US20040107172A1 - Optical pulse-coupled artificial neurons - Google Patents
Optical pulse-coupled artificial neurons Download PDFInfo
- Publication number
- US20040107172A1 US20040107172A1 US09/963,971 US96397101A US2004107172A1 US 20040107172 A1 US20040107172 A1 US 20040107172A1 US 96397101 A US96397101 A US 96397101A US 2004107172 A1 US2004107172 A1 US 2004107172A1
- Authority
- US
- United States
- Prior art keywords
- optical
- input signal
- neuron
- intensity
- artificial neuron
- 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.)
- Granted
Links
Images
Classifications
-
- 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
- G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
Definitions
- This invention relates to artificial neurons and, more particularly, to pulse coupled optical neurons and a method for making optical pulse-coupled artificial neurons.
- pulse-coupled neural networks that is, a neural network that simulates the neural network found in biological systems.
- the inputs to each neuron are weighted and summed. This sum is then compared with a threshold value. If the sum of the neural input signals exceeds the threshold value, a pulse signal will be produced by a pulse generator.
- an artificial neuron operable for receiving an optical input signal having an intensity and providing an optical output signal comprising a sequence of optical pulses having a pulse output frequency that depends on the intensity of the input signal
- the artificial neuron comprising: (a) an input signal integrator operable for receiving and summing a plurality of optical input signals to provide an integrated electrical or optical input signal having an integrated signal intensity; and (b) an optical pulse generator operable for receiving said integrated electrical or optical input signal from said input signal integrator and provide optical output pulses having a frequency that is determined by said integrated signal intensity.
- an optical neural network comprising a plurality of artificial neurons, each artificial neuron being operable for receiving an optical input signal having an intensity and providing an optical output signal comprising a sequence of optical pulses having a pulse output frequency that depends on said intensity, the artificial neuron further comprising: (a) an input signal integrator operable for receiving and summing a plurality of optical input signals to provide an integrated electrical or optical input signal having an integrated signal intensity; and (b) an optical pulse generator operable for receiving said integrated electrical or optical input signal from said input signal integrator and provide optical output pulses having a frequency that is determined by said integrated signal intensity.
- an artificial neuron comprising an optoelectronic circuit
- the artificial neuron comprising an optoelectronic circuit being adapted to be coupled with other similar artificial neurons by means of optical pulses.
- FIG. 1 is a plan view of an optoelectronic embodiment of an artificial neuron in accordance with a first embodiment of the present invention.
- FIG. 2 illustrates the relationship between the pulse frequency and waveform of the output signal of the optoelectronic neuron of FIG. 1 in response to a change in the intensity of an input signal.
- FIG. 3 is a graph illustrating the relationship between the pulse output frequency of an optoelectronic neuron in accordance with FIG. 1 and the sum of pulse amplitudes of a plurality of optical signals comprising the input signal.
- FIG. 4 is a plan view of an embodiment of an optoelectronic neuron in accordance with the present invention wherein the signal input port comprises a pair of photosensitive elements, one of which is stimulatory, and the other inhibitory, in response to an input signal.
- FIG. 5 is a schematic diagram of the optoelectronic embodiment of an artificial neuron shown in FIG. 1 wherein the frequency of optical pulses from an LED disposed in the signal output section of an astable electronic multivibrator is dependent on the intensity of light incident on a photosensitive element disposed to comprise an element in the RC time constant of the signal input section of the circuit.
- FIG. 6 is a schematic diagram of an astable multivibrator in accordance with FIG. 5, illustrating the operation of such a circuit when the photosensitive elements comprising the RC time constants of the circuit, as shown in FIG. 5, are replaced with fixed resistors.
- FIG. 7 is a schematic diagram of an optoelectronic neuron comprising an astable multivibrator that is similar to the astable multivibrator of FIG. 5, but wherein a second photosensitive element provides an inhibitory signal to the output LED that depends on the intensity of an inhibitory optical input signal.
- FIG. 8 is a perspective view of a two layer device comprised of a pulse coupled artificial neuron network employing an array of forward and rearward directed optoelectronic neurons, as shown in FIG. 1, disposed between two parallel mirrors.
- FIG. 9 a is a cross-sectional view of an optical pulse generator employing a nonlinear optical material disposed between two parallel mirrors that receives and sums a plurality of optical signal input pulses and generates a pulsed output signal having a pulse frequency that depends on the sum of the input signals.
- FIG. 9 b is a cross-sectional view of another embodiment of an optical pulse generator employing two nonlinear optical materials disposed between parallel mirrors to form a nonlinear Fabry-Perot etalon, as shown in FIG. 9 a .
- the device receives and sums a plurality of optical signal input pulses and generates a pulsed output signal having a pulse frequency that depends on the sum of the input signals.
- FIG. 10 is a cross-sectional view of a further embodiment of an optical pulse generator wherein a special light modulator comprising a sheet of a photosensitive material is disposed on one side of a nonlinear Fabry-Perot etalon such as the configuration shown in FIG. 9 a.
- an optoelectronic artificial neuron 10 is an astable circuit contained in two conductive tubes 11 and 12 .
- the conductive tubes 11 and 12 are joined with a nonconductive member 13 .
- the conductive tubes are connected to a power supply (not shown) having a voltage output of +V.
- a number of such neurons 10 are interconnected by a common conductive medium, as, for example, by stacking, they can share the same single power supply lines.
- Input signals 14 from an external source are collected by a photodetector 15 .
- the output pulses comprising the output signal 16 of the neuron 10 are provided by a light-emitting diode (LED) 17 , or a laser diode (LD).
- LED light-emitting diode
- LD laser diode
- the pulsation frequency is dependent on the light energy collected by the photodetector 15 .
- An example of the input and output pulse waveform is illustrated in FIG. 2.
- the input light signals 14 are summed over space or time.
- an output light pulse 16 will be generated. Every individual output pulse 16 comprising the output signal spikes to the same level as an all-or-nothing event. That is, once the positive-feedback cycle starts, nothing will stop the full amplitude spiking. If the threshold value is not reached, however, the slight increase in the output light intensity will subside when the stimulus is removed.
- Each output pulse 16 lasts only for a short period of time.
- the pulse frequency is a function of the input signal summation, either temporally or spatially.
- FIG. 3 shows the relationship between the output pulse frequency and the summation of the input signals.
- the optoelectronic neuron 10 can be modified to 11 include both excitatory and inhibitory synapses as illustrated in the artificial optoelectronic neuron shown at 40 in FIG. 4.
- the input port consists of two photodetectors, one photodetector 15 being excitatory, and the other photodetector 41 being inhibitory.
- Individual neurons 40 decide whether or not to fire an output pulse 16 by summing excitatory and inhibitory inputs. Signals received by the excitatory synapses make the neuron more likely to fire a pulse. Conversely, signals received by the inhibitory synapses render the neuron 40 less likely to fire.
- FIG. 5 An example of a first preferred embodiment of an optoelectronic neuron based on an astable multivibrator is illustrated in FIG. 5.
- the astable circuit consists of two transistors T 1 and T 2 , two capacitors C 1 and C 2 , three resistors R 1 , R 2 and R 3 , a photodetector PD, and a light-emitting diode LED.
- the photodetector is the input port (i.e., receptor) for the optoelectric neuron 10 .
- the light-emitting diode serves as the output port, emitting signals to be received at the synaptic inputs of neighboring neurons.
- the circuit 60 of FIG. 6 if the photodetector PD of FIG. 5, which may be a photosensitive resistor, a photodiode or a photosensitive transistor, is replaced with a resistor R 4 , then the circuit 60 will act as an ordinary astable multivibrator.
- Multivibrators are classified according to the number of steady (stable) states of the circuit. A steady state exists when circuit operation is essentially constant; that is, one transistor remains in conduction and the other remains “off” until an external signal is applied.
- the three types of multivibrators are the astable, monostable, and bistable.
- the astable circuit has no stable state. With no external signal applied, the transistors alternately switch from cutoff to saturation at a frequency determined by the RC time constants of the coupling circuits. In the circuit 60 illustrated in FIG. 6, durations of the two unstable periods are
- the oscillation frequency of the multivibrator circuit will be determined primarily by the light energy collected by the photodectector PD.
- output signal pulses 16 are produced if the input power reaches an activation threshold.
- the frequency of the light pulses emitted from the LED increases with an increase of input light power.
- the frequency of the pulses comprising the output signal 16 increases with the summation of energy contained in each pulse.
- FIG. 7 illustrates an optoelectronic neuron 40 containing both excitatory and inhibitory synapses.
- Input signals 14 collected by photodetector PD 1 can increase the output signal 16 pulsation frequency.
- inhibitory input signals 71 detected by PD 2 , can reduce the output signal pulsation frequency.
- a two-layer system can be assembled to simulate a neural network 80 as illustrated in FIG. 8.
- Forward oriented neurons 10 form the first layer
- some backward oriented neurons 81 in the assembly form the second layer.
- the input signal of a neuron 81 in the second layer is the light signal reflected from the partially reflective mirror 2 , which light signal originated from its neighboring neurons 10 in the first layer.
- synchronization and momental contrast enhancement can be performed.
- all the neurons are synchronized and pulse periodically.
- neurons in different areas are simultaneously exposed to light intensity at different levels, the neurons will pulse in groups.
- All-optical neurons can be constructed by advantageously using the regenerative pulsations in optical bistable devices.
- Regenerative pulsation in optical bistable devices is a suitable mechanism for realization of optical neurons due to the following properties:
- the pulsation has a distinctive threshold.
- the pulsation frequency can be increased by increasing of the power of the excitation light.
- Regenerative pulsation in optical bistable devices is an optical phenomenon, first observed by McCall, using a hybrid optical device (S. L. McCall, “Instability and regenerative pulsation phenomena in Fabry-Perot nonlinear optic media devices,” Appl. Phys. Lett., Vol. 32, 284(1978)). Regenerative pulsation can also take place in an all-optical bistable device, when the nonlinearity had two contributions of opposite sign and different time constants. When regenerative pulsation takes place, the bistable device turns into an astable device. As of the present date, this phenomenon has only been observed in several bistable devices using nonlinear materials such as GaAs, InSb, ZnS and liquid crystals.
- the pulsation frequency is relatively stable, but not dependent on the incident light intensity as desired for constructing optical neurons. Accordingly, these materials are not suitable candidates for all-optical neuron applications at the present time.
- the nonlinear material needs to have a response time constant that varies with the light intensity of light incident thereon. In certain materials, such as dye-doped liquid crystals and photorefractive crystals, the nonlinear response time constants are sensitive to the light intensity and these materials are, therefore suitable for building optical neurons.
- the pulse coupled, all-optical neuron in accordance with the present invention consists two elements: an input signal integrator and a bistable device, the latter being a pulse generator.
- the function of the integrator is to sum the input light pulses, and transmit a continuous wave light beam, the intensity of the CW beam fluctuating with the summation of the input signals.
- An example of an integrator suitable for use in an optical neuron application is a spatial light modulator comprising liquid crystals.
- the readout light intensity of such a spatial light modulator is a function of the incident light signal energies, with a response speed determined by the frequency of the LC driver.
- the pulse generator is a key element in the all-optical neuron of the present invention.
- the function of the pulse generator is to generate a train of pulses having a frequency that varies with the light intensity it receives from the integrator.
- the pulse generator is actually a nonlinear Fabry-Perot etalon.
- the optical behavior of the nonlinear medium contained in the etalon cavity generally depends on two nonlinear mechanisms. One mechanism tends to turn the device “on” at higher input light intensities, while the other (slower) mechanism tends to turn the device “off”.
- Nonlinear medium that can be adapted for disposition within the cavity of a Fabry-Perot etalon is a dye-doped liquid crystal, which has both a large thermo-optic coefficient and large orientational nonlinearity.
- thermal effects result in a change in the index of refraction of the medium, which can turn the device “on” automatically.
- the device is turned “on”, the light intensity build-up inside the etalon cavity triggers the orientational nonlinearity, which will turn the device “off”. After the device is turned “off”, the thermal effect will again dominate and the device turns “on”.
- the contrast ratio of the device is preferably maximized.
- the Fabry-Perot is basically a simple interferometer relying on the interference of multiply reflected beams.
- the simplest Fabry-Perot etalon consists of a pair of parallel mirrors. The incident beam of light undergoes multiple reflections between two parallel mirrors that define the cavity.
- T T 0 1 + F ⁇ ⁇ sin 2 ⁇ ( ⁇ ) ( 1 )
- F 4 ⁇ RD ( 1 - RD ) 2 ( 2 )
- T 0 D ⁇ ( 1 - R ) 2 ( 1 - RD ) 2 ( 3 )
- D exp ⁇ ( - ⁇ ⁇ ⁇ L - ⁇ ⁇ ⁇ L ) ( 4 )
- ⁇ 2 ⁇ ⁇ ⁇ ⁇ n ⁇ ⁇ L ⁇ + ⁇ o ′ ( 5 )
- ⁇ is the single-trip phase shift of the cavity
- R is the reflectivity of the mirrors
- T 0 is the peak transmission of the filter
- ⁇ is the absorption coefficient of the material
- ⁇ is the scattering coefficient
- L is the etalon cavity length
- ⁇ is the light wavelength
- ⁇ ′ o is a constant.
- the cavity loss is due to two main factors: (a) absorption by the dye molecules; and (b) scattering by the liquid crystal molecules. From the equations shown above, the contrast ratio of the device in the “on” and “off” states is
- a medium with low absorption and scattering is preferred. Since the photosensitivity of the material is dependent on the concentration of the doped dye, which can not be too low, the only way to enhance the F factor is to lower the scattering loss of the liquid crystal. When the liquid crystal molecules are well aligned in the cavity, high contrast ratio can be achieved.
- the use of vacuum deposited dielectric films as an alignment layer is a common practice. In experiments, it has been found that some dielectric mirrors may also be used as the alignment layers. In the film deposition process, if the substrate is tilted an angle to the evaporation source, the molecular orientation within the film will have a determined direction, which may be used to align the liquid crystal molecules.
- regenerative pulsation can also take place in a nonlinear Fabry-Perot etalon when the etalon has two layers of nonlinear materials 90 b disposed in the cavity, as illustrated in FIG. 9 b .
- the structure of the optical pulse generator 90 ′ is more advantageous than the optical pulse generator 90 because of the greater flexibility in the selection of the nonlinear materials 90 b disposed in the cavity of the etalon 91 . If only one material with two nonlinear mechanisms is used, such as indicated at 90 a in FIG. 9 a , the response properties of the mechanisms are usually fixed.
- ⁇ and ⁇ are the phase shifts of light in the two materials
- P i , P o are incident and transmitted light power
- ⁇ 1 , ⁇ 2 are the response time constants of the two nonlinear materials
- a, r are the nonlinear coefficients.
- ⁇ o and ⁇ 0 are the phase shifts under steady state
- ⁇ 0 ⁇ P i T( ⁇ 0 , ⁇ 0 ) (7)
- the nonlinear Fabry-Perot etalon turns from a bistable device into an astable device.
- the pulsation frequency can be a function of the incident intensity, satisfying the requirements for optical neuron operations.
- FIG. 10 shows such an all-optical neuron 100 having a laminate structure.
- the optical neuron 100 consists of a photoconductive layer 101 such as, for example, a layer of amorphous silicon, a pair of dielectric mirrors 102 , and a liquid crystal layer 103 .
- the mirrors 102 and the liquid crystal layer 103 form a nonlinear Fabry-Perot etalon 104 .
- the impedance of the photoconductive layer 101 is extremely high.
- the electric field in the liquid crystal layer which field is dependent on the intensity of light reaching either side of the photoconductive layer 101 , induces a phase change in the liquid crystal.
- the nonlinear Fabry-Perot etalon 104 can be switched from the reflection state to the transmission state as a result of the phase change in the liquid crystal layer 103 .
- the light intensity inside cavity may trigger nonlinearity due to the thermal effect and/or orientational effects inside the Fabry-Perot etalon, and switch the Fabry-Perot etalon from transmission state back to reflection state.
- This process continues ad infinitum, with a frequency depending on the intensity of the incident light signal.
- the shape of output pulses 108 produced by optical neuron 100 is opposite to the shape of the input pulses 105 . Accordingly, a converter such as a special light modulator is needed to transform the waveform.
- both the optoelectronic neurons 10 and 40 and all-optical neurons 90 and 100 may also be used as optical analog-to-digital converters.
- a large scale network can be constructed using an array 80 of such neurons as indicated in FIG. 8. Interconnection among these neurons can be performed with off-the-shelf optical components such as mirrors, lenses, gratings and optical fibers.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Neurology (AREA)
- Optical Modulation, Optical Deflection, Nonlinear Optics, Optical Demodulation, Optical Logic Elements (AREA)
- Optical Communication System (AREA)
Abstract
The present invention discloses low noise, optically coupled optoelectronic and all-optical artificial neuron devices that can be configured in an array to simulate the function of biological neural networks, and methods for making the artificial neurons. In a first optoelectronic embodiment, the device employs the regenerative pulsation property of astable multivibrators as optical pulse generators. Prior art pulse-coupled artificial neurons are subject to undesirable noise interference because the interconnection of such prior art neurons is based on electrical signals conducted through a grid of wires. The present invention obviates the need for hard-wired interconnection of individual neurons in order to cofigure the neurons into a network. In an optoelectronic embodiment, the neuron receives an optical input signal from an external source. A photosensitive detector, disposed in a circuit to control the state of an astable or bistable multivibrator, converts the intensity of the input light into a train of light pulses having a frequency that is a function of the intensity of the input signal. In an all-optical embodiment of an artificial neuron, an input signal is first integrated and the integrated signal transmitted to an optical pulse generator comprised of a nonlinear material disposed within the cavity of a Fabry-Perot etalon. The output of the etalon is a train of light pulses having a frequency that depends upon the intensity of the integrated input signal. When a weak light signal reaches the neuron's input port, there is no light pulse emitted from the output port. By contrast, a strong signal, or a group of weak signals, triggers a short-lived light pulse. The output pulse frequency is a function of the summed input signal power.
Description
- [0001] This work was supported by the Defense Advanced Research Projects Agency of the U.S. Government under Contract No. DAAH01-99-C-R202. The Government may have certain rights in the invention.
- 1. Field of the Invention
- This invention relates to artificial neurons and, more particularly, to pulse coupled optical neurons and a method for making optical pulse-coupled artificial neurons.
- 2. Prior Art
- Although electronic computers are powerful tools for digital computation and information processing, their pattern recognition and image processing capabilities are far from satisfactory. Recently, many new computation architectures have been explored, including pulse-coupled neural networks: that is, a neural network that simulates the neural network found in biological systems. In such neural networks, the inputs to each neuron are weighted and summed. This sum is then compared with a threshold value. If the sum of the neural input signals exceeds the threshold value, a pulse signal will be produced by a pulse generator.
- The essential element of most neural networks is the massive number of weighted interconnections used to combine relatively simple processing elements (neurons) together into a useful architecture. In prior art artificial neural network hardware architecture, interconnection among the neurons is realized using wires. The wiring is necessarily complicated when each neuron is connected to a large number of other neurons. The inherent parallelism and interconnection capability of optics make it an attractive means for the implementing the neural network interconnection process.
- It is an object of the invention to provide an artificial neuron operable for receiving an optical input signal having an intensity and providing an optical output signal comprising a sequence of optical pulses having a pulse output frequency that depends on the intensity of the input signal, the artificial neuron comprising: (a) an input signal integrator operable for receiving and summing a plurality of optical input signals to provide an integrated electrical or optical input signal having an integrated signal intensity; and (b) an optical pulse generator operable for receiving said integrated electrical or optical input signal from said input signal integrator and provide optical output pulses having a frequency that is determined by said integrated signal intensity.
- It is a further object of the invention to provide an optical neural network comprising a plurality of artificial neurons, each artificial neuron being operable for receiving an optical input signal having an intensity and providing an optical output signal comprising a sequence of optical pulses having a pulse output frequency that depends on said intensity, the artificial neuron further comprising: (a) an input signal integrator operable for receiving and summing a plurality of optical input signals to provide an integrated electrical or optical input signal having an integrated signal intensity; and (b) an optical pulse generator operable for receiving said integrated electrical or optical input signal from said input signal integrator and provide optical output pulses having a frequency that is determined by said integrated signal intensity.
- It is still a further object of the invention to provide an artificial neuron adapted for optical communication with other artificial neurons and a method for making an artificial neuron.
- It is yet a further object of the invention to provide an artificial neuron comprising an optoelectronic circuit, the artificial neuron comprising an optoelectronic circuit being adapted to be coupled with other similar artificial neurons by means of optical pulses.
- It is another object of the invention to provide an artificial, all-optical neuron comprising a nonlinear Fabry-Perot etalon, the all-optical artificial neuron being adapted to be coupled with similar artificial neurons by means of optical pulses, and a method for making an all-optical artificial neuron.
- It is another object of the invention to provide an artificial neuron network comprising an optically coupled array of artificial neurons adapted for applications in neural networks and/or image processing devices.
- The features of the invention believed to be novel are set forth with particularity in the appended claims. However the invention itself, both as to organization and method of operation, together with further objects and advantages thereof may be best understood by reference to the following description taken in conjunction with the accompanying drawings in which:
- FIG. 1 is a plan view of an optoelectronic embodiment of an artificial neuron in accordance with a first embodiment of the present invention.
- FIG. 2 illustrates the relationship between the pulse frequency and waveform of the output signal of the optoelectronic neuron of FIG. 1 in response to a change in the intensity of an input signal.
- FIG. 3 is a graph illustrating the relationship between the pulse output frequency of an optoelectronic neuron in accordance with FIG. 1 and the sum of pulse amplitudes of a plurality of optical signals comprising the input signal.
- FIG. 4 is a plan view of an embodiment of an optoelectronic neuron in accordance with the present invention wherein the signal input port comprises a pair of photosensitive elements, one of which is stimulatory, and the other inhibitory, in response to an input signal.
- FIG. 5 is a schematic diagram of the optoelectronic embodiment of an artificial neuron shown in FIG. 1 wherein the frequency of optical pulses from an LED disposed in the signal output section of an astable electronic multivibrator is dependent on the intensity of light incident on a photosensitive element disposed to comprise an element in the RC time constant of the signal input section of the circuit.
- FIG. 6 is a schematic diagram of an astable multivibrator in accordance with FIG. 5, illustrating the operation of such a circuit when the photosensitive elements comprising the RC time constants of the circuit, as shown in FIG. 5, are replaced with fixed resistors.
- FIG. 7 is a schematic diagram of an optoelectronic neuron comprising an astable multivibrator that is similar to the astable multivibrator of FIG. 5, but wherein a second photosensitive element provides an inhibitory signal to the output LED that depends on the intensity of an inhibitory optical input signal.
- FIG. 8 is a perspective view of a two layer device comprised of a pulse coupled artificial neuron network employing an array of forward and rearward directed optoelectronic neurons, as shown in FIG. 1, disposed between two parallel mirrors.
- FIG. 9a is a cross-sectional view of an optical pulse generator employing a nonlinear optical material disposed between two parallel mirrors that receives and sums a plurality of optical signal input pulses and generates a pulsed output signal having a pulse frequency that depends on the sum of the input signals.
- FIG. 9b is a cross-sectional view of another embodiment of an optical pulse generator employing two nonlinear optical materials disposed between parallel mirrors to form a nonlinear Fabry-Perot etalon, as shown in FIG. 9a. The device receives and sums a plurality of optical signal input pulses and generates a pulsed output signal having a pulse frequency that depends on the sum of the input signals.
- FIG. 10 is a cross-sectional view of a further embodiment of an optical pulse generator wherein a special light modulator comprising a sheet of a photosensitive material is disposed on one side of a nonlinear Fabry-Perot etalon such as the configuration shown in FIG. 9a.
- Turning now to FIG. 1, an optoelectronic
artificial neuron 10 is an astable circuit contained in twoconductive tubes conductive tubes nonconductive member 13. The conductive tubes are connected to a power supply (not shown) having a voltage output of +V. When a number ofsuch neurons 10 are interconnected by a common conductive medium, as, for example, by stacking, they can share the same single power supply lines. Input signals 14 from an external source are collected by aphotodetector 15. The output pulses comprising theoutput signal 16 of theneuron 10 are provided by a light-emitting diode (LED) 17, or a laser diode (LD). In theoutput signal 16 of theoptoelectronic neuron 10, the pulsation frequency is dependent on the light energy collected by thephotodetector 15. An example of the input and output pulse waveform is illustrated in FIG. 2. - In the
optoelectronic neuron 10, the input light signals 14 are summed over space or time. When the sum exceeds a threshold value, anoutput light pulse 16 will be generated. Everyindividual output pulse 16 comprising the output signal spikes to the same level as an all-or-nothing event. That is, once the positive-feedback cycle starts, nothing will stop the full amplitude spiking. If the threshold value is not reached, however, the slight increase in the output light intensity will subside when the stimulus is removed. Eachoutput pulse 16 lasts only for a short period of time. The pulse frequency is a function of the input signal summation, either temporally or spatially. FIG. 3 shows the relationship between the output pulse frequency and the summation of the input signals. - Neurons in biological neural networks receive information from both excitatory and inhibitory synapses. Accordingly, the
optoelectronic neuron 10 can be modified to 11 include both excitatory and inhibitory synapses as illustrated in the artificial optoelectronic neuron shown at 40 in FIG. 4. In theoptoelectronic neuron 40, the input port consists of two photodetectors, onephotodetector 15 being excitatory, and theother photodetector 41 being inhibitory.Individual neurons 40 decide whether or not to fire anoutput pulse 16 by summing excitatory and inhibitory inputs. Signals received by the excitatory synapses make the neuron more likely to fire a pulse. Conversely, signals received by the inhibitory synapses render theneuron 40 less likely to fire. - An example of a first preferred embodiment of an optoelectronic neuron based on an astable multivibrator is illustrated in FIG. 5. The astable circuit consists of two transistors T1 and T2, two capacitors C1 and C2, three resistors R1, R2 and R3, a photodetector PD, and a light-emitting diode LED. As with the receptor in biological neurons, the photodetector is the input port (i.e., receptor) for the
optoelectric neuron 10. The light-emitting diode serves as the output port, emitting signals to be received at the synaptic inputs of neighboring neurons. - With reference now to the
multivabrator circuit 60 of FIG. 6, if the photodetector PD of FIG. 5, which may be a photosensitive resistor, a photodiode or a photosensitive transistor, is replaced with a resistor R4, then thecircuit 60 will act as an ordinary astable multivibrator. Multivibrators are classified according to the number of steady (stable) states of the circuit. A steady state exists when circuit operation is essentially constant; that is, one transistor remains in conduction and the other remains “off” until an external signal is applied. The three types of multivibrators are the astable, monostable, and bistable. The astable circuit has no stable state. With no external signal applied, the transistors alternately switch from cutoff to saturation at a frequency determined by the RC time constants of the coupling circuits. In thecircuit 60 illustrated in FIG. 6, durations of the two unstable periods are - T1=0.69R4C1
- T 2=0.69R3C2
- The free-running frequency of such a multivibrator is given by
- f=1/T=1/(T 1 +T 2)=1.45/(R 4 C 1 +R 3 C 2).
- When a photodetector is included in the astable circuit to form an optoelectronic neuron, the oscillation frequency of the multivibrator circuit will be determined primarily by the light energy collected by the photodectector PD. With a continuous wave (CW) light input,
output signal pulses 16 are produced if the input power reaches an activation threshold. The frequency of the light pulses emitted from the LED increases with an increase of input light power. With pulsed light input, the frequency of the pulses comprising theoutput signal 16 increases with the summation of energy contained in each pulse. - FIG. 7 illustrates an
optoelectronic neuron 40 containing both excitatory and inhibitory synapses. Input signals 14 collected by photodetector PD1 can increase theoutput signal 16 pulsation frequency. On the other hand, inhibitory input signals 71, detected by PD2, can reduce the output signal pulsation frequency. - By using a plurality of the
optoelectronic neurons 10, shown in FIG. 1, a two-layer system can be assembled to simulate aneural network 80 as illustrated in FIG. 8. Forward orientedneurons 10 form the first layer, and some backward orientedneurons 81 in the assembly form the second layer. The input signal of aneuron 81 in the second layer is the light signal reflected from the partiallyreflective mirror 2, which light signal originated from its neighboringneurons 10 in the first layer. In this 2-layer system 80, synchronization and momental contrast enhancement can be performed. When uniformly distributed light intensity is projected to the input end of the first layer, all the neurons are synchronized and pulse periodically. When neurons in different areas are simultaneously exposed to light intensity at different levels, the neurons will pulse in groups. - 3. All-Optical Neuron
- The objectives of the present invention can also be met by all-optical artificial neurons. All-optical neurons can be constructed by advantageously using the regenerative pulsations in optical bistable devices. Regenerative pulsation in optical bistable devices is a suitable mechanism for realization of optical neurons due to the following properties:
- 1. The pulsation has a distinctive threshold.
- 2. By using proper nonlinear materials, the pulsation frequency can be increased by increasing of the power of the excitation light.
- 3. Occurrence of regenerative pulsation in optical bistable devices have favorable switching and memory properties.
- Regenerative pulsation in optical bistable devices is an optical phenomenon, first observed by McCall, using a hybrid optical device (S. L. McCall, “Instability and regenerative pulsation phenomena in Fabry-Perot nonlinear optic media devices,” Appl. Phys. Lett., Vol. 32, 284(1978)). Regenerative pulsation can also take place in an all-optical bistable device, when the nonlinearity had two contributions of opposite sign and different time constants. When regenerative pulsation takes place, the bistable device turns into an astable device. As of the present date, this phenomenon has only been observed in several bistable devices using nonlinear materials such as GaAs, InSb, ZnS and liquid crystals. In these prior art devices, the pulsation frequency is relatively stable, but not dependent on the incident light intensity as desired for constructing optical neurons. Accordingly, these materials are not suitable candidates for all-optical neuron applications at the present time. For optical neuron applications, the nonlinear material needs to have a response time constant that varies with the light intensity of light incident thereon. In certain materials, such as dye-doped liquid crystals and photorefractive crystals, the nonlinear response time constants are sensitive to the light intensity and these materials are, therefore suitable for building optical neurons.
- The pulse coupled, all-optical neuron in accordance with the present invention consists two elements: an input signal integrator and a bistable device, the latter being a pulse generator. The function of the integrator is to sum the input light pulses, and transmit a continuous wave light beam, the intensity of the CW beam fluctuating with the summation of the input signals. The greater the number of input signals the integrator receives, the higher the intensity of the CW beam it transmits to the pulse generator. An example of an integrator suitable for use in an optical neuron application is a spatial light modulator comprising liquid crystals. The readout light intensity of such a spatial light modulator is a function of the incident light signal energies, with a response speed determined by the frequency of the LC driver.
- The pulse generator is a key element in the all-optical neuron of the present invention. The function of the pulse generator is to generate a train of pulses having a frequency that varies with the light intensity it receives from the integrator. In the preferred embodiment of the present invention, the pulse generator is actually a nonlinear Fabry-Perot etalon. The optical behavior of the nonlinear medium contained in the etalon cavity generally depends on two nonlinear mechanisms. One mechanism tends to turn the device “on” at higher input light intensities, while the other (slower) mechanism tends to turn the device “off”.
- An example of a nonlinear medium that can be adapted for disposition within the cavity of a Fabry-Perot etalon is a dye-doped liquid crystal, which has both a large thermo-optic coefficient and large orientational nonlinearity. When the light intensity incident upon the liquid crystal is higher than the activation threshold of the dye-doped liquid crystal, thermal effects result in a change in the index of refraction of the medium, which can turn the device “on” automatically. When the device is turned “on”, the light intensity build-up inside the etalon cavity triggers the orientational nonlinearity, which will turn the device “off”. After the device is turned “off”, the thermal effect will again dominate and the device turns “on”. Under a constant incident power, this on-and-off process is continuous, and the etalon mirror transmits a train of pulse with a fixed frequency. When the incident power changes, the response time constants in the dye-doped liquid crystal changes, which induces a change in the pulsation frequency.
- When designing a pulse generator for use in an optical neuron for practical applications, the contrast ratio of the device is preferably maximized. The Fabry-Perot is basically a simple interferometer relying on the interference of multiply reflected beams. The simplest Fabry-Perot etalon consists of a pair of parallel mirrors. The incident beam of light undergoes multiple reflections between two parallel mirrors that define the cavity. In a Fabry-Perot etalon containing a nonlinear material within the cavity, the transmission of light by the etalon is governed by:
- where θ is the single-trip phase shift of the cavity, R is the reflectivity of the mirrors, T0 is the peak transmission of the filter, α is the absorption coefficient of the material, β is the scattering coefficient, L is the etalon cavity length, λ is the light wavelength, and θ′o is a constant. The transmission peaks at θ=0, π, 2π, . . . nπ, can be made very sharp in a low loss cavity (α≈β≈0) by increasing the reflectivities of the mirror surfaces.
- In the Fabry-Perot, when a dye-doped liquid crystal is disposed within the cavity, the cavity loss is due to two main factors: (a) absorption by the dye molecules; and (b) scattering by the liquid crystal molecules. From the equations shown above, the contrast ratio of the device in the “on” and “off” states is
- Δ=1+F
- In order to obtain a high F factor, a medium with low absorption and scattering is preferred. Since the photosensitivity of the material is dependent on the concentration of the doped dye, which can not be too low, the only way to enhance the F factor is to lower the scattering loss of the liquid crystal. When the liquid crystal molecules are well aligned in the cavity, high contrast ratio can be achieved. The use of vacuum deposited dielectric films as an alignment layer is a common practice. In experiments, it has been found that some dielectric mirrors may also be used as the alignment layers. In the film deposition process, if the substrate is tilted an angle to the evaporation source, the molecular orientation within the film will have a determined direction, which may be used to align the liquid crystal molecules.
- Just as the nonlinear Fabry-
Perot etalon 91 containing a medium 90 a having two nonlinear mechanisms, shown in theoptical neuron 90 of FIG. 9a, regenerative pulsation can also take place in a nonlinear Fabry-Perot etalon when the etalon has two layers of nonlinear materials 90 b disposed in the cavity, as illustrated in FIG. 9b. The structure of theoptical pulse generator 90′ is more advantageous than theoptical pulse generator 90 because of the greater flexibility in the selection of the nonlinear materials 90 b disposed in the cavity of theetalon 91. If only one material with two nonlinear mechanisms is used, such as indicated at 90 a in FIG. 9a, the response properties of the mechanisms are usually fixed. -
-
- θo and φ0 are the phase shifts under steady state,
- θ0=αPiT(θ0,φ0) (7)
- φ0=rθ0
- When Equation (4) is satisfied, the nonlinear Fabry-Perot etalon turns from a bistable device into an astable device. By employing proper nonlinear materials that have time constants related to the irradiance, the pulsation frequency can be a function of the incident intensity, satisfying the requirements for optical neuron operations.
- In order to make the nonlinear material more sensitive to incident light, a spatial light modulator can be combined with a nonlinear Fabry-Perot etalon. FIG. 10 shows such an all-
optical neuron 100 having a laminate structure. Theoptical neuron 100 consists of aphotoconductive layer 101 such as, for example, a layer of amorphous silicon, a pair ofdielectric mirrors 102, and a liquid crystal layer 103. Themirrors 102 and the liquid crystal layer 103 form a nonlinear Fabry-Perot etalon 104. In the state where noinput signal 105 is incident on thedevice 100, and no reading light 106 reaches thephotoconductive layer 101 by passing through the Fabry-Perot in a right to left direction, the impedance of thephotoconductive layer 101 is extremely high. When an AC voltage is applied onto thetransparent electrode layers 107, the electric field in the liquid crystal layer, which field is dependent on the intensity of light reaching either side of thephotoconductive layer 101, induces a phase change in the liquid crystal. The nonlinear Fabry-Perot etalon 104 can be switched from the reflection state to the transmission state as a result of the phase change in the liquid crystal layer 103. In the transmission state, the light intensity inside cavity may trigger nonlinearity due to the thermal effect and/or orientational effects inside the Fabry-Perot etalon, and switch the Fabry-Perot etalon from transmission state back to reflection state. This process continues ad infinitum, with a frequency depending on the intensity of the incident light signal. The shape ofoutput pulses 108 produced byoptical neuron 100 is opposite to the shape of theinput pulses 105. Accordingly, a converter such as a special light modulator is needed to transform the waveform. - In addition to potential applications in pulse coupled neural networks, both the
optoelectronic neurons optical neurons array 80 of such neurons as indicated in FIG. 8. Interconnection among these neurons can be performed with off-the-shelf optical components such as mirrors, lenses, gratings and optical fibers. - While particular embodiments of the present invention have been illustrated and described, it would be obvious to those skilled in the art that various other changes and modifications can be made without departing from the spirit and scope of the invention. It is therefore intended to cover in the appended claims all such changes and modifications that are within the scope of this invention.
Claims (13)
1. An artificial neuron operable for receiving an optical input signal having an intensity and providing an optical output signal comprising a sequence of optical pulses having a pulse output frequency that depends on the intensity of the input signal, the artificial neuron comprising: (a) an input signal integrator operable for summing a plurality of optical input signals to provide an integrated electrical or optical input signal having an integrated signal intensity; and (b) an optical pulse generator operable for receiving said integrated electrical or optical input signal from said input signal integrator and provide optical output pulses having a frequency that is determined by said integrated signal intensity.
2. The artificial neuron in accordance with claim 1 wherein said input signal integrator is a photosensitive element.
3. The artificial neuron in accordance with claim 2 wherein said photosensitive element is a resistor.
4. The artificial neuron in accordance with claim 2 wherein said photosensitive element is a photosensitive diode.
5. The artificial neuron in accordance with claim 2 wherein said photosensitive element is a photosensitive transistor.
6. An artificial neuron in accordance with claim 2 wherein said optical pulse generator is an astable multivibrator comprising a photoemissive device having an optical pulse output frequency that depends upon said integrated signal intensity.
7. An artificial neuron in accordance with claim 1 wherein said optical pulse generator comprises a nonlinear Fabry-Perot etalon.
8. An artificial neuron in accordance with claim 7 wherein said nonlinear Fabry-Perot etalon comprises a dye-doped liquid crystal layer disposed between two parallel, partially reflecting mirrors.
9. An artificial neuron in accordance with claim 8 further comprising a second layer of nonlinear optical material disposed between said two parallel, partially reflecting mirrors.
10. An artificial neuron in accordance with claim 8 wherein said Fabry-Perot etalon is disposed between a pair of optically transparent electrodes.
11. An artificial neuron in accordance with claim 9 wherein said Fabry-Perot etalon is disposed between a pair of optically transparent electrodes.
12. An optical neural network comprising a plurality of artificial neurons, each artificial neuron being operable for receiving an optical input signal having an intensity and providing an optical output signal comprising a sequence of optical pulses having a pulse output frequency that depends on said intensity, the artificial neuron further comprising: (a) an input signal integrator operable for receiving and summing a plurality of optical input signals to provide an integrated electrical or optical input signal having an integrated signal intensity; and (b) an optical pulse generator operable for receiving said integrated electrical or optical input signal from said input signal integrator and provide optical output pulses having a frequency that is determined by said integrated signal intensity.
13. An optical neural network comprising a plurality of artificial neurons in accordance with claim 12 , further comprising reflective means operable for reflecting at least a portion of said optical output pulses to provide reflected optical output pulses and directing at least a portion of said reflected optical output pulses to be received by said input signal generator of at least one optical neuron comprising said optical neuron network.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US09/963,971 US6754646B1 (en) | 2001-09-25 | 2001-09-25 | Optical pulse-coupled artificial neurons |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US09/963,971 US6754646B1 (en) | 2001-09-25 | 2001-09-25 | Optical pulse-coupled artificial neurons |
Publications (2)
Publication Number | Publication Date |
---|---|
US20040107172A1 true US20040107172A1 (en) | 2004-06-03 |
US6754646B1 US6754646B1 (en) | 2004-06-22 |
Family
ID=32393906
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US09/963,971 Expired - Fee Related US6754646B1 (en) | 2001-09-25 | 2001-09-25 | Optical pulse-coupled artificial neurons |
Country Status (1)
Country | Link |
---|---|
US (1) | US6754646B1 (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2007071070A1 (en) * | 2005-12-23 | 2007-06-28 | Universite De Sherbrooke | Spatio-temporal pattern recognition using a spiking neural network and processing thereof on a portable and/or distributed computer |
WO2015028504A1 (en) * | 2013-09-02 | 2015-03-05 | Koninklijke Philips N.V. | Artificial neuron |
US10142034B2 (en) | 2013-09-02 | 2018-11-27 | Philips Lighting Holding B.V. | Optically transmissive electronic device having an optically transmissive light emitting device to transmit optical signal to a second optically transmissive light receiving device through a first optically transmissive light receiving device |
US20190130909A1 (en) * | 2016-04-18 | 2019-05-02 | The Industry & Academic Cooperation In Chungnam National University (Iac) | Voice frequency analysis system, voice frequency analysis method, and voice recognition system and voice recognition method using the same |
CN109784486A (en) * | 2018-12-26 | 2019-05-21 | 中国科学院计算技术研究所 | A kind of optical neural network processor and its training method |
US10429580B2 (en) | 2015-08-27 | 2019-10-01 | Bar-Ilan University | Multi optically-coupled channels module and related methods of computation |
WO2019217835A1 (en) * | 2018-05-10 | 2019-11-14 | The Board Of Trustees Of The Leland Stanford Junior University | Training of photonic neural networks through in situ backpropagation |
WO2020147282A1 (en) * | 2019-01-16 | 2020-07-23 | 南方科技大学 | Image recognition method based on optical neural network structure, apparatus and electronic device |
US11460753B2 (en) | 2018-05-10 | 2022-10-04 | The Board Of Trustees Of The Leland Stanford Junior University | Systems and methods for activation functions for photonic neural networks |
US11797836B1 (en) * | 2019-12-23 | 2023-10-24 | Waymo Llc | Sensor-integrated neural network |
US11796794B2 (en) | 2020-05-12 | 2023-10-24 | The Board Of Trustees Of The Leland Stanford Junior University | Multi-objective, robust constraints enforced global topology optimizer for optical devices |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
RU2604331C2 (en) * | 2014-11-05 | 2016-12-10 | Айыысхан Иванович Алексеев | Artificial neuron (versions) |
KR20200097369A (en) * | 2019-02-07 | 2020-08-19 | 삼성전자주식회사 | Optical device and optical neural network apparatus including the same |
CN111244270B (en) * | 2020-01-16 | 2021-08-24 | 北京大学 | Electronic device and method for simulating function of biological neuron |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5099114A (en) * | 1989-04-28 | 1992-03-24 | Nippon Telegraph & Telephone Corporation | Optical wavelength demultiplexer |
US5684586A (en) * | 1996-06-13 | 1997-11-04 | Hewlett Packard Company | Apparatus for characterizing short optical pulses |
US5960111A (en) * | 1997-02-10 | 1999-09-28 | At&T Corp | Method and apparatus for segmenting images prior to coding |
US6650915B2 (en) * | 2001-09-13 | 2003-11-18 | Fovioptics, Inc. | Non-invasive measurement of blood analytes using photodynamics |
-
2001
- 2001-09-25 US US09/963,971 patent/US6754646B1/en not_active Expired - Fee Related
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2007071070A1 (en) * | 2005-12-23 | 2007-06-28 | Universite De Sherbrooke | Spatio-temporal pattern recognition using a spiking neural network and processing thereof on a portable and/or distributed computer |
US20090287624A1 (en) * | 2005-12-23 | 2009-11-19 | Societe De Commercialisation De Produits De La Recherche Applique-Socpra-Sciences Et Genie S.E.C. | Spatio-temporal pattern recognition using a spiking neural network and processing thereof on a portable and/or distributed computer |
US8346692B2 (en) | 2005-12-23 | 2013-01-01 | Societe De Commercialisation Des Produits De La Recherche Appliquee-Socpra-Sciences Et Genie S.E.C. | Spatio-temporal pattern recognition using a spiking neural network and processing thereof on a portable and/or distributed computer |
WO2015028504A1 (en) * | 2013-09-02 | 2015-03-05 | Koninklijke Philips N.V. | Artificial neuron |
CN105531723A (en) * | 2013-09-02 | 2016-04-27 | 飞利浦照明控股有限公司 | Artificial neuron |
US10142034B2 (en) | 2013-09-02 | 2018-11-27 | Philips Lighting Holding B.V. | Optically transmissive electronic device having an optically transmissive light emitting device to transmit optical signal to a second optically transmissive light receiving device through a first optically transmissive light receiving device |
US10429580B2 (en) | 2015-08-27 | 2019-10-01 | Bar-Ilan University | Multi optically-coupled channels module and related methods of computation |
US20200003948A1 (en) * | 2015-08-27 | 2020-01-02 | Bar-Ilan University | Multi optically-coupled channels module and related methods of computation |
US10838139B2 (en) * | 2015-08-27 | 2020-11-17 | Bar-Ilan University | Multi optically-coupled channels module and related methods of computation |
US20190130909A1 (en) * | 2016-04-18 | 2019-05-02 | The Industry & Academic Cooperation In Chungnam National University (Iac) | Voice frequency analysis system, voice frequency analysis method, and voice recognition system and voice recognition method using the same |
US10930279B2 (en) * | 2016-04-18 | 2021-02-23 | The Industry & Academic Cooperation In Chungnam National University (Iac) | Voice frequency analysis system, voice frequency analysis method, and voice recognition system and voice recognition method using the same |
WO2019217835A1 (en) * | 2018-05-10 | 2019-11-14 | The Board Of Trustees Of The Leland Stanford Junior University | Training of photonic neural networks through in situ backpropagation |
US11460753B2 (en) | 2018-05-10 | 2022-10-04 | The Board Of Trustees Of The Leland Stanford Junior University | Systems and methods for activation functions for photonic neural networks |
US12026615B2 (en) | 2018-05-10 | 2024-07-02 | The Board Of Trustees Of The Leland Stanford Junior University | Training of photonic neural networks through in situ backpropagation |
CN109784486A (en) * | 2018-12-26 | 2019-05-21 | 中国科学院计算技术研究所 | A kind of optical neural network processor and its training method |
WO2020147282A1 (en) * | 2019-01-16 | 2020-07-23 | 南方科技大学 | Image recognition method based on optical neural network structure, apparatus and electronic device |
US11797836B1 (en) * | 2019-12-23 | 2023-10-24 | Waymo Llc | Sensor-integrated neural network |
US11796794B2 (en) | 2020-05-12 | 2023-10-24 | The Board Of Trustees Of The Leland Stanford Junior University | Multi-objective, robust constraints enforced global topology optimizer for optical devices |
Also Published As
Publication number | Publication date |
---|---|
US6754646B1 (en) | 2004-06-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US6754646B1 (en) | Optical pulse-coupled artificial neurons | |
US5157537A (en) | Distributed resonant cavity light beam modulator | |
US5337183A (en) | Distributed resonant cavity light beam modulator | |
Lian et al. | Photonic (computational) memories: tunable nanophotonics for data storage and computing | |
EP0207157A1 (en) | Optical logic arrangement with self electro-optic effect devices | |
JP2656694B2 (en) | Light source, photon switching device and optical switching device | |
EP0083520A1 (en) | Optical beam switching device and telephone exchange including the same | |
JP2572052B2 (en) | Method and apparatus for manufacturing a reflection hologram for interconnecting nonlinear optical elements | |
Kurokawa et al. | Spatial light modulators using ferroelectric liquid crystal | |
US6967765B2 (en) | Active optical system for changing the wavelength of an image | |
Arsenault et al. | An introduction to optics in computers | |
US4403833A (en) | Electrooptical multipliers | |
Eldering et al. | Electrooptic polymer materials and devices for global optical interconnects | |
EP0434139A2 (en) | Frequency doubling optical waveguide with active phase matching | |
US4767196A (en) | Asymmetric optical logic element | |
McCormick et al. | All-Optical Shift Register Using Symmetric Self-Electro-Optic Effect Devices | |
FR2575284A1 (en) | OPTICAL POSITION MARKING DEVICE | |
US4126901A (en) | Photovoltaic-ferroelectric correlation devices | |
Shamir | Fundamental limits of optical computing | |
Semenov et al. | Optical waveguide processors | |
EP0189706B1 (en) | Optical device for position detection | |
JPS59112233A (en) | Measuring device for characteristics of optical beam | |
Radehaus et al. | Maximum detection with a two-dimensional optoelectronic pnpn winner-take-all network | |
Itoh et al. | 1.06-um optically addressable liquid crystal spatial light modulator for an HV target of the indoor laser radar system | |
Eldering et al. | Fixed optical interconnects for concurrent computing systems |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
REMI | Maintenance fee reminder mailed | ||
LAPS | Lapse for failure to pay maintenance fees | ||
STCH | Information on status: patent discontinuation |
Free format text: PATENT EXPIRED DUE TO NONPAYMENT OF MAINTENANCE FEES UNDER 37 CFR 1.362 |
|
FP | Expired due to failure to pay maintenance fee |
Effective date: 20080622 |