Nonlinear activation layer and nonlinear activation method for optical neural network

Abstract

The embodiment of the invention provides a nonlinear activation layer and a nonlinear activation method for an optical neural network, and relates to the technical field of optical signal processing. The non-linear activation layer comprises a binary non-linear activation device which comprises a two-dimensional material liquid crystal which deflects under the action of an electric field; the polarizer is arranged on the first side of the binary nonlinear activation device; the polarization analyzer is arranged on the second side of the binary nonlinear activation device, the second side is opposite to the first side, and the polarization direction of the polarization analyzer is perpendicular to the polarization direction of the polarizer; wherein the linearly polarized light enters the polarizer, the polarizer filters the linearly polarized light, the filtered linearly polarized light enters the binary nonlinear activation device, the deflected two-dimensional material liquid crystal modulates the filtered linearly polarized light, and the modulated linearly polarized light enters the polarization analyzer to realize optical nonlinear activation. The nonlinear activation layer of the application has a dynamic adjustable characteristic and is suitable for an all-optical architecture.

Description

Technical Field

[0001] The present application relates to the technical field of optical signal processing, and in particular to a nonlinear activation layer and a nonlinear activation method for an optical neural network. Background Art

[0002] Optical neural network is an architectural paradigm that uses optical devices such as optical modulators and photodetectors to achieve computational reasoning functions by transmitting and processing parameters such as the amplitude, phase and frequency of light. Compared with traditional electrical neural networks, optical neural networks have the advantages of low energy consumption, low latency, parallelism and anti-interference because the signal transmission carrier is photons, and have received great attention from academia and industry. The nonlinear activation layer is one of the core components of optical neural networks and is used for feature learning and representation. However, common optical devices used in free-space optical neural networks, such as spatial light modulators and digital micromirrors, do not have nonlinear optical effects and cannot be used as nonlinear activation layers. At present, there are two common technical paths applied to free-space optical neural networks to achieve nonlinear activation. First, optical signals are converted into electrical signals through photodiodes or electronic circuits to indirectly achieve nonlinear activation. However, this method cannot achieve full-link optical signal processing and calculation, and loses the rich information carried by optical signals while increasing response time. Second, nonlinear activation is achieved by using the absorption and transmission characteristics of optical nonlinear materials. However, current nonlinear optical activation is mainly based on light intensity coding, which cannot achieve negative activation / only has an odd function activation form that is symmetric about the origin, making it difficult for the network to capture the true distribution of data and may also lose important features. In addition, optical nonlinear materials only have fixed response characteristics, and their activation methods cannot be dynamically controlled, making them unable to adapt to diverse and more complex learning tasks and network architectures.

[0003] In summary, how to develop a nonlinear activation layer with dynamically controllable properties suitable for all-optical architecture has become a technical problem that needs to be solved urgently. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to propose a nonlinear activation layer and a nonlinear activation method for an optical neural network, wherein the nonlinear activation layer has dynamically adjustable characteristics and is suitable for an all-optical architecture.

[0005] To achieve the above-mentioned purpose, a first aspect of an embodiment of the present application proposes a nonlinear activation layer for an optical neural network, wherein the nonlinear activation layer comprises:

[0006] Binary nonlinearly activated devices, comprising liquid crystals of two-dimensional materials that deflect under the action of an electric field;

[0007] A polarizer, the polarizer being arranged on a first side of the binary nonlinear activation device;

[0008] An analyzer, wherein the analyzer is arranged on a second side of the binary nonlinear activation device, the second side being a side opposite to the first side, and a polarization direction of the analyzer is perpendicular to a polarization direction of the polarizer;

[0009] Among them, the linear polarized light is incident on the polarizer, and the polarizer filters the linear polarized light. The filtered linear polarized light is incident on the binary nonlinear activation device, and the filtered linear polarized light is modulated by the deflected two-dimensional material liquid crystal. The modulated linear polarized light is incident on the analyzer to achieve optical nonlinear activation.

[0010] Optionally, the binary nonlinear activation device includes a first nonlinear activation device and a second nonlinear activation device with perpendicular electric field directions, the polarizer includes a first polarizer and a second polarizer with perpendicular polarization directions, the analyzer includes a first polarizer and a second polarizer with perpendicular polarization directions, the polarization direction of the first polarizer is the same as the polarization direction of the second polarizer, and the polarization direction of the second polarizer is the same as the polarization direction of the first analyzer;

[0011] The nonlinear activation layer further comprises:

[0012] A beam splitter prism, the beam splitter prism being arranged on one side of the first nonlinear activation device and on one side of the second nonlinear activation device;

[0013] Among them, the linear polarized light is incident on the beam splitter prism, and the beam splitter prism splits the linear polarized light into a first linear polarized light and a second linear polarized light, and the polarization direction of the first linear polarized light is perpendicular to the polarization direction of the second linear polarized light; the first linear polarized light is incident on the first polarizer, and the second linear polarized light is incident on the second polarizer.

[0014] Optionally, the first nonlinear activation device is used for positive modulation, the second nonlinear activation device is used for negative modulation, the angle between the polarization direction of the first polarizer and the electric field direction of the first nonlinear activation device is 45°, and the angle between the polarization direction of the second polarizer and the electric field direction of the second nonlinear activation device is 45°;

[0015] If the polarization azimuth angle of the linearly polarized light is 45°, then after the first linearly polarized light is incident on the first polarizer, it passes through the first nonlinear activation device and the first analyzer in sequence;

[0016] If the polarization azimuth angle of the linearly polarized light is -45°, then after the second linearly polarized light is incident on the second polarizer, it passes through the second nonlinear activation device and the second analyzer in sequence.

[0017] Optionally, the nonlinear activation layer further comprises: a quarter wave plate, wherein the quarter wave plate is arranged on one side of the beam splitter prism.

[0018] Wherein, the circularly polarized light is incident on the quarter-wave plate, and the quarter-wave plate is used to convert the circularly polarized light into linearly polarized light.

[0019] Optionally, the first nonlinear activation device is used for positive modulation, the second nonlinear activation device is used for negative modulation, the angle between the polarization direction of the first polarizer and the electric field direction of the first nonlinear activation device is 45°, and the angle between the polarization direction of the second polarizer and the electric field direction of the second nonlinear activation device is 45°;

[0020] If the phase difference of the circularly polarized light is π / 2, then after the first linearly polarized light is incident on the first polarizer, it passes through the first nonlinear activation device and the first analyzer in sequence;

[0021] If the phase difference of the linearly polarized light is -π / 2, then after the second linearly polarized light is incident on the second polarizer, it passes through the second nonlinear activation device and the second analyzer in sequence.

[0022] Optionally, the first nonlinear activation device and the second nonlinear activation device include: Kerr cells of two-dimensional material liquid crystal.

[0023] Optionally, the Kerr cell of the two-dimensional material liquid crystal comprises:

[0024] A Kerr cell container having a built-in functional layer, wherein the functional layer contains a two-dimensional material liquid crystal;

[0025] Two electrodes, the two electrodes are respectively arranged on opposite sides of the Kerr cell container;

[0026] The two electrodes are connected to an external power source to apply an electric field to the functional layer, and the orientation of the two-dimensional material liquid crystal is deflected under the action of the electric field.

[0027] Optionally, the two-dimensional liquid crystal material is a two-dimensional vermiculite aqueous dispersion, which can achieve π / 2 phase modulation under the action of an electric field.

[0028] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application proposes a nonlinear activation method for an optical neural network, which is applied to the nonlinear activation layer as described in the first aspect, and the method comprises:

[0029] Controlling linear polarized light to be incident on the polarizer, the polarizer filters the linear polarized light, and the filtered linear polarized light is incident on the nonlinear activation device;

[0030] An electric field is applied to the nonlinear activation device to modulate the filtered linearly polarized light through the two-dimensional material liquid crystal that is deflected under the action of the electric field, and the modulated linearly polarized light is incident on the analyzer to achieve optical nonlinear activation.

[0031] Optionally, before the linearly polarized light is controlled to be incident on the polarizer, the method further comprises:

[0032] The circularly polarized light is controlled to be incident on a quarter wave plate, and the quarter wave plate converts the circularly polarized light into linearly polarized light.

[0033] The present application proposes a nonlinear activation layer and a nonlinear activation method for an optical neural network. The nonlinear activation process includes: linear polarized light is incident on a polarizer, the polarizer filters the linear polarized light, the filtered linear polarized light is incident on a binary nonlinear activation device, the filtered linear polarized light is modulated by a deflected two-dimensional material liquid crystal, and the modulated linear polarized light is incident on an analyzer to achieve optical nonlinear activation. The process utilizes the optical path formed by the polarizer, the binary nonlinear activation device and the analyzer, and utilizes the binary nonlinear activator to adjust different incident light transmittances by changing the electric field intensity, and can follow Malus's law to achieve reconfigurable optical activation. Moreover, the two-dimensional material liquid crystal has a large electro-optical Kerr coefficient, and the phase delay is positively correlated with the electro-optical Kerr coefficient, so the binary nonlinear activation device can achieve full-precision full modulation of the phase within a smaller electric field range. In addition, since the two-dimensional material will not be photodegraded and has good stability to ultraviolet radiation, it can be used for ultraviolet light calculations. In summary, this nonlinear activation layer has dynamically controllable characteristics and is suitable for all-optical architecture. It can improve the performance of optical neural networks in complex tasks and promote the development of the field of optical neural networks. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 is a schematic diagram of the structure of a nonlinear activation layer provided in an embodiment of the present application;

[0035] Figure 2 A photo of a two-dimensional vermiculite liquid crystal provided in an embodiment of the present application;

[0036] Figure 3 A binary nonlinear activation device with a diagonal dimension of 46.1 mm provided in an embodiment of the present application;

[0037] Figure 4 The phase full modulation effect diagram of the two-dimensional vermiculite aqueous dispersion with a concentration of 9 g / L provided in the embodiment of the present application before and after 840 hours under different electric field intensities;

[0038] Figure 5The modulation effect diagram of the two-dimensional vermiculite aqueous dispersion with a concentration of 9 g / L provided in the embodiment of the present application before and after strong ultraviolet light irradiation for 12 hours;

[0039] Figure 6 A phase modulation comparison diagram of a two-dimensional vermiculite liquid crystal and a two-dimensional graphene oxide liquid crystal at a concentration of 9 g / L provided in an embodiment of the present application in the same electric field intensity range;

[0040] Figure 7 A binary nonlinear activation device with a diagonal dimension of 127.3 mm provided in an embodiment of the present application;

[0041] Figure 8 A schematic diagram of the structure of a nonlinear activation layer for linearly polarized light provided in an embodiment of the present application;

[0042] Fig. 9 A schematic diagram of the structure of a nonlinear activation layer for circularly polarized light provided in an embodiment of the present application;

[0043] Fig.10 It is a comparison diagram of the binary activation effect and the linear rectification function activation effect of the two-dimensional vermiculite liquid crystal with a concentration of 9 g / L of the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0045] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0046] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0047] First, some nouns involved in this application are analyzed:

[0048] Optical neural network: It is a new type of network formed by using optical technology, such as optical connection technology and optical device technology. It has the unique advantages of ultra-parallel processing and information transmission capabilities, high-density wiring capabilities and the ability to directly process images. Its basic components are optical neuron devices and optical synapse devices, among which optical neural chips are crucial. Similar to electrical neural networks, optical neural networks also require nonlinear activation processing, so it is necessary to set up a nonlinear activation layer for optical neural networks.

[0049] Nonlinear activation layers for optical neural networks: mainly introduce nonlinear characteristics to enable processing of more complex tasks and data patterns.

[0050] The embodiments of the present application provide a nonlinear activation layer and a nonlinear activation method for an optical neural network, which are specifically illustrated by the following embodiments. First, the nonlinear activation layer for an optical neural network in the embodiments of the present application is described.

[0051] See also Figure 1 In one embodiment, the nonlinear activation layer comprises:

[0052] A binary non-linear activation device 100 comprising a two-dimensional material liquid crystal that deflects under the action of an electric field;

[0053] A polarizer 200, which is disposed on a first side of the binary nonlinear activation device 100;

[0054] The analyzer 300 is disposed on a second side of the binary nonlinear activation device 100 , where the second side is a side opposite to the first side. The polarization direction of the analyzer 300 is perpendicular to the polarization direction of the polarizer 200 .

[0055] The process of nonlinear activation may include: linear polarized light X is incident on the polarizer 200, the polarizer 200 filters the linear polarized light X, the filtered linear polarized light is incident on the binary nonlinear activation device 100, the filtered linear polarized light is modulated by the deflected two-dimensional material liquid crystal, the modulated linear polarized light is incident on the analyzer 300, and the activated linear polarized light is obtained. In this way, optical nonlinear activation can be achieved.

[0056] It should be noted that the azimuth of the linearly polarized light can be positive or negative. When the azimuth of the linearly polarized light is positive, the above-mentioned electric field is specifically the first electric field. When the azimuth of the linearly polarized light is negative, the above-mentioned electric field is specifically the second electric field. The electric field strength of the first electric field is different from the electric field strength of the second electric field, and the deflection of the two-dimensional material liquid crystal under the first electric field is different from the deflection under the second electric field. The filtered linearly polarized light is positively modulated by applying the first electric field, and the filtered linearly polarized light is negatively modulated by applying the second electric field. This embodiment can at least overcome the following technical obstacles: the current nonlinear optical activation is mainly based on light intensity coding, and negative value activation cannot be achieved / only has an odd function activation form that is symmetric about the origin, which makes it difficult for the network to capture the true distribution of data and may also lose important features.

[0057] In one embodiment, the binary nonlinear activation device 100 is a first nonlinear activation device for positive modulation, or a second nonlinear activation device for negative modulation. The first nonlinear activation device and the second nonlinear activation device both include a Kerr cell of a two-dimensional material liquid crystal. After the linear polarized light is incident on the polarizer 200 to obtain the filtered linear polarized light, the filtered linear polarized light is incident on the first side of the Kerr cell and is transmitted from the second side of the Kerr cell to obtain the modulated linear polarized light. The second side of the Kerr cell is the side opposite to the first side of the Kerr cell.

[0058] It should be noted that the Kerr box is an important device in optics that is realized by the Kerr effect. Kerr effect: refers to the phenomenon that the refractive index of certain materials changes under the action of an electric field. Usually, the change in refractive index is proportional to the square of the electric field strength. This effect causes the phase and amplitude of the light signal to change when passing through the material. The structure of the Kerr box: It is composed of a two-dimensional material liquid crystal with the Kerr effect. By applying an electric field, the refractive index of the material can be controlled, thereby modulating the light signal.

[0059] In one embodiment, referring to Figure 1 The Kerr cell of two-dimensional material liquid crystal includes: a Kerr cell container 110, which has a built-in functional layer, and the functional layer contains two-dimensional material liquid crystal; two electrodes 120, which are respectively arranged on opposite sides of the Kerr cell container 110. Specifically, the two electrodes are connected to an external power source to apply an electric field to the functional layer, and the orientation of the two-dimensional material liquid crystal is deflected under the action of the electric field.

[0060] It should be noted that the filtered polarized light is incident from one side of the Kerr cell of the two-dimensional material liquid crystal, and the orientation of the two-dimensional material liquid crystal in the functional layer is deflected under the action of the electric field. The filtered polarized light passes through the two-dimensional material liquid crystal to generate a phase delay and is transmitted from the Kerr cell. Since the binary nonlinear activation device uses photons as the transmission carrier of the signal, the calculation speed v can be expressed as Where c is the speed of light in free space, and n is the refractive index of the medium to the electromagnetic wave of this frequency, the response time of the binary nonlinear activation device is only in the order of picoseconds. The binary nonlinear activation device can modulate the intensity of the transmitted light based on the following law, namely Where I0 is the incident light, θ is the angle between the optical axis of the two-dimensional material liquid crystal and the filtered polarized light, and δ is the phase delay. Therefore, the binary nonlinear activation device can adjust the transmittance of different incident light by changing the electric field intensity to achieve reconfigurable optical activation. The two-dimensional material liquid crystal has a large electro-optical Kerr coefficient K, and the phase delay δ = 2πLKE 2 , where L is the optical path perpendicular to the electric field and E is the electric field intensity, so the binary nonlinear activation device can achieve full-precision full modulation of the phase within a smaller electric field range. In addition, since two-dimensional materials do not undergo photodegradation and have good stability to ultraviolet radiation, they can be used for ultraviolet light computing.

[0061] It should be noted that the binary nonlinear activation device can be used for positive modulation or negative adjustment by setting different electric field strengths. In order to reduce the complexity of the setting, a first nonlinear activation device for positive modulation and a second nonlinear activation device for negative modulation can be specially set. The first nonlinear activation device and the second nonlinear activation device are both binary nonlinear activation devices.

[0062] In one embodiment, the two-dimensional material liquid crystal includes an inorganic two-dimensional material and a dispersant.

[0063] In another embodiment, the two-dimensional liquid crystal material is a two-dimensional vermiculite aqueous dispersion, which can achieve π / 2 phase modulation under the action of an electric field, thereby improving the adjustment accuracy and thus improving the accuracy of nonlinear activation.

[0064] The two-dimensional vermiculite aqueous dispersion refers to the aqueous dispersion of the two-dimensional vermiculite material. In one example, the dispersion concentration of the two-dimensional vermiculite material is 9 g / L. The optical photograph of the two-dimensional vermiculite material liquid crystal is as follows: Figure 2 shown.

[0065] In one example, the Kerr cell container 110 can be an isotropic cuvette with an outer dimension of 12.5*12.5*45 mm. The electrode 120 can be a 45*10*0.06 mm double-sided conductive copper foil tape. After the aqueous dispersion of the two-dimensional vermiculite material is added to the Kerr cell container 110, the opening of the cuvette is sealed with a tetrafluoroethylene cover to obtain a binary nonlinear activation device with a diagonal dimension of 46.1 mm, such as Figure 3 shown.

[0066] In one embodiment, this embodiment provides a method for determining whether the binary nonlinear activation device in the above example has full phase modulation capability. Specifically, since the modulation object is an analog optical signal, full precision modulation can be achieved by reducing the electric field interval. After 35 days (840 hours) of testing, the phase modulation intervals of the binary nonlinear activation device with a test duration of 0 hours and 840 hours are basically overlapped, such as Figure 4 As shown, it has good cycle stability. Figure 4 In the figure, the abscissa is the electric field intensity, and the ordinate is the optical activation.

[0067] In one embodiment, this embodiment provides a method for determining the ultraviolet light stability of the binary nonlinear activation device in the above example. Specifically, at 4.1×10 7 mW -2 Under the irradiation of extremely strong ultraviolet light of level 1, the modulation effect of the binary nonlinear activation device was tested every 3 hours. After a cumulative 12 hours, the modulation effect of the binary nonlinear activation device was attenuated by only 1.29%. Figure 5 As shown, it can meet the application requirements under ultraviolet light. Figure 5 In the figure, the abscissa is the electric field intensity, and the ordinate is the optical activation.

[0068] In a comparative example, this comparative example provides a Kerr cell of a two-dimensional material liquid crystal, Figure 3 The difference between the corresponding examples is that the material of the functional layer is different, that is, the material of the functional layer is a two-dimensional graphene oxide aqueous dispersion. In the same electric field strength range, the phase modulation effect of the comparative example is tested, such as Figure 6 As shown. Figure 6 In the figure, the horizontal axis is the electric field intensity and the vertical axis is the optical activation. According to the test results, the response sensitivity of the two-dimensional vermiculite aqueous dispersion is higher, and full phase modulation can be achieved in a small electric field intensity range.

[0069] In another example, the Kerr cell container 110 can be a custom container made of isotropic material with an outer dimension of 95*15*92.5 mm, resulting in a binary nonlinear activation device with a diagonal dimension of 127.3 mm (5 inches), such as Figure 7 shown.

[0070] In another example, the Kerr cell container 110 may have an outer dimension of 1*0.04*1 mm, resulting in a binary nonlinear activation device with a diagonal dimension of 1.4 mm.

[0071] In one embodiment, the electric field applied to the functional layer is an alternating electric field. Specifically, the waveform of the alternating electric field can be a sine wave, a square wave, etc.; the frequency of the alternating electric field can be 10 1 Hz to 10 7Hz; the strength of the AC electric field can be selected as 10 2 V / m to 10 8 V / m.

[0072] In one embodiment, the AC electric field may be a sinusoidal AC electric field, and the frequency of the AC electric field may be 10 4 Hz, the strength of the AC electric field can be 10 3 V / m.

[0073] In one embodiment, the characteristic size of the binary nonlinear activation device 100 is adjustable. Specifically, the diagonal size is in the range of micrometer to centimeter.

[0074] In one embodiment, the diagonal size of the binary non-linear activation device 100 may be 1.4 mm, 46.1 mm, or 127.3 mm (5 inches).

[0075] In one embodiment, the incident direction of the linearly polarized light is selected to be the third side of the Kerr cell, that is, the adjacent surface where the electrode is installed. The incident angle of the linearly polarized light is selected to be perpendicular to the surface of the Kerr cell. Specifically, the linearly polarized light is orthogonal to the direction of the applied electric field in the functional layer.

[0076] In one embodiment, the linear polarized light can be selected as a laser with a wavelength of 300nm-700nm. For example, the wavelength of the linear polarized light can be 303nm, 450nm and 532nm.

[0077] In one embodiment, referring to Figure 1 and Figure 8The binary nonlinear activation device 100 includes a first nonlinear activation device 1001 and a second nonlinear activation device 1002 with perpendicular electric field directions, the polarizer 200 includes a first polarizer 2001 and a second polarizer 2002 with perpendicular polarization directions, and the analyzer 300 includes a first analyzer 3001 and a second analyzer 3002 with perpendicular polarization directions. The polarization direction of the first polarizer 2001 is the same as that of the second analyzer 3002, and the polarization direction of the second polarizer 2002 is the same as that of the first analyzer 3001. The nonlinear activation layer further includes a beam splitter 400, which is disposed on one side of the first nonlinear activation device 1001 and on one side of the second nonlinear activation device 1002. Wherein, the linear polarized light is incident on the beam splitter 400, and the beam splitter 400 divides the linear polarized light into a first linear polarized light and a second linear polarized light, and the polarization direction of the first linear polarized light is perpendicular to the polarization direction of the second linear polarized light; the first linear polarized light is incident on the first polarizer 2001, and the second linear polarized light is incident on the second polarizer 2002. Regardless of whether the azimuth angle of the polarized light is positive or negative, it will be divided by the beam splitter 400 into the first linear polarized light and the second linear polarized light, which are both positive, or the first linear polarized light and the second linear polarized light, which are both negative, and are incident on the first polarizer and the second polarizer respectively. The positive first linear polarized light can pass through the first polarizer 2001, but the negative first linear polarized light cannot (the polarization direction of the negative first linear polarized light is perpendicular to the polarization direction of the first polarizer, so it cannot pass through), and the negative second linear polarized light can pass through the second polarizer 2002, but the positive second linear polarized light cannot (the polarization direction of the positive second linear polarized light is perpendicular to the polarization direction of the second polarizer, so it cannot pass through). The benefit of this embodiment is that automatic positive and negative differentiation is achieved.

[0078] The first linear polarized light and the second linear polarized light not only have perpendicular polarization directions but also have equal light intensities. The first linear polarized light is incident on the first polarizer, the first polarizer filters the first linear polarized light, the filtered first linear polarized light is incident on the nonlinear activation device, the nonlinear activation device modulates the filtered first linear polarized light, and the modulated first linear polarized light is incident on the analyzer to achieve optical nonlinear activation. The second linear polarized light is incident on the second polarizer, the second polarizer filters the second linear polarized light, the filtered second linear polarized light is incident on the nonlinear activation device, the nonlinear activation device modulates the filtered second linear polarized light, and the modulated second linear polarized light is incident on the analyzer to achieve optical nonlinear activation.

[0079] In one embodiment, a linearly polarized light source provides linearly polarized light as an input optical signal, wherein the polarization azimuth of the linearly polarized light is 45° and is encoded as positive, and the polarization azimuth is -45° and is encoded as negative.

[0080] In one embodiment, the angle between the polarization direction of the first polarizer 2001 and the electric field direction of the first nonlinear activation device 1001 is 45°, and the angle between the polarization direction of the second polarizer 2002 and the electric field direction of the second nonlinear activation device 1002 is 45°. If the polarization azimuth of the linear polarized light is 45°, the first linear polarized light passes through the first nonlinear activation device 1001 and the first analyzer 3001 in sequence after entering the first polarizer 2001; if the polarization azimuth of the linear polarized light is -45°, the second linear polarized light passes through the second nonlinear activation device 1002 and the second analyzer 3002 in sequence after entering the second polarizer 2002.

[0081] The benefit of this embodiment is that the nonlinear activation layer can automatically perform nonlinear activation on linearly polarized light with polarization angles of 45° and -45°, thereby improving the degree of automation of nonlinear activation and facilitating the integration of optical neural networks.

[0082] In one embodiment, referring to Fig. 9 The nonlinear activation layer further includes: a quarter wave plate 500, which is disposed on one side of the beam splitter 400. The circularly polarized light Y is incident on the quarter wave plate 500, and the quarter wave plate 500 is used to convert the circularly polarized light Y into linearly polarized light. After the linearly polarized light is converted, the linearly polarized light is incident on the beam splitter prism. The optical path after the linearly polarized light is incident on the beam splitter prism has been described above (e.g. Figure 7 The embodiment shown in the figure) is described in detail and will not be repeated here.

[0083] It should be noted that after the incident circularly polarized light is converted into linearly polarized light by the quarter wave plate 500 (also called 1 / 4 wave plate), it passes through the splitter prism 400 and the first polarizer 2001 and the second polarizer 2002, and is selectively incident from one side of the Kerr cell of the two-dimensional material liquid crystal in the direction perpendicular to the electric field from a branch (the optical path where the first polarizer is located or the optical path where the second polarizer is located). The orientation of the two-dimensional material liquid crystal in the functional layer of the Kerr cell is deflected under the action of the electric field, and the filtered linearly polarized light generates a phase delay through the two-dimensional material liquid crystal, and is transmitted and emitted from the Kerr cell, achieving different activation effects on the linear polarized light with a positive or negative azimuth angle.

[0084] In one embodiment, the circularly polarized light source provides circularly polarized light as an input light signal, the left-handed circularly polarized light has a phase difference of -π / 2 and is encoded as negative, and the right-handed circularly polarized light has a phase difference of π / 2 and is encoded as positive.

[0085] In one embodiment, the angle between the polarization direction of the first polarizer 2001 and the electric field direction of the first nonlinear activation device 1001 is 45°, and the angle between the polarization direction of the second polarizer 2002 and the electric field direction of the second nonlinear activation device 1002 is 45°. If the phase difference of circularly polarized light is π / 2, the first linearly polarized light passes through the first nonlinear activation device 1001 and the first analyzer 3001 in sequence after entering the first polarizer 2001; if the phase difference of linearly polarized light is -π / 2, the second linearly polarized light passes through the second nonlinear activation device 1002 and the second analyzer 3002 in sequence after entering the second polarizer 2002.

[0086] The benefit of this embodiment is that the nonlinear activation layer can automatically perform nonlinear activation on circularly polarized light with phase differences of π / 2 and -π / 2, respectively, thereby improving the degree of automation of nonlinear activation and facilitating the integration of optical neural networks.

[0087] It can be understood that the present application completes the reconfigurable binary nonlinear activation layer in free space by encoding polarization characteristics as positive and negative signs, combining the relationship between the incident light and the outgoing light of the Kerr cell of the two-dimensional material liquid crystal, and utilizing the optical path design, thereby realizing reconfigurable binary optical activation of positive and negative optical input signals (linearly polarized light or circularly polarized light) in free space. The size of the binary nonlinear activation device can be expanded from microns to inches, the response speed is in the picosecond level, the response range covers the visible light range, and full-precision full modulation of the phase can be achieved in a small electric field range. The reconfigurable binary nonlinear activation layer solves the problem of delay and fixed response of optical activation, and the inability to process positive and negative inputs differently, expands the function of optical activation, and provides the possibility for the realization of all-optical neural networks for complex tasks.

[0088] In one embodiment, a nonlinear activation method for an optical neural network is applied to the nonlinear activation layer as described above, the method comprising:

[0089] Controlling the linear polarized light to be incident on the polarizer, the polarizer filters the linear polarized light, and the filtered linear polarized light is incident on the nonlinear activation device;

[0090] An electric field is applied to the nonlinear activation device to modulate the filtered linearly polarized light through the two-dimensional material liquid crystal that is deflected under the action of the electric field. The modulated linearly polarized light is incident on the analyzer to achieve optical nonlinear activation.

[0091] Specifically, a nonlinear activation layer based on incident linear polarized light is set, the incident linear polarized light is encoded positively and negatively based on the polarization azimuth, and the negative value response rule is independently defined to achieve reconfigurable binary optical activation in the entire range.

[0092] In one embodiment, before controlling the linear polarized light to be incident on the polarizer, the method further includes: controlling the circular polarized light to be incident on a quarter wave plate, and the quarter wave plate converts the circular polarized light into linear polarized light. After the conversion to obtain linear polarized light, the linear polarized light is incident on a beam splitter prism. The optical path after the linear polarized light is incident on the beam splitter prism has been described above (e.g. Figure 3 The embodiment shown in the figure) is described in detail and will not be repeated here.

[0093] In one example, a nonlinear activation implementation process for incident linearly polarized light may include:

[0094] The nonlinear activation layer based on incident linear polarized light is shown in the figure below. Figure 8 As shown. Its nonlinear activation layer includes: a beam splitter prism 400, a first polarizer 2001, a first nonlinear activation device 1001, a first analyzer 3001, a second polarizer 2002, a second nonlinear activation device 1002, and a second analyzer 3002. After passing through the beam splitter prism 400, the linearly polarized light X provides incident light to the first nonlinear activation device 1001 and the second nonlinear activation device 1002 of the vertical optical path. The first polarizer 2001 is arranged between the beam splitter prism 400 and the first nonlinear activation device 1001, and the second polarizer 2002 is arranged between the beam splitter prism 400 and the second nonlinear activation device 1002. The polarization direction of the first polarizer 2001 of the two vertical optical paths is perpendicular to the polarization direction of the second polarizer 2002, and forms an angle of 45° with the electric field direction. The first analyzer 3001 is arranged at the end of the optical path and on the opposite side of the first nonlinear activation device 1001 and the first polarizer 2001. The second polarizer 3002 is disposed at the end of the optical path and on the side opposite to the second nonlinear active device 1002 and the second polarizer 2002. The polarization direction of the first polarizer 3001 is orthogonal to the polarization direction of the first polarizer 2001. The polarization direction of the second polarizer 3002 is orthogonal to the polarization direction of the second polarizer 2002.

[0095] In this example, the linear polarized light X as the input optical signal is encoded as positive and negative when the polarization azimuth angle is 45° and -45°, respectively, and is divided into two perpendicular paths with equal light intensity and the same phase by the beam splitter prism 400. The first polarizer 2001 and the second polarizer 2002 filter the positive and negative inputs respectively. The incident light after filtering (also called beam splitting) is selectively incident from one side of the binary nonlinear activation device (the first nonlinear activation device 1001 or the second nonlinear activation device 1002) from a branch in a direction perpendicular to the electric field. The two-dimensional material liquid crystal in the functional layer of the Kerr cell is 10 4Hz, 1V / mm electric field, the input light signal is deflected, the input light signal passes through the two-dimensional material liquid crystal to generate phase delay, and is transmitted from the Kerr cell, achieving different activation effects for the input of positive and negative polarization azimuth encoding. Under this setting, the nonlinear activation characteristic curve of the binary nonlinear activation device is as follows Fig.10 As shown. Fig.10 In the figure, the horizontal axis is the electric field intensity and the vertical axis is the optical activation characteristics, which mainly shows the comparison between the binary activation of the two-dimensional vermiculite aqueous dispersion and the activation effect of the traditional linear rectification function.

[0096] In one example, a nonlinear activation implementation process for incident circularly polarized light may include:

[0097] The nonlinear activation layer based on incident circularly polarized light is shown in the following figure. Fig. 9 As shown. Its nonlinear activation layer includes: a quarter wave plate 500, a beam splitter 400, a first polarizer 2001, a first nonlinear activation device 1001, a first analyzer 3001, a second polarizer 2002, a second nonlinear activation device 1002, and a second analyzer 3002. The circularly polarized light Y passes through the quarter wave plate 500 and the beam splitter 400 in sequence and provides incident light to the first nonlinear activation device 1001 and the second nonlinear activation device 1002 of the vertical optical path. The first polarizer 2001 is arranged between the beam splitter 400 and the first nonlinear activation device 1001, and the second polarizer 2002 is arranged between the beam splitter 400 and the second nonlinear activation device 1002. The polarization direction of the first polarizer 2001 of the two vertical optical paths is perpendicular to the polarization direction of the second polarizer 2002, and forms an angle of 45° with the electric field direction. The first polarizer 3001 is disposed at the end of the optical path and on the side opposite to the first nonlinear activation device 1001 and the first polarizer 2001. The second polarizer 3002 is disposed at the end of the optical path and on the side opposite to the second nonlinear activation device 1002 and the second polarizer 2002. The polarization direction of the first polarizer 3001 is orthogonal to the polarization direction of the first polarizer 2001. The polarization direction of the second polarizer 3002 is orthogonal to the polarization direction of the second polarizer 2002.

[0098] In this example, circularly polarized light Y is used as an input optical signal and is encoded as positive and negative when the phase difference is π / 2 and -π / 2, respectively. It is converted into linearly polarized light by a quarter wave plate 500 and is divided into two perpendicular paths with equal light intensity and the same phase by a beam splitter prism 400. The first polarizer 2001 and the second polarizer 2002 filter the positive and negative inputs respectively. The incident light after beam splitting is selectively incident from one side of the binary nonlinear activation device (the first nonlinear activation device 1001 or the second nonlinear activation device 1002) from a branch in a direction perpendicular to the electric field. The two-dimensional material liquid crystal in the functional layer of the Kerr cell is 10 4Hz, deflected under the action of an electric field of 1V / mm, the input light signal (phase difference of π / 2 corresponds to a polarization azimuth of 45°, and phase difference of -π / 2 corresponds to a polarization azimuth of -45°) produces a phase delay through the two-dimensional material liquid crystal and is transmitted out from the Kerr cell, achieving different activation effects for the positive and negative inputs encoded with the polarization azimuth.

[0099] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0100] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0101] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0102] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0103] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0104] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0105] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0106] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable an electronic device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0107] The preferred embodiments of the present application are described above with reference to the accompanying drawings, but the scope of the rights of the present application is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present application should be within the scope of the rights of the present application.

Claims

1. A nonlinear activation layer for an optical neural network, characterized in that The nonlinear activation layer comprises: Binary nonlinearly activated devices, comprising a two-dimensional material liquid crystal that deflects under the action of an electric field; A polarizer, the polarizer being arranged on a first side of the binary nonlinear activation device; An analyzer, wherein the analyzer is arranged on a second side of the binary nonlinear activation device, the second side being a side opposite to the first side, and a polarization direction of the analyzer is perpendicular to a polarization direction of the polarizer; Among them, the linear polarized light is incident on the polarizer, and the polarizer filters the linear polarized light. The filtered linear polarized light is incident on the binary nonlinear activation device, and the filtered linear polarized light is modulated by the deflected two-dimensional material liquid crystal. The modulated linear polarized light is incident on the analyzer to achieve optical nonlinear activation.

2. The nonlinear activation layer according to claim 1, characterized in that The binary nonlinear activation device comprises a first nonlinear activation device and a second nonlinear activation device with perpendicular electric field directions, the polarizer comprises a first polarizer and a second polarizer with perpendicular polarization directions, the analyzer comprises a first polarizer and a second polarizer with perpendicular polarization directions, the polarization direction of the first polarizer is the same as the polarization direction of the second polarizer, and the polarization direction of the second polarizer is the same as the polarization direction of the first analyzer; The nonlinear activation layer further comprises: A beam splitter prism, the beam splitter prism being arranged on one side of the first nonlinear activation device and on one side of the second nonlinear activation device; Among them, the linear polarized light is incident on the beam splitter prism, and the beam splitter prism splits the linear polarized light into a first linear polarized light and a second linear polarized light, and the polarization direction of the first linear polarized light is perpendicular to the polarization direction of the second linear polarized light; the first linear polarized light is incident on the first polarizer, and the second linear polarized light is incident on the second polarizer.

3. The nonlinear activation layer according to claim 2, characterized in that: The first nonlinear activation device is used for positive modulation, the second nonlinear activation device is used for negative modulation, the angle between the polarization direction of the first polarizer and the electric field direction of the first nonlinear activation device is 45°, and the angle between the polarization direction of the second polarizer and the electric field direction of the second nonlinear activation device is 45°; If the polarization azimuth angle of the linearly polarized light is 45°, then after the first linearly polarized light is incident on the first polarizer, it passes through the first nonlinear activation device and the first analyzer in sequence; If the polarization azimuth angle of the linearly polarized light is -45°, then after the second linearly polarized light is incident on the second polarizer, it passes through the second nonlinear activation device and the second analyzer in sequence.

4. The nonlinear activation layer according to claim 2, characterized in that: The nonlinear activation layer further comprises: a quarter wave plate, wherein the quarter wave plate is arranged on one side of the beam splitter prism. Wherein, the circularly polarized light is incident on the quarter-wave plate, and the quarter-wave plate is used to convert the circularly polarized light into linearly polarized light.

5. The nonlinear activation layer according to claim 4, characterized in that: The first nonlinear activation device is used for positive modulation, the second nonlinear activation device is used for negative modulation, the angle between the polarization direction of the first polarizer and the electric field direction of the first nonlinear activation device is 45°, and the angle between the polarization direction of the second polarizer and the electric field direction of the second nonlinear activation device is 45°; If the phase difference of the circularly polarized light is π / 2, then after the first linearly polarized light is incident on the first polarizer, it passes through the first nonlinear activation device and the first analyzer in sequence; If the phase difference of the linearly polarized light is -π / 2, then after the second linearly polarized light is incident on the second polarizer, it passes through the second nonlinear activation device and the second analyzer in sequence.

6. The nonlinear activation layer according to any one of claims 2 to 5, characterized in that: The first non-linear activation device and the second non-linear activation device include: Kerr cells of two-dimensional material liquid crystal.

7. The nonlinear activation layer according to claim 6, characterized in that: The Kerr cell of the two-dimensional material liquid crystal comprises: A Kerr cell container having a built-in functional layer, wherein the functional layer contains a two-dimensional material liquid crystal; Two electrodes, the two electrodes are respectively arranged on opposite sides of the Kerr cell container; The two electrodes are connected to an external power source to apply an electric field to the functional layer, and the orientation of the two-dimensional material liquid crystal is deflected under the action of the electric field.

8. The nonlinear activation layer according to claim 7, characterized in that: The two-dimensional liquid crystal material is a two-dimensional vermiculite aqueous dispersion, which can achieve π / 2 phase modulation under the action of an electric field.

9. A nonlinear activation method for an optical neural network, characterized in that: Applied to the nonlinear activation layer according to any one of claims 1 to 8, the method comprises: Controlling linear polarized light to be incident on the polarizer, the polarizer filters the linear polarized light, and the filtered linear polarized light is incident on the nonlinear activation device; An electric field is applied to the nonlinear activation device to modulate the filtered linearly polarized light through the two-dimensional material liquid crystal that is deflected under the action of the electric field, and the modulated linearly polarized light is incident on the analyzer to achieve optical nonlinear activation.

10. The nonlinear activation method according to claim 9, characterized in that: Before the linear polarized light is controlled to be incident on the polarizer, the method further comprises: The circularly polarized light is controlled to be incident on a quarter wave plate, and the quarter wave plate converts the circularly polarized light into linearly polarized light.

Images