本原理は従来例と同じで
ある。本実施例では光学マスタの代わりに,外部信号に
て変調可能な空間光変調器(10)を用いている。空間光
変調器(10)は第3図に示すように、平面的に分布した
光を入射し、2次元的に光透過量を変調するものであ
る。第3図において、黒い部分は光強度がゼロ(すなわ
ち、全く光が通らない)であり、白い部分は光が全部通
過しており、中間色(ドッド、ハッチング、クロスハッ
チングで順に濃くなる様子を模式的に示している)は一
部光が通過していることを示している。このような空間
光変調器(10)として例えば液晶ディスプレイを使用す
ることができるが、これに限るものではなく、光磁気効
果を利用したものなど、他のどのような材料、原理を用
いたものであっても等しく適用することができる。Next, the operation will be described. The basic principle is the same as the conventional example. In this embodiment, a spatial light modulator (10) that can be modulated by an external signal is used instead of the optical master. As shown in FIG. 3, the spatial light modulator (10) is a device that receives light distributed in a plane and modulates the amount of light transmission two-dimensionally. In Fig. 3, the black part has zero light intensity (that is, no light passes through it), and the white part has all of the light passing through it, showing the intermediate colors (dots, hatching, cross-hatching, and darkening sequentially). Indicates that some light is passing through. For example, a liquid crystal display can be used as such a spatial light modulator (10), but it is not limited to this, and any other material or principle such as one utilizing a magneto-optical effect is used. Can be equally applied.
空間光変調器(10)では各単位毎に光強度を変調するこ
とができる。この単位のことを画素と呼んでおり、第3
図では各々の小さな四角が画素に相当している。各画素
は別々の電圧を印加することによって独立に駆動するこ
とができる。例えば、正弦波状の電圧を印加すれば、そ
れに応じて透過光強度は正弦波状に変化する。従って、
記憶情報の正値,負値に対するTijマトリクスの成分を
それぞれTij (+)、Tij (-)とし、それぞれを角周波数ω
1、ω2で変調した信号の和を で表わす。このようにすれば空間光変調器(10)のi行
目j列の画素を正と負の要素で多重化することができ
る。j番目の発光ダイオード(1)で空間光変調器(1
0)のj行目を照射し,空間光変調器(10)のi列目の
光をi番目の受光素子(3)で受けると なる信号が得られる。この信号をバンドパスフイルタ
(10)を通して,角周波数ω1とω2の成分を分離した
後,整流器(12)で整流する。The spatial light modulator (10) can modulate the light intensity for each unit. This unit is called a pixel,
In the figure, each small square corresponds to a pixel. Each pixel can be driven independently by applying different voltages. For example, when a sinusoidal voltage is applied, the intensity of the transmitted light changes in a sinusoidal manner accordingly. Therefore,
The components of the T ij matrix for the positive and negative values of the stored information are T ij (+) and T ij (-) , respectively, and the respective angular frequencies ω
The sum of the signals modulated by 1 and ω2 Express with. By doing so, the pixel in the i-th row and the j-th column of the spatial light modulator (10) can be multiplexed with positive and negative elements. The spatial light modulator (1
When the j-th row of 0) is irradiated and the light of the i-th column of the spatial light modulator (10) is received by the i-th light receiving element (3) Signal is obtained. This signal is passed through a bandpass filter (10) to separate the angular frequency components ω 1 and ω 2 , and then rectified by a rectifier (12).
すなわち, なる信号が得られる。これらの信号を差動増幅器(4)
で引算して Ui=Ui(+)−Ui(-) なる信号を得,これを比較器(5)でしきい値処理し,
発光ダイオードアレイ(1)へフイードバツクする。That is, Signal is obtained. These signals are applied to the differential amplifier (4)
Subtracted by to obtain the signal Ui = Ui (+) −Ui (-) , which is thresholded by the comparator (5),
Feed back to the light emitting diode array (1).
なお,上記実施例では,空間光変調器(10)を周波数多
重化信号で駆動するものを示したが,同一周波数で位相
の異なる信号で,例えばTijの正値はTij(+)cosωt,負値
はTij(-)sinωtのように駆動しても良い。このとき,Ti
jの正値,負値は位相によつて弁別できる(位相変調
法)。Although the spatial light modulator (10) is driven by the frequency-multiplexed signal in the above embodiment, signals having the same frequency but different phases, for example, a positive value of Tij is Tij (+) cos ωt, a negative value. The value may be driven as Tij (-) sin ωt. At this time, Ti
The positive and negative values of j can be distinguished by the phase (phase modulation method).
また,時分割多重化してもよい。その例を第2図に示
す。Alternatively, time division multiplexing may be performed. An example thereof is shown in FIG.
図において,(1)は発光ダイオードアレイ,(3)は
受光素子アレイ,(4)は差動増幅器,(5)は比較
器,(10)は空間光変調器,(14a),(14b)は空間光
変調器ドライバ,(15a)(15b)はパルス発生器,(16
a)(16b)はサンプルホールドアンプである。In the figure, (1) is a light emitting diode array, (3) is a light receiving element array, (4) is a differential amplifier, (5) is a comparator, (10) is a spatial light modulator, (14a), (14b). Is a spatial light modulator driver, (15a) (15b) is a pulse generator, (16
a) and (16b) are sample and hold amplifiers.
空間光変調器(10)はそのドライバ(14a)(14b)を通
じて,パルサ(15a)(15b)によつてTij(+)とTij(-)信
号で交互にパルス駆動される。このとき受光素子(3)
出力を空間光変調器(10)駆動信号パルスに同期させて
サンプルホールドすることにより,Tijマトリクスの正値
Tij(+),負値Tij(-)を分離できる。後の動作は第1図の
実施例と同じである。The spatial light modulator (10) is pulse-driven by the pulsers (15a) and (15b) through the drivers (14a) and (14b) alternately with the Tij ( +) and Tij (-) signals. At this time, the light receiving element (3)
Positive value of Tij matrix is obtained by sampling and holding the output in synchronization with the spatial light modulator (10) drive signal pulse.
Tij (+) and negative value Tij (-) can be separated. The subsequent operation is the same as that of the embodiment shown in FIG.
以上のように,この発明によれば,ニユーロンの状態ベ
クトルに応じて点滅する発光素子アレイ,記憶情報の正
値,負値に対してそれぞれ異なつた周波数もしくは位
相,または上記正値,負値に対して交互の時分割により
多重化して駆動することにより相関行列を実現し,上記
発光素子アレイからの光を変調させて上記ニユーロンの
状態ベクトルと相関行列との積信号光を透過する空間光
変調器,この空間光変調器からの透過光を受信し上記積
信号の和を得る受光素子アレイ,この受光素子アレイの
出力信号を弁別して上記正値,負値に相当する信号を得
る信号処理器,この信号処理器で得られた正値,負値に
相当する信号の差を得る差動増幅器,およびこの差動増
幅器で得られた差の値をしきい値処理して上記発光素子
アレイにフイードバツクする比較器を備えたので,光学
系が一系統でよく,装置が簡単となり,大規模化が容易
となる効果がある。As described above, according to the present invention, the light emitting element array that blinks according to the state vector of the euronne, the frequency or phase different from the positive value and the negative value of the stored information, or the positive value and the negative value described above. On the other hand, a spatial light modulation that realizes a correlation matrix by multiplexing and driving by alternating time division, modulates the light from the light emitting element array, and transmits the product signal light of the state vector of the Nyron and the correlation matrix. Device, a light receiving element array for receiving transmitted light from the spatial light modulator to obtain the sum of the product signals, and a signal processor for discriminating output signals of the light receiving element array to obtain signals corresponding to the positive and negative values , A differential amplifier for obtaining a difference between signals corresponding to a positive value and a negative value obtained by the signal processor, and threshold value processing of the difference value obtained by the differential amplifier to the light emitting element array. Feedback Because with a comparator for click, the optical system is better in one system, device is simple, there is an effect that it is easy to scale-up.
第1図はこの発明の一実施例による光ニユーロコンピユ
ータを示す構成図,第2図ははこの発明の他の実施例に
よる光ニユーロコンピユータを示す構成図,第3図はこ
の発明に係る空間光変調器の動作を説明する説明図、第
4図は従来の光ニユーロコンピユータを示す構成図であ
る。 図において,(1),(1a),(1b)は発光ダイオード
アレイ,(2a),(2b)は光学マスク,(3),(3
a),(3b)は受光素子アレイ,(4)は差動増幅器,
(5)は比較器,(6)は不完全情報入力,(7)は完
全情報出力,(10)空間光変調器,(11)はバンドパス
フイルタ,(12)は整流器,(13),(14a),(14b)
はドライバ,(15a),(15b)はパルス発生器,(16
a),(16b)はサンプルホールドアンプである。 なお、各図中同一符号は同一または相当部分を示すもの
とする。1 is a block diagram showing an optical euro computer according to an embodiment of the present invention, FIG. 2 is a structural diagram showing an optical euro computer according to another embodiment of the present invention, and FIG. 3 is related to the present invention. FIG. 4 is an explanatory diagram for explaining the operation of the spatial light modulator, and FIG. 4 is a configuration diagram showing a conventional optical neuro-computer. In the figure, (1), (1a) and (1b) are light emitting diode arrays, (2a) and (2b) are optical masks, and (3) and (3
a) and (3b) are light receiving element arrays, (4) is a differential amplifier,
(5) is a comparator, (6) is incomplete information input, (7) is complete information output, (10) spatial light modulator, (11) is bandpass filter, (12) is rectifier, (13), (14a), (14b)
Is a driver, (15a), (15b) is a pulse generator, and (16
a) and (16b) are sample and hold amplifiers. In the drawings, the same reference numerals indicate the same or corresponding parts.
Claims (1)
積和演算を行う光ニユーロコンピユータにおいて,上記
ニユーロンの状態ベクトルに応じて点滅する発光素子ア
レイ, 記憶情報の正値,負値に対してそれぞれ異なつた周波数
もしくは位相,または上記正値,負値に対して交互の時
分割により多重化して駆動することにより上記相関行列
を実現し,上記発光素子アレイからの光を変調させて上
記ニユーロンの状態ベクトルと相関行列との積信号光を
透過する空間光変調器, この空間光変調器からの透過光を受信し上記積信号の和
を得る受光素子アレイ, この受光素子アレイの出力信号を弁別して上記正値,負
値に相当する信号を得る信号処理器, この信号処理器で得られた正値,負値に相当する信号の
差を得る差動増幅器. およびこの差動増幅器で得られた差の値をしきい値処理
して上記発光素子アレイにフイードバツクする比較器を
備えたことを特徴とする光ニユーロコンピユータ。1. In an optical neutron computer for performing a product-sum operation of a neutron state vector and a correlation matrix, a light emitting element array blinking according to the neutron state vector, and positive and negative values of stored information. The above-mentioned correlation matrix is realized by multiplexing and driving the different frequencies or phases, or the positive and negative values by alternating time division, and the light from the light-emitting element array is modulated to realize the above-mentioned A spatial light modulator that transmits a product signal light of a state vector and a correlation matrix, a light receiving element array that receives transmitted light from the spatial light modulator and obtains the sum of the product signals, and an output signal of the light receiving element array Separately, a signal processor that obtains the signals corresponding to the positive and negative values, and a differential amplifier that obtains the difference between the signals corresponding to the positive and negative values obtained by this signal processor. An optical euro-computer which is provided with a comparator for thresholding the value of the difference obtained by the differential amplifier and feeding back to the light emitting element array.
Priority Applications (4)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP19754788A JPH077315B2 (en) | 1988-08-08 | 1988-08-08 | Optical euro computer |
| GB8915308A GB2220780B (en) | 1988-07-05 | 1989-07-04 | Neurocomputer |
| US07/375,813 US5095459A (en) | 1988-07-05 | 1989-07-05 | Optical neural network |
| DE3922129A DE3922129C2 (en) | 1988-07-05 | 1989-07-05 | Neurocomputer |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP19754788A JPH077315B2 (en) | 1988-08-08 | 1988-08-08 | Optical euro computer |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| JPH0245815A JPH0245815A (en) | 1990-02-15 |
| JPH077315B2 true JPH077315B2 (en) | 1995-01-30 |
Family
ID=16376298
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| JP19754788A Expired - Lifetime JPH077315B2 (en) | 1988-07-05 | 1988-08-08 | Optical euro computer |
Country Status (1)
| Country | Link |
|---|---|
| JP (1) | JPH077315B2 (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH05158097A (en) * | 1991-12-06 | 1993-06-25 | Hitachi Ltd | Parallel light converter |
| US11398871B2 (en) | 2019-07-29 | 2022-07-26 | Lightmatter, Inc. | Systems and methods for analog computing using a linear photonic processor |
-
1988
- 1988-08-08 JP JP19754788A patent/JPH077315B2/en not_active Expired - Lifetime
Also Published As
| Publication number | Publication date |
|---|---|
| JPH0245815A (en) | 1990-02-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| GB2220780A (en) | Neurocomputer | |
| Xu et al. | 1000 fps computational ghost imaging using LED-based structured illumination | |
| US20210003498A1 (en) | Dynamic high-speed high-sensitivity imaging device and imaging method | |
| DE69816876T2 (en) | IMPROVEMENTS REGARDING PATTERN RECOGNITION | |
| WO2019186548A1 (en) | Optical neural network unit and optical neural network configuration | |
| Lu et al. | Generation and characterization of ultrabroadband polarization–frequency hyperentangled photons | |
| JPS5939687B2 (en) | Color separation method and device for pattern creation using light spectral distribution | |
| Xue et al. | Scattering reduction by structured light illumination in line-scanning temporal focusing microscopy | |
| Jiang et al. | Fast hyperspectral single-pixel imaging via frequency-division multiplexed illumination | |
| CN105388486A (en) | Ghost imaging system and imaging method based on fiber array pseudo-thermal light | |
| Nikolenko et al. | Spatial light modulator microscopy | |
| US20250021127A1 (en) | Optical Computing Device and Computing Method | |
| Crombez et al. | Computational hyperspectral light-sheet microscopy | |
| JPH077315B2 (en) | Optical euro computer | |
| US20170059483A1 (en) | Laser microscope and microscopy method | |
| Deutsch et al. | Multichannel single-output color pattern recognition by use of a joint-transform correlator | |
| Li et al. | High-speed two-photon microscopy with adaptive line-excitation | |
| Wu et al. | Only-train-electrical-to-optical-conversion (OTEOC): simple diffractive neural networks with optical readout | |
| Zhao et al. | Multi-input and multi-objective design for a mini-LED LCD integrated with MIMO visible light communication | |
| CN109718476A (en) | 3-D scanning two-photon stimulating system and its stimulating method | |
| CN109253997A (en) | A kind of Raman computed tomography (SPECT) system based on frequency modulation(PFM) and space encoding | |
| JP2786275B2 (en) | Optical neural network optical system | |
| Hoshi et al. | Single-pixel imaging for edge images using deep neural networks | |
| DE69431377T2 (en) | Self-organizing pattern learning system | |
| JPH06161586A (en) | Optical neural processor |