31results about How to "Effective convergence" patented technology

The invention discloses a broadband high-efficiency 1bit circularly polarized reconfigurable transmission unit and an array antenna. The reconfigurable transmission unit comprises a circularly polarized wave receiving patch, an upper-layer dielectric substrate, an upper-layer metal ground plane, a middle dielectric substrate, a DC bias line, a lower-layer metal ground plane, a lower-layer dielectric substrate and a radiation patch which are sequentially arranged from top to bottom, and can directly realize two unit phases of 0 degree and 180 degrees of circularly polarized transmission waves; the equivalent rotation phase is free of dispersion; the state of each unit in the transmission array is controlled through the FPGA, + / -60-degree large-angle scanning of circular polarization transmission beams can be achieved, the beam performance is good, and high circular polarization purity is achieved. The array antenna provided by the invention has the advantages of wide axial ratio bandwidth and high aperture efficiency, and has relatively high application potential and value in scenes requiring real-time electric control circular polarization beam scanning antennas, such as application of long-distance radar detection, communication systems and the like.

The invention discloses a classification processing method and device for target characteristic data. The method comprises the steps that a target characteristic data set is acquired; the target characteristic data set comprises target characteristic data; preprocessing the target characteristic data set to obtain a preprocessed data set; and carrying out classification identification processing on the preprocessed data set to obtain category information. According to the method, effective classification processing and service of complex multi-source heterogeneous target characteristic data are realized, correct extraction of target characteristics is realized, and the production efficiency and competitiveness of equipment are effectively improved.

The invention discloses a tomato maturity detection system and method based on deep learning, and relates to the technical field of fruit and vegetable detection models. Comprising the following steps: acquiring data from a robot perspective based on a natural scene of a tomato plantation, and making a tomato maturity detection data set; on the basis of a YOLO v5s model, an improved model YOLOv5s BCD model fusing a CA attention mechanism, a BiFPN neck feature fusion network and a Decup led Head head decoupling network is provided and serves as a tomato maturity detection model of the invention; experimental research of model training and effect comparison is introduced, the conditions of model evaluation and optimization are performed through a model comparison test, the tomato maturity grade is judged on the basis of the fruit red surface coloring proportion and in combination with fruit appearance characteristics, tomatoes with the required maturity grade are collected, the light weight and accuracy of edge side deployment are adapted, and the tomato maturity grade evaluation method is suitable for the tomato maturity grade evaluation and the tomato maturity grade evaluation and the tomato maturity grade evaluation and the tomato maturity grade evaluation and the tomato maturity grade evaluation. The system can be used as a visual system of a tomato harvesting robot, and technical support is provided for automatic picking.

The invention discloses a quick charging cable for a new energy charging pile, and relates to the technical field of new energy charging equipment, the quick charging cable comprises a plurality of conductor layers, insulating layers are sleeved among the outer surface walls of the plurality of conductor layers, filling layers are sleeved among the outer surface walls of the plurality of insulating layers, and temperature conducting layers are fixedly inserted among the inner surface walls of the filling layers. An isolation plate is fixedly inserted into the inner surface wall of the heat conduction layer, and a plurality of air holes are formed in one side of the outer wall of the isolation plate. According to the invention, through the interaction of all layers of materials, heat generated in the electric power transmission process can be effectively gathered to the center, heat dissipation is carried out in a wind power circulation mode, and under the cooperation of multi-layer isolation, internal and external heat isolation is realized, and the problem of interference caused by mutual circulation is avoided; the equipment guarantees the safety of the cable and the surrounding environment from the aspects of heat insulation, fire prevention and heat dissipation, greatly improves the operation stability of a charging system, and also guarantees the electrical safety in the charging process.

The invention discloses a pancreatic focus CT image segmentation method based on CF-HRnet. According to the method, HRNet with imported pre-training weight is used for training on an NIH Pancreas data set, a coarse-to-fine training framework in an RSTN algorithm is used for training, two steps of preprocessing and joint training are added on the basis of original RSTN coarse segmentation and fine segmentation steps, and the network performance is further improved by increasing learning levels and training batches, so that the network performance is improved. The key optimization point is to introduce a model parameter transfer function and refine the RSTN segmentation process, and more accurately guide the neural network to identify the position information and detail information of the pancreatic focus. According to the method, the number of model parameters is greatly reduced while the focus region segmentation accuracy is improved, and powerful support is provided for diagnosis of pancreatic diseases.

The invention discloses a mismatching disturbance observation method and device, and is applied to the technical field of control, and the method comprises the steps: building a motion equation of a controlled system and an auxiliary system corresponding to the controlled system; based on the motion equation of the controlled system and the auxiliary system, establishing an error auxiliary system; designing a sliding mode variable and a control law of the auxiliary system according to the state of the error auxiliary system; based on the control law, the error auxiliary system converges to a sliding mode surface with the sliding mode variable being zero in the approaching process; and designing an observation value of mismatching disturbance through the motion equation of the controlled system and the sliding mode variable.

The utility model provides a heliostat mirror frame, include: main girder, set up first support layer on main girder, the second support layer who is configured for bearing sub mirror, second support layer includes a plurality of second support frame interval setting on first support layer, and at least two second support frames form a sub mirror mounting area for bearing sub mirror, at least part second support frame's length is different, and / or, at least part second support frame's extension direction is different to form different sub mirror mounting area. The utility model provides technical scheme to design to a plurality of second support frame length and extension direction, make a plurality of sub mirror mounting area formed by a plurality of second support frame have at least part difference, and then make a plurality of sub mirror mounting area can be according to actual demand and splice the heliostat of different contour and install splice out different shape contour, be favorable to according to actual environment and other factors more accurately design and control the focusing effect of heliostat, improve photo-thermal conversion efficiency.

Single image super-resolution aims at recovering a high-resolution image from a low-resolution image, and is an important and challenging task in computer vision. With wide application of streaming media and high-definition equipment, development of an efficient and high-precision super-resolution method becomes an important subject. A traditional deep learning method generally faces relatively high calculation overhead although the image reconstruction performance is remarkably improved, and the application of the traditional deep learning method on resource-limited equipment is limited. In order to solve the problem, the invention provides a lightweight super-resolution model based on a Transform structure. An efficient multi-scale attention module (PEMA) and a simplified feed-forward network (SAB) are designed, so that the calculation complexity is effectively reduced, and the model performance is improved.

The utility model discloses rapid deicing equipment after nuclear magnetic resonance quenching, and relates to the field of nuclear magnetic resonance structure maintenance equipment, the rapid deicing equipment after nuclear magnetic resonance quenching comprises a heating piece and an accelerating piece, the heating piece is provided with a gas inlet and a gas outlet, gas enters the heating piece through the gas inlet, the heating piece heats the gas, and the accelerating piece accelerates the heating of the gas. The heated gas is discharged through the gas outlet; the heating part is connected with the gas outlet through a pipeline, the accelerating part is connected with the gas outlet through a pipeline, the accelerating part is communicated with the gas outlet through a pipeline, the gas is accelerated through the accelerating part, and'ice blocks' in the nuclear magnetic resonance equipment can be effectively removed through the heating part and the accelerating part.

The invention discloses a pumping ring structure-based active optical fiber with high pumping conversion efficiency and an amplifier structure, and relates to the field of active optical fibers and optical fiber amplifiers, the pumping ring structure-based active optical fiber comprises a layered structure which is sequentially arranged from inside to outside, and the layered structure comprises a fiber core, a pumping ring structure, a quartz cladding and a coating layer, the optical fiber is located at the geometric center of the active fiber; the pumping ring structure is arranged around the fiber core and is used for guiding and converging pumping light to gather towards a fiber core area, the pumping ring structure is composed of one or more layers of concentric annular modulation areas, and the refractive indexes of all the layers of modulation areas are gradually increased layer by layer from outside to inside in the radial direction; according to the active optical fiber and amplifier structure, one or more layers of concentric pumping rings with refractive indexes gradually increased layer by layer from outside to inside are arranged in the cladding, and the concentration degree and energy density of pumping light towards a fiber core area can be remarkably improved to a certain extent by utilizing the guiding effect of refractive index gradients on a pumping light propagation path; therefore, higher pumping absorption efficiency and gain performance are realized.

The present invention relates to the field of space technology and vibration control, and in particular to a vibration suppression control method for a space microgravity vibration isolation system driven by both knowledge and data. First, a dynamic model of a space microgravity vibration isolation system is constructed and used as an initial knowledge-driven module. Then, a knowledge-driven nominal preset performance control law is designed; based on the constructed vibration isolation dynamics model, a preliminary vibration suppression control strategy is designed. Secondly, a data-driven vibration suppression strategy and update law are designed. Through the data-driven module and real-time monitored vibration data, a neural network algorithm is used to approximate the uncertainty and disturbance conditions in the current environment. Furthermore, sample generation and network training based on the nominal controller are carried out to simulate the system dynamics behavior under different disturbances. Finally, online training and application based on dual drive of knowledge and data are performed. The present invention can perform adaptive adjustments for different vibration sources and provide efficient and accurate vibration isolation control.

The invention provides a liquefied gas leakage video image autonomous monitoring method and system based on deep learning. The method comprises the following steps: S1, acquiring a liquefied gas leakage image data set; s2, performing data enhancement on the liquefied gas leakage image data set; s3, a liquefied gas leakage monitoring model Yolov5s-LPG is constructed on the basis of the network model Yolov5s; s4, using the liquefied gas leakage image data set after data enhancement to train the liquefied gas leakage monitoring model Yolov5s-LPG; and S5, on the basis of the trained liquefied gas leakage monitoring model Yolov5s-LPG, carrying out real-time monitoring on liquefied gas leakage. According to the invention, liquefied gas leakage monitoring based on deep learning is realized, the problems of poor real-time performance and lack of autonomous discrimination ability in the existing liquefied gas leakage monitoring means can be solved, and the problems existing in the liquefied gas leakage monitoring means based on deep learning are also solved.

The utility model discloses a flash lamp shade of eccentric light compensation, include: main part, stretch out, light entrance surface and light exit surface, main part with stretch out department link, light entrance surface set up in the bottom of main part, it has annular fresnel tooth, light exit surface sets up in the top of stretch out department, light exit surface is inclined surface. The utility model has beneficial effect in that, set up the angle of light exit of inclined light exit surface adjustment, suitable for the case of luminescent device and lampshade eccentricity, guarantee light -emitting even, and the annular fresnel tooth of light entrance surface can carry out effective convergence and homogenization treatment to incident light, thereby promote overall light compensation effect.

The embodiment of the present specification provides an optical sensor. The optical sensor comprises a microlens array composed of at least two microlenses and a sensor chip; the sensor chip is provided with photosensitive elements corresponding to the microlenses respectively; the microlens converges the light irradiated to the microlens to the photosensitive element corresponding to the microlens; the microlenses in the microlens array located at different field positions on the sensor chip have different lens structures, so as to improve the ratio of the illumination intensity of the light received by the photosensitive element corresponding to the microlens at the edge position to the illumination intensity of the light received by the photosensitive element corresponding to the microlens at the center position; and the photosensitive element is used for generating a corresponding photosensitive electrical signal according to the illumination intensity of the received light. The above-mentioned optical sensor improves the relative illumination of the array as a whole, and further improves the presentation effect of the finally obtained image.

The utility model belongs to the field of automobile bulbs, and particularly relates to a high-light-transmission automobile bulb which comprises a lampshade, a lamp wall, a lamp holder, a lamp tube, a scattering mirror, a reflecting mirror and a threaded connection pipe, the lampshade is connected with the lamp wall through the threaded connection pipe, and the scattering mirror and the reflecting mirror are arranged in the lampshade. The scattering mirror is arranged in an arc shape, has an opening facing the lamp tube and is used for uniformly scattering light emitted by the lamp tube; the reflecting mirror is arranged in a paraboloid shape, the focus is located in the center of the lamp tube, and the reflecting mirror is used for efficiently converging light emitted by the lamp tube and reflecting the light to the scattering mirror. The lampshade is made of high-light-transmittance materials, the light transmittance is not lower than 95%, and light loss is reduced to the maximum extent. By optimizing the light path design and selecting the high-light-transmittance material, the bulb lighting efficiency and quality are improved, meanwhile, the structure is compact, assembling is convenient, and the bulb can be widely applied to the field of various automobile lighting.

The invention provides a plant leaf chloroplast three-dimensional imaging axial resolution enhancement method and system, and the method comprises the steps: constructing an in-vitro model sample through fluorescent microspheres, and the in-vitro model sample comprises a non-scattering true value image and an axial stretching distortion image caused by scattering of the same visual field imaging; pre-training the deep learning network by using the in-vitro model sample, and establishing a pre-training model; constructing a plant real sample of plant leaf chloroplast, and training the pre-training model to obtain an axial resolution enhancement model; and inputting a to-be-processed chloroplast three-dimensional fluorescence image collected in a complete leaf into the axial resolution enhancement model, and outputting a chloroplast image used for accurate three-dimensional reconstruction after axial resolution enhancement. According to the method, hardware does not need to be transformed, the axial deformation of the plant leaf chloroplast in three-dimensional imaging is specifically corrected through an innovative training data preparation strategy and supervised deep learning, and the real three-dimensional form of the chloroplast is recovered.

The invention discloses a generation method, a simulation method, a constraint method and a device for a magnetic configuration of a magnetic material, and belongs to the technical field of material science and computers, and the generation method comprises the following steps: obtaining an initial magnetic configuration and an initial atomic configuration of the magnetic material, and randomly generating a plurality of random mechanism type combinations; selecting one of a target magnetic configuration and a target atomic configuration, and obtaining a constraint constant and a wave function of the target magnetic configuration by adopting a self-consistent iteration mode according to the Hamiltonian of the magnetic configuration; determining and storing magnetic system energy, temperature and pressure intensity of the target magnetic configuration; returning to select a new follow-up mechanism type combination, obtaining an initial magnetic configuration and an initial atomic configuration using new different magnetic materials after all the combinations are completed, and repeating the steps; and obtaining the magnetic material with the specified phase transition temperature through the magnetic configuration library by using a deep learning model. According to the method, the target magnetic configuration can be simulated, a magnetic configuration library is formed, and synthesis of a target magnetic material is guided.

This application discloses an optical lens, a camera module, and an electronic device that can meet the requirements of miniaturization and high imaging quality, and can stably image under complex lighting conditions. The optical lens has six refractive lenses, which are arranged sequentially from the object side to the image side along the optical axis: the first lens has negative refractive power, and its object side is concave near the optical axis, and its image side is concave near the optical axis; the second lens has positive refractive power, and its object side is convex near the optical axis, and its image side is convex near the optical axis; the third lens has positive refractive power, and its object side is convex near the optical axis; the fourth lens has positive refractive power, and its object side is convex near the optical axis, and its image side is convex near the optical axis; the fifth lens has negative refractive power, and its object side is concave near the optical axis; the sixth lens has negative refractive power, and its image side is concave near the optical axis; the optical lens satisfies the following relationship: 30°≤FOV≤40°, 1≤CT4 / CT5≤1.4.

The sewage filter material centrifugal machine comprises an installation platform, a lower barrel body is fixedly installed on the installation platform, an upper cover plate is installed on the lower barrel body in an overturning mode through a cover overturning power device, a bearing seat is fixedly installed in the lower barrel body, and a driving shaft is rotatably installed on the bearing seat; the driving shaft is fixedly connected with the rotary drum and is rotatably mounted in the lower barrel body, the driving shaft penetrates through the lower barrel body and the mounting platform, a plurality of centrifugal drainage holes are formed in the side wall of the rotary drum, a filter bag rotating synchronously is mounted in the rotary drum, and a pushing rod for pushing the filter bag is mounted in the driving shaft in a lifting manner. The pushing rod is adjusted to ascend and descend through a pushing device, the driving shaft is driven to rotate through a rotating power device, a feeding pipe and an exhaust pipe are arranged on the upper cover plate, a liquid outlet is formed in the bottom of the lower barrel body, and the liquid outlet penetrates through the mounting platform; the centrifugal machine can effectively separate flocculates in sewage to complete solid-liquid separation, and meanwhile, can effectively remove odor.

The invention relates to the technical field of geotechnical engineering and artificial intelligence crossing, and discloses a grouting data enhancement method and system for an improved table generative adversarial network, and the method comprises the steps: obtaining grouting table data, carrying out the preprocessing, obtaining the preprocessed grouting table data, and generating a condition vector; processing the preprocessed grouting table data and the condition vector by using a data enhancement model; outputting the synthesized grouting table data and the authenticity judgment result; training a generator module and a discriminator module by using a loss function according to the authenticity discrimination result, respectively configuring learning rate adaptive schedulers for the generator module and the discriminator module, and adjusting the learning rate according to the loss change; and based on the trained data enhancement model, generating enhanced grouting table data and performing similarity evaluation. A small sample grouting data set can be effectively expanded, and sufficient training samples are provided for a data-driven grouting parameter prediction model.

The present invention relates to an ellipsoidal MPCVD device for growing diamond materials, which belongs to the field of diamond material growth equipment. The present invention aims to solve the problem of low microwave energy coupling efficiency during the operation of the existing MPCVD resonant cavity. A microwave power supply is provided at one end of the rectangular waveguide in the ellipsoidal MPCVD device for growing diamond materials of the present invention, a short-circuit piston is provided at the other end of the rectangular waveguide, an ellipsoidal resonant cavity is provided at the lower part of the rectangular waveguide, and one end of the microwave antenna extends into the ellipsoidal resonant cavity through the waveguide. The cavity wall of the ellipsoidal resonant cavity is ellipsoidal, and the interior of the cavity wall of the ellipsoidal resonant cavity is a hollow structure. The sample stage is provided at the bottom of the ellipsoidal resonant cavity, and the exhaust system and the air intake system are respectively connected to the ellipsoidal resonant cavity through pipelines. The present invention utilizes the dual-focus characteristics of the ellipsoid to achieve high-density plasma excitation and deposition growth of diamond.

This invention relates to a low-frequency receiving antenna based on a gradient metamaterial with critical coupling, belonging to the field of low-frequency magnetic induction communication technology. The low-frequency receiving antenna includes: a receiving coil connected in series with a capacitor and connected to a receiver load via the capacitor; and a gradient metamaterial composed of multiple concentric resonant units with size gradients, each resonant unit consisting of a coil and a capacitor connected in series to form an independent resonant circuit. The receiving coil and the gradient metamaterial are concentrically nested, with the receiving coil located on the innermost side. This invention's low-frequency receiving antenna achieves matching between the antenna output impedance and the receiver load by precisely controlling the coupling strength between the gradient metamaterial and the receiving coil, as well as the capacitor configuration parameters. Simultaneously, the wide spacing between adjacent coils significantly reduces proximity losses. With the optimal combination of impedance matching and loss performance, the antenna reaches a critical coupling state. Combined with the increasing equivalent permeability distribution from the outside to the inside formed by the gradient metamaterial, it effectively converges low-frequency magnetic fields, improving the pickup capability of weak low-frequency electromagnetic signals. This invention maintains a compact size for the low-frequency antenna while possessing high receiving sensitivity, making it suitable for low-frequency magnetic field communication and detection in complex environments such as underwater and underground.

This utility model discloses a ventilation window guide hood for livestock, including a connecting sleeve and a guide plate. The connecting sleeve has multiple connecting through holes for fixing. The guide plate is located on the front side of the connecting sleeve and includes a central converging plate, a connecting transition plate, an open guide plate, and side plates. The central converging plate is located in the middle of the entire guide plate and extends upward and forward at an angle. The connecting transition plate and the open guide plate are connected to both sides of the central converging plate from back to front. The connecting transition plate is inclined downward and forward, and the open guide plate is inclined upward and forward. The side plates are fixedly connected to the ends of the connecting transition plate and the open guide plate. This utility model has a simple structure, is quick to install, requires no subsequent debugging, and has a wide range of applications. It can effectively gather fresh air, ensuring that even when the ventilation window is opened at a very small angle, fresh air can flow to a farther area through the central air-gathering jet area of ​​the guide hood, achieving an ideal ventilation effect.

The invention provides an eye protection lighting lamp capable of thickening choroid and further controlling eye axis growth, and relates to the field of lamps, the eye protection lighting lamp comprises a base plate, a supporting rod, a main lamp and a phototherapy lamp, the phototherapy lamp comprises a lamp shell, the lamp shell comprises an upper lamp shell and a lower lamp shell, a heat dissipation shell is arranged at the top of the upper lamp shell, and a lamp panel electrically connected with a main substrate in the main lamp is arranged in the heat dissipation shell; a plurality of 650nm red light lamp beads are arranged on the lamp panel, an inner lamp shell is arranged in the lower lamp shell, a lens is fixed on the inner lamp shell, a groove is formed in the position, located at the irradiation position of the red light lamp beads, of the top of the lens, a first convex lens is arranged at the bottom of the groove, a trapezoidal groove is formed in the bottom of the lens, and a second convex lens is arranged at the bottom of the trapezoidal groove. Through the design of an external phototherapy lamp lens structure, effective convergence of a light source is achieved, a 650nm red light lamp bead light source is converged to the periphery of the eyes of a user, the functions of thickening the choroid and further controlling the growth of the eye axis are achieved, and the eye nursing effect on the user is improved.

The utility model discloses an air duct structure of a water anion machine, which comprises an air inlet, an air outlet, a water inlet and a water outlet, wherein the diameter of the air inlet is larger than that of a fan; the air outlet is trumpet-shaped, and the area of the air inlet end of the air outlet is larger than that of the air outlet end. The negative ion generator has the technical effects and advantages that through the structural design of the air inlet and the air outlet, air can be effectively gathered, larger air pressure can be generated, and negative ions can be blown to a farther distance, so that the negative ion generator is adaptive to a space with a larger area.

The present invention provides a method for automatically identifying alarm categories and alarm convergence rules, the method comprising: obtaining batch alarm data; constructing an alarm convergence rule base according to the batch alarm data; processing real-time alarm data based on the alarm convergence rule base, converging alarm information, and pushing the alarm information to operation and maintenance personnel. The present invention has good learning and business adaptability. While achieving effective convergence of synchronous / asynchronous alarm information and reducing the interference of invalid alarms to operation and maintenance personnel, it also reduces the complexity of convergence rule configuration and improves the utilization rate of operation and maintenance resources.

The invention discloses a carrier speed estimation method and device based on a millimeter wave radar, and a storage medium, and relates to the technical field of carrier speed estimation. The method comprises the following steps: firstly, carrying out interference suppression processing on a distance-velocity spectrum matrix to obtain an interference suppression spectrum matrix; the first path performs velocity dimension energy aggregation on the interference suppression spectrum matrix to generate a static background energy distribution spectrum and determine a first velocity estimation value according to the static background energy distribution spectrum; a second path screens and verifies an effective ground reflection target from the interference suppression spectrum matrix, and a second speed estimation value is obtained by combining the pitch angle of the effective ground reflection target; and finally, respectively carrying out time sequence filtering and weighted fusion on the outputs of the two paths, and outputting the final carrier speed. According to the method, the problem that a traditional point cloud method fails in an open scene is fundamentally solved, and the robustness, accuracy and stability of a spectrum analysis method in a complex interference environment are remarkably improved.

The invention discloses an oil control liquid composition with a pore minimizing function and a preparation method of the oil control liquid composition, and belongs to the technical field of daily chemical products. The composition is prepared from the following components in parts by weight: 2 to 5 parts of mesenchymal stem cell extract, 10 to 15 parts of herba portulacae extract, 10 to 15 parts of rosemary extract, 5 to 10 parts of sodium hyaluronate, 5 to 10 parts of salicylic acid, 10 to 15 parts of tea tree essential oil, 10 to 15 parts of honey paste and 15 to 48 parts of sterile water. According to the invention, the biological extract is used as a main component, so that pores can be effectively astringed, the problem of excessive skin grease secretion is solved, the irritation to the skin is small, and the skin is not excessively dry or even sensitive or decrustation phenomenon is avoided; meanwhile, the mesenchymal stem cell extract in the formula contains rich growth factors and cell factors, can promote regeneration and repair of skin cells, can promote metabolism and regeneration of the skin cells, can accelerate apoptosis of coarse pores with problems, accelerates growth of new pores, and fundamentally solves the problem of coarse pores.

The present application relates to an imaging lens, an endoscope objective lens and an endoscope. The imaging lens comprises: a first lens with negative focal length, the image side surface is concave at the near optical axis; a second lens with positive focal length, the image side surface is convex at the near optical axis; a third lens with positive focal length, the object side surface and the image side surface are both convex at the near optical axis; a fourth lens with negative focal length, the object side surface is concave at the near optical axis, and the image side surface is convex at the near optical axis; and the imaging lens satisfies: 150 deg / mm <= FOV / SD11 <= 189 deg / mm. The above imaging lens can balance miniaturization design and good imaging quality.

The invention discloses a gene interaction relationship prediction method and device based on a graph neural network, and the method comprises the steps: carrying out the preprocessing of the transcription expression data of each experimental group, including the standardization of a gene ID, and representing the gene transcription expression data of different experimental groups through a unified gene ID; obtaining original data for subsequent analysis; performing gene similarity calculation according to each sample group to obtain similarity data between genes in the sample groups; performing statistics and screening on data of all sample groups, and selecting gene pairs meeting set conditions as a basis for constructing a graph data structure; further processing the original data to ensure that all sample groups only contain genes defined in the graph structure; then, carrying out standardization processing on the data of the sample groups, so that the data dimensions of all the sample groups are consistent; and carrying out model training on the generated graph data set to finally obtain edge weight data in a graph structure, and accurately reflecting an interaction relationship between genes. The method can be used for analyzing gene expression data, predicting the interaction relationship between genes, and carrying out unsupervised learning under the condition of no label data.