All optical neural network
An all-optical neural network that utilizes light beams and optical components to implement layers of the neural network is disclosed herein. The all-optical neural network includes an input layer, zero or more hidden layers, and an output layer. Each layer of the neural network is configured to simulate linear and nonlinear operations of a conventional artificial neural network neuron on an optical signal. In an embodiment, the optical linear operation is performed by a spatial light modulator and an optical lens. The optical lens performs a Fourier transformation on the set of light beams and sums light beams with similar propagation orientations. The optical nonlinear operation is implemented utilizing a nonlinear optical medium having an electromagnetically induced transparency characteristic whose transmission of a probe beam of light is controlled by the intermediate output of a coupling beam of light from the optical linear operation.
1. A system for implementing an optical neuron, the system comprising:
a linear subsystem configured to perform an optical summation operation that combines one or more beams of light to generate a coupling beam of light as an intermediate signal, wherein the linear subsystem comprises:
a first spatial light modulator (SLM) for modulating the one or more beams of light; and
an optical lens that combines the one or more beams of light to generate the coupling beam of light; and
a nonlinear subsystem configured to perform an optical nonlinear operation based on the coupling beam of light to generate an activation response signal, wherein the nonlinear subsystem includes a nonlinear optical medium that has an electromagnetically induced transparency (EIT) characteristic, and wherein the activation response signal comprises a probe beam of light transmitted through the nonlinear optical medium;
wherein the nonlinear subsystem comprises a probe laser configured to generate the probe beam of light, wherein the probe beam of light is directed at the nonlinear optical medium, and the transmission of the probe beam of light through the nonlinear optical medium is controlled based on the coupling beam of light.
2. The system of claim 1 , wherein the optical lens is configured to perform a Fourier transform.
3. The system of claim 2 , wherein the optical lens generates at least two intermediate signals for two or more optical neurons by combining beams of light, each optical neuron corresponding to a particular propagating orientation, and each intermediate signal located at a different location on a focal plane of the optical lens.
4. The system of claim 1 , further comprising:
a second spatial light modulator (SLM) configured to modulate one or more additional beams of light by a set of weights to generate one or more weighted beams of light as the output of the optical neuron, wherein the one or more additional beams of light are split from the activation response signal.
5. The system of claim 4 , wherein the second SLM is tuned using a weighted Gerchberg-Saxton (GSW) algorithm; and
wherein the system further comprises a photosensor configured to measure the output from the second SLM.
6. The system of claim 4 , wherein the set of weights is learned by training a neural network based on a set of training data, and wherein the neural network includes an input layer, one or more hidden layers, and an output layer, and the set of weights is associated with an interface between the input layer and a hidden layer of the one or more hidden layers, a hidden layer of the one or more hidden layers and a subsequent hidden layer of the one or more hidden layers, or a hidden layer of the one or more hidden layers and the output layer.
7. The system of claim 1 , wherein the transmission of the probe beam of light in the nonlinear optical medium is controlled by at least an intensity or a frequency of the coupling beam of light.
8. The system of claim 1 , wherein the nonlinear optical medium comprises at least one of atoms, molecules, quantum dots, or solid-state materials in which a population of particles are controllable in one or more quantum or classical states.
9. The system of claim 1 , wherein the optical lens is configured to generate two or more intermediate signals for two or more optical neurons by combining beams of light, wherein each intermediate signals of the two or more intermediate signals corresponds to a spot at a front focal plane, wherein each spot corresponds to a different optical neuron of the two or more optical neurons.
10. A system for implementing an all-optical neural network (AONN), the system comprising:
an input layer including one or more optical neurons; and
an output layer including one or more optical neurons,
wherein each optical neuron of the one or more optical neurons in the input layer and of the one or more optical neurons in the output layer comprises a linear subsystem and a nonlinear subsystem;
wherein the linear subsystem comprises:
a first spatial light modulator (SLM) for modulating one or more beams of light; and
an optical lens configured to combine the one or more beams of light to generate a coupling beam of light as an intermediate signal; and
wherein the nonlinear subsystem comprises:
a nonlinear optical medium that has an electromagnetically induced transparency (EIT) characteristic controlled by the coupling beam of light; and
a probe laser configured to generate a probe beam of light, wherein the probe beam of light is directed at the nonlinear optical medium, and wherein the transmission of the probe beam of light through the nonlinear optical medium is controlled based on the coupling beam of light.
11. The system of claim 10 , wherein each optical neuron in the input layer is configured to receive one beam of light as an input to the optical neuron, and wherein each optical neuron in the output layer transmits one beam of light as an output of the optical neuron.
12. The system of claim 10 , further comprising:
one or more hidden layers, wherein each hidden layer includes one or more optical neurons,
wherein each optical neuron in the input layer, a hidden layer of the one or more hidden layers, or the output layer further comprises:
a second spatial light modulator (SLM) for modulating an activation response signal output by the nonlinear subsystem for the optical neuron by a set of weights to generate weighted output signals as inputs for a subsequent layer of the AONN, wherein the activation response signal comprises the probe beam of light transmitted through the nonlinear optical medium that is nonlinearly controlled by the coupling beam of light.
13. The system of claim 12 , wherein the system further comprises:
a photosensor configured to capture the weighted output signals generated by the second SLM; and
a light source configured to generate one or more additional beams of light in accordance with the weighted output signals to implement a subsequent layer of the AONN, wherein the subsequent layer is a hidden layer of the one or more hidden layers or the output layer.
14. The system of claim 13 , wherein the first SLM is configured to spatially modulate at least one of an amplitude or a phase of a light beam incident on a surface of the first SLM.
15. The system of claim 14 , wherein the first SLM is tuned using a weighted Gerchberg-Saxton (GSW) algorithm.
16. The system of claim 10 , wherein the optical lens is configured to combine the one or more beams of light on a focal plane of the optical lens.
17. The system of claim 10 , wherein the optical lens is configured to generate two or more intermediate signals for two or more optical neurons by combining beams of light, wherein each intermediate signals of the two or more intermediate signals corresponds to a spot at a front focal plane, wherein each spot corresponds to a different optical neuron of the two or more optical neurons.