Feature vector generation method and device

Abstract

The invention discloses a feature vector generation method and device, and relates to the technical field of computers. A specific embodiment of the method comprises the steps of obtaining feature data and a feature score of a target object, generating an initial vector of the target object based on a first neural network, and further generating a prediction score of the target object through a second neural network, and adjusting parameters of the first neural network and the second neural network based on the prediction score and the feature score, so that a vector output by the first neural network after parameter adjustment is a feature vector of the target object. According to the embodiment, the feature vector of the target object can be determined based on the feature data of the target object, and specifically, the feature vector can be accurately calculated through the trained neural network.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and device for generating a feature vector. Background Art

[0002] In the logistics industry, in order to improve the freight transfer efficiency and service quality of each region, it is often necessary to predict data such as cargo volume and delivery rate based on regional geographic entities (Area of ​​Interest, AOI) (such as a community) in map data based on machine learning, deep learning and other technical solutions. In these technical solutions, the basic information of AOI is the key data of the learning model, and the feature representation based on the key data largely determines the effect of the learning model.

[0003] In the prior art, there are two main types of features that characterize AOI, numerical and categorical. When a numerical feature is input into a machine learning model, it is represented as a real number, and when a categorical feature is input into the model, it is generally represented as an integer. However, the dimensions of the two types of feature data and each feature are different, and there may be data with larger values ​​and data with smaller values ​​at the same time. When performing machine learning, it is impossible to balance the influence of each feature well, which affects the results of machine learning. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a method and device for generating a feature vector, which can accurately calculate a feature vector representing the characteristics of a target object based on the feature data of the target object and a neural network.

[0005] To achieve the above object, according to one aspect of an embodiment of the present invention, a method for generating a feature vector is provided, comprising:

[0006] Acquire feature data of a target object and a feature score of the target object;

[0007] Inputting the characteristic data of the target object into a first neural network to generate an initial vector of the target object;

[0008] Inputting the initial vector of the target object into a second neural network to obtain a predicted score of the target object;

[0009] Repeating the process of adjusting the parameters of the first neural network and the second neural network based on the prediction score and the feature score until a preset stop condition is met, and using the first neural network after parameter adjustment as a feature extraction network;

[0010] A feature vector of the target object is generated based on the feature data of the target object and the feature extraction network.

[0011] Optionally, adjusting the parameters of the first neural network and the second neural network based on the prediction score and the feature score includes: calculating a loss value based on the prediction score and the feature score; and adjusting the parameters of the first neural network and the second neural network based on the loss value.

[0012] Optionally, the preset stop condition is that the loss value is not greater than a preset loss threshold, or the number of parameter adjustments is greater than or equal to a preset number threshold.

[0013] Optionally, obtaining the feature score of the target object includes: inputting the feature data of the target object into a preset scoring system to obtain the feature score of the target object.

[0014] Optionally, the feature data of the target object is input into a preset scoring system to obtain a feature score of the target object, including: determining the scoring information of each feature of the target object through the preset scoring system and the feature data of the target object; and calculating the feature score of the target object based on the scoring information of each feature of the target object.

[0015] Optionally, the scoring information of each feature of the target object is determined through the preset scoring system and the feature data of the target object, including: determining a scoring rule corresponding to the target object according to a type feature in the feature data of the target object; and calculating the scoring information of each feature of the target object according to the scoring rule corresponding to the target object through the preset scoring system.

[0016] According to another aspect of an embodiment of the present invention, there is provided a device for generating a feature vector, comprising:

[0017] An acquisition module, used to acquire feature data of a target object and a feature score of the target object;

[0018] An input module, used to input the characteristic data of the target object into a first neural network to generate an initial vector of the target object; input the initial vector of the target object into a second neural network to obtain a predicted score of the target object;

[0019] an adjustment module, configured to repeatedly adjust the parameters of the first neural network and the second neural network based on the prediction score and the feature score until a preset stop condition is met, and use the first neural network after parameter adjustment as a feature extraction network;

[0020] A generating module is used to generate a feature vector of the target object based on the feature data of the target object and the feature extraction network.

[0021] Optionally, an adjustment module, used to adjust the parameters of the first neural network and the second neural network based on the predicted score and the feature score, includes: a calculation module, used to calculate a loss value based on the predicted score and the feature score; and an adjustment module, used to adjust the parameters of the first neural network and the second neural network based on the loss value.

[0022] Optionally, the preset stop condition is that the loss value is not greater than a preset loss threshold, or the number of parameter adjustments is greater than or equal to a preset number threshold.

[0023] Optionally, an acquisition module, used for the feature score of the target object, includes: an input module, further used to input the feature data of the target object into a preset scoring system to obtain the feature score of the target object.

[0024] Optionally, the input module is also used to input the feature data of the target object into a preset scoring system to obtain the feature score of the target object, including: a determination module, used to determine the scoring information of each feature of the target object through the preset scoring system and the feature data of the target object; and a calculation module, further used to calculate the feature score of the target object based on the scoring information of each feature of the target object.

[0025] Optionally, a determination module is used to determine the scoring information of each feature of the target object through the preset scoring system and the feature data of the target object, including: a determination module is used to determine the scoring rule corresponding to the target object according to the type feature in the feature data of the target object; and a calculation module is also used to calculate the scoring information of each feature of the target object according to the scoring rule corresponding to the target object through the preset scoring system.

[0026] According to another aspect of an embodiment of the present invention, there is provided an electronic device, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method for generating a feature vector provided in an embodiment of the present invention.

[0027] According to another aspect of the embodiments of the present invention, a computer-readable medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method for generating a feature vector provided by the embodiments of the present invention is implemented.

[0028] According to another aspect of the embodiments of the present invention, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the method for generating a feature vector provided by the embodiments of the present invention is implemented.

[0029] An embodiment of the above invention has the following advantages or beneficial effects: by inputting the feature data of the target object into the first neural network to obtain an initial vector, and inputting the initial vector into the second neural network to obtain a predicted score, the parameters of the first neural network and the second neural network are adjusted based on the predicted score and the obtained feature score of the target object, so that the vector output by the first neural network is the feature vector of the target object, and accurate calculation of the feature vector corresponding to the target object data is achieved. The feature vector calculated by the above method is a high-dimensional embedding vector, which can better characterize the characteristics of the target object and can be used in various other machine learning scenarios.

[0030] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

[0032] Figure 1 is a schematic diagram of main steps of a method for generating a feature vector according to an embodiment of the present invention;

[0033] Figure 2 A schematic diagram of a training process of a neural network provided by an embodiment of the present invention;

[0034] Figure 3 is a schematic diagram of main modules of a device for generating a feature vector according to an embodiment of the present invention;

[0035] Figure 4 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0036] Figure 5 It is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0038] It should be noted that the collection, collection, updating, analysis, processing, use, transmission, storage and other aspects of user personal information involved in the technical solution disclosed in the present invention are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information, network security and national security.

[0039] It should be noted that in the technical solution of the present invention, the collection, use, storage, sharing and transfer of user personal information involved are in compliance with the provisions of relevant laws and regulations, and it is necessary to inform the user and obtain the user's consent or authorization. When applicable, the user's personal information is de-identified and / or anonymized and / or encrypted.

[0040] Figure 1 FIG. 1 is a schematic diagram of the main steps of the method for generating a feature vector according to an embodiment of the present invention. Figure 1 As shown, the method for generating a feature vector in the embodiment of the present invention mainly includes the following steps S101 to S105.

[0041] Step S101, acquiring feature data of a target object and a feature score of the target object.

[0042] Get the feature data of AOI (target object), which is the data related to the difficulty of logistics and express delivery of AOI and the parameter data of AOI itself, including: weekly average number of successful delivery, monthly average number of successful delivery, timely delivery rate, customer complaint rate, location code, whether it is a residential area, etc. Get the feature score of the target object determined based on the above features, which is the score of the difficulty of AOI delivery, used to characterize the difficulty of logistics delivery within the AOI, and can be used as a label for the target object feature vector generation.

[0043] Optionally, different features can be selected in different application scenarios, and feature scores can also be calculated in different ways, not limited to the score of the difficulty of collection and delivery in the above-mentioned logistics scenario. For example, in the product recommendation scenario, the accuracy of the product recommendation is calculated based on data such as the user's click-through rate, order rate, and add-to-cart rate. Different feature scores can also be used in the same scenario. For example, in the logistics scenario, the score of delivery efficiency, the score of delivery quantity, etc. can be flexibly selected according to actual needs, and there is no restriction on this.

[0044] Step S102: inputting feature data of the target object into the first neural network to generate an initial vector of the target object.

[0045] The acquired AOI feature data is input into the first neural network to generate an N-dimensional AOI embedding vector as the initial vector of the target object, which is used to characterize the difficulty of picking up objects within the AOI range. The value of the dimension N of the AOI embedding vector output by the first neural network can be set according to the actual situation. For example, it can be set according to the data input requirements of the downstream task. That is, if the downstream task requires the input of a four-dimensional vector, then N can be set to 4, and the first neural network outputs the corresponding four-dimensional vector. The first neural network is a one-layer neural network; since the initial parameters of the first neural network are randomly initialized through Gaussian distribution, the initial vector of the output target object is also a random vector.

[0046] Step S103: input the initial vector of the target object into the second neural network to obtain a predicted score of the target object.

[0047] The initial vector of the target object output by the first neural network is input into the second neural network to generate a prediction score based on the initial vector, which is used to reflect the difficulty of AOI dispatch. The second neural network is a three-layer neural network, and the generated prediction score is the predicted value of the AOI dispatch difficulty score.

[0048] Step S104, repeatedly adjusting the parameters of the first neural network and the second neural network based on the prediction score and the feature score until a preset stop condition is met, and using the first neural network after parameter adjustment as the feature extraction network.

[0049] Based on the predicted score of the target object output by the second neural network and the obtained characteristic score of the target object, the parameters of the two neural networks are adjusted so that the predicted score output by the second neural network gradually approaches the characteristic score. According to the preset stop condition, the parameters of the first neural network and the second neural network are repeatedly adjusted until the condition is met. Specifically, if it is judged that the neural network does not meet the preset stop condition, the parameters of the two neural networks are adjusted according to the predicted score and characteristic score output by the second neural network, and the neural network after adjusting the parameters is predicted to obtain a new predicted score output by the second neural network. If it is judged that the preset stop condition is still not met, the parameters of the neural network are adjusted based on the new predicted score until the neural network meets the preset stop condition. At this time, the first neural network completes the training and is set as a feature extraction network, which can output an accurate feature vector.

[0050] Optionally, the process of inputting the feature data of the AOI into the first neural network and then inputting the vector output by the first neural network into the second neural network to obtain the predicted score is called the forward propagation of the neural network; the process of adjusting the parameters of the first neural network and the second neural network in turn according to the predicted score and the feature score is called the backward propagation. After multiple rounds of forward propagation and backward propagation, the random initial vector output by the first neural network can be adjusted to an N-dimensional vector with embedded features that can reflect the difficulty of AOI assignment.

[0051] Step S105: generating a feature vector of the target object based on the feature data of the target object and the feature extraction network.

[0052] The feature output of the target object is input into the feature extraction network (the trained first neural network) to output the feature vector of the target object.

[0053] According to the method for generating a feature vector provided by an embodiment of the present invention, it is possible to generate an initial embedding vector and a predicted score of the target object based on the neural network and the feature data of the target object, and to adjust the parameters of the neural network based on the predicted score and the acquired feature score of the target object, so that the adjusted neural network can output the feature vector of the target object, thereby realizing accurate generation of the N-dimensional feature vector of the AOI through the neural network, converting the multi-dimensional feature data of the AOI into an embedding vector, which can better characterize the characteristics of the AOI and facilitate subsequent machine learning tasks.

[0054] Optionally, adjusting the parameters of the first neural network and the second neural network based on the predicted score and the feature score includes: calculating a loss value based on the predicted score and the feature score; and adjusting the parameters of the first neural network and the second neural network based on the loss value. When adjusting the parameters of the neural network, firstly, the output predicted score can be used to calculate the loss value. The loss value is calculated by the obtained feature score y to reflect the difference between the two score data. Specifically, the MSE loss value e of the two scores is calculated using the following formula:

[0055]

[0056] Where n is the number of input samples of the neural network. This formula can measure the average of the mean square error of the prediction scores and feature scores of a batch of samples.

[0057] After obtaining the loss value e, the two neural networks are back-propagated based on the loss value e. Back-propagation will make the first neural network parameter , the value of the output embedding vector and the second neural network parameter The value of is adjusted in the direction of reducing the MSE loss value e. After multiple rounds of forward and backward propagation, that is, multiple rounds of parameter adjustment, the vector output by the first neural network is the feature vector of AOI (target object).

[0058] Optionally, when calculating the loss value, you can also choose to output the prediction score The difference between the obtained feature score y is used as the loss value, which is used for back propagation to adjust in the direction of reducing the loss value.

[0059] Optionally, the preset stop condition is that the loss value is not greater than a preset loss threshold, or the number of parameter adjustments is greater than or equal to a preset number threshold. The preset stop condition for adjusting the parameters of the neural network is that the loss value is not greater than a preset loss threshold, or the number of parameter adjustments is greater than or equal to a preset number threshold. When the neural network meets this condition, forward and backward propagation is stopped, and it is determined that the first neural network at this time is a trained feature extraction network, and its output vector is the feature vector of the target object.

[0060] Optionally, the preset stop condition can be selected according to the actual situation to suit different scenarios. In addition to the loss value not being greater than the preset loss threshold, or the number of parameter adjustments being greater than or equal to the preset number threshold, other conditions may also be that the adjustment time is not greater than the preset time threshold, and there is no restriction on this.

[0061] According to the method for generating a feature vector provided by an embodiment of the present invention, the feature data of the target object can be input into a neural network to obtain a feature vector of the target object output by the neural network. Before the neural network outputs the feature vector, the parameters of the neural network can be adjusted according to the output prediction score and the feature score of the target object, so that the feature vector output by the neural network is more accurate.

[0062] Optionally, obtaining the characteristic score of the target object includes: inputting the characteristic data of the target object into a preset scoring system to obtain the characteristic score of the target object. By inputting the characteristic data of the target object into the preset scoring system, an output characteristic score, i.e., a difficulty score of the target object, can be obtained.

[0063] Optionally, the feature score of the target object can also be obtained by calculating by calling a preset scoring rule. Specifically, after the feature data is obtained, the stored user preset scoring rule is called, and the feature score is calculated according to the scoring rule.

[0064] Optionally, the characteristic data of the target object is input into a preset scoring system to obtain the characteristic score of the target object, including: determining the scoring information of each characteristic of the target object through the preset scoring system and the characteristic data of the target object; calculating the characteristic score of the target object based on the scoring information of each characteristic of the target object. First, the scoring information of each characteristic in the target characteristic is determined through the preset scoring system, that is, the scoring value corresponding to each characteristic, or the scoring operation, for example: based on the numerical value of feature a, the corresponding scoring value is determined to be 5 points; based on the numerical value of feature b, the corresponding scoring operation is determined to be plus 1 point. The final characteristic score is calculated based on the scoring information of each characteristic of the target object.

[0065] Optionally, the scoring information of each feature of the target object is determined by a preset scoring system and the feature data of the target object, including: determining the scoring rule corresponding to the target object according to the type feature in the feature data of the target object; calculating the scoring information of each feature of the target object according to the scoring rule corresponding to the target object by the preset scoring system. When determining the score of each feature, due to the influence of the type feature in the feature data of the target object, there may be differences in the scores of different features, such as whether the type feature is a residential area. Since the difference in the number of orders in residential and non-residential areas may be large, the corresponding average number of delivered orders in residential and non-residential areas obviously cannot use the same scoring rule. Therefore, before determining the scoring information of each feature, the scoring rule corresponding to the target object is first determined according to the type feature of the target object, and then the scoring information of each feature is determined based on the scoring rule by the scoring system, and finally the score of the target object is calculated. For example, the scoring of the average weekly number of successfully delivered orders in Rule 1 is: 0-200 is 1 point, 201-400 is 2 points, and 401-600 is 3 points; the scoring of the average weekly number of successfully delivered orders in Rule 2 is: 0-100 is 1 point, 101-200 is 2 points, and 201-300 is 3 points.

[0066] Optionally, when determining the scoring rules, a comprehensive selection is made based on multiple type features, or a selection is made based on only one type feature; different scoring levels in the scoring rules may not be evenly distributed, and can be set according to actual conditions, without limitation.

[0067] Optionally, the scoring rules can also be personalized according to the feature type, value range, etc., to make them more suitable for different application scenarios and improve the accuracy of the scoring results.

[0068] According to the method for generating a feature vector provided by an embodiment of the present invention, it is possible to generate an initial embedding vector and a predicted score of a target object based on a neural network and the feature data of the target object, and determine the feature score of the target object based on a scoring system. The parameters of the neural network are adjusted in combination with the predicted score and the obtained feature score of the target object, so that the adjusted neural network can output the feature vector of the target object, and the N-dimensional feature vector of the AOI is accurately generated through the neural network, and the multi-dimensional feature data of the AOI is converted into an embedding vector, which can better characterize the characteristics of the AOI. At the same time, by freely setting the value of the dimension N, the embedding vector input by the neural network can be made more suitable for subsequent machine learning tasks, thereby improving the application effect of the AOI feature data.

[0069] like Figure 2 The figure shows a schematic diagram of the training process of the neural network provided by an embodiment of the present invention, including: obtaining AOI feature data including weekly average number of successfully delivered orders, monthly average number of successfully delivered orders, timely rate of successfully delivered orders, customer complaint rate, etc., inputting the feature data into a layer of neural network to obtain an output N-dimensional initial random vector, inputting the vector into a multi-layer neural network to obtain an output AOI pickup and dispatch difficulty prediction value; calculating the MSE loss value based on the AOI pickup and dispatch difficulty prediction value and the obtained AOI pickup and dispatch difficulty score, and performing back propagation based on the loss value to adjust the parameter values ​​of a layer of neural network and a multi-layer neural network, and stopping the back propagation when the preset stop condition is met. At this time, the N-dimensional vector output by the layer of neural network is the feature vector of the AOI.

[0070] Figure 3 is a schematic diagram of a device 300 for generating a feature vector provided by an embodiment of the present invention, such as Figure 3 As shown, the data distribution device 300 according to the embodiment of the present invention mainly includes: an acquisition module 301 , an input module 302 , an adjustment module 303 and a generation module 304 .

[0071] An acquisition module 301 is used to acquire feature data of a target object and a feature score of the target object;

[0072] An input module 302 is used to input the characteristic data of the target object into the first neural network to generate an initial vector of the target object; and input the initial vector of the target object into the second neural network to obtain a predicted score of the target object;

[0073] An adjustment module 303, configured to repeatedly adjust the parameters of the first neural network and the second neural network based on the prediction score and the feature score until a preset stop condition is met, and use the first neural network after the parameter adjustment as a feature extraction network;

[0074] The generating module 304 is used to generate a feature vector of the target object based on the feature data of the target object and the feature extraction network.

[0075] The device for generating a feature vector provided by an embodiment of the present invention can generate an initial embedding vector and a predicted score of the target object based on the neural network and the feature data of the target object, and adjust the parameters of the neural network based on the predicted score and the acquired feature score of the target object, so that the adjusted neural network can output the feature vector of the target object, thereby realizing accurate generation of the N-dimensional feature vector of the AOI through the neural network, converting the multi-dimensional feature data of the AOI into an embedding vector, which can better characterize the characteristics of the AOI and facilitate subsequent machine learning tasks.

[0076] Optionally, an adjustment module 303, used to adjust parameters of the first neural network and the second neural network based on the predicted score and the feature score, includes: a calculation module 305, used to calculate a loss value based on the predicted score and the feature score; an adjustment module 303, used to adjust parameters of the first neural network and the second neural network based on the loss value.

[0077] Optionally, the preset stop condition is that the loss value is not greater than a preset loss threshold, or the number of parameter adjustments is greater than or equal to a preset number threshold.

[0078] Optionally, the acquisition module 301 is used for scoring the characteristics of the target object, including: an input module 302, which is further used to input the characteristic data of the target object into a preset scoring system to obtain the characteristic score of the target object.

[0079] Optionally, the input module 302 is also used to input the feature data of the target object into a preset scoring system to obtain the feature score of the target object, including: a determination module 306, used to determine the scoring information of each feature of the target object through the preset scoring system and the feature data of the target object; a calculation module 305, also used to calculate the feature score of the target object based on the scoring information of each feature of the target object.

[0080] Optionally, the determination module 306 is used to determine the scoring information of each feature of the target object through a preset scoring system and the feature data of the target object, including: the determination module 306 is used to determine the scoring rule corresponding to the target object according to the type feature in the feature data of the target object; the calculation module 305 is also used to calculate the scoring information of each feature of the target object according to the scoring rule corresponding to the target object through the preset scoring system.

[0081] According to the device for generating feature vectors provided by the embodiment of the present invention, it is possible to generate an initial embedding vector and a predicted score of a target object based on a neural network and the feature data of the target object, and determine the feature score of the target object based on a scoring system. The parameters of the neural network are adjusted in combination with the predicted score and the acquired feature score of the target object, so that the adjusted neural network can output the feature vector of the target object, and the N-dimensional feature vector of the AOI is accurately generated through the neural network, and the multi-dimensional feature data of the AOI is converted into an embedding vector, which can better characterize the characteristics of the AOI. At the same time, by freely setting the value of the dimension N, the embedding vector input by the neural network can be made more suitable for subsequent machine learning tasks, thereby improving the application effect of the AOI feature data.

[0082] Figure 4 An exemplary system architecture 400 is shown to which the method for generating a feature vector or the apparatus for generating a feature vector according to the embodiment of the present invention can be applied.

[0083] like Figure 4 As shown, system architecture 400 may include terminal devices 401, 402, 403, network 404 and server 405. Network 404 is used to provide a medium for communication links between terminal devices 401, 402, 403 and server 405. Network 404 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0084] Users can use terminal devices 401, 402, 403 to interact with server 405 through network 404 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 401, 402, 403, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0085] The terminal devices 401 , 402 , and 403 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0086] Server 405 may be a server that provides various services, such as a background management server that supports websites browsed by users using terminal devices 401, 402, and 403. The background management server may obtain feature data and feature scores of the target object based on the received feature vector generation request and other data; input the feature data of the target object into the first neural network to generate an initial vector of the target object; input the initial vector of the target object into the second neural network to obtain a predicted score of the target object; repeatedly adjust the parameters of the first neural network and the second neural network based on the predicted score and the feature score until a preset stop condition is met, and use the first neural network after parameter adjustment as a feature extraction network; generate a feature vector of the target object based on the feature data of the target object and the feature extraction network, and feed back the processing results (such as feature vectors - just an example) to the terminal device.

[0087] It should be noted that the method for generating a feature vector provided in the embodiment of the present invention is generally executed by the server 405 , and accordingly, the device for generating a feature vector is generally disposed in the server 405 .

[0088] It should be understood that Figure 4 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0089] Reference below Figure 5 , which shows a schematic diagram of the structure of a computer system 500 of a terminal device suitable for implementing an embodiment of the present invention. Figure 5 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0090] like Figure 5 As shown, the computer system 500 includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage part 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the system 500 are also stored. The CPU 501, the ROM 502, and the RAM 503 are connected to each other through a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0091] The following components are connected to the I / O interface 505: an input section 506 including a keyboard, a mouse, etc.; an output section 507 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN card, a modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the I / O interface 505 as needed. A removable medium 511, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 510 as needed, so that a computer program read therefrom is installed into the storage section 508 as needed.

[0092] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 509, and / or installed from the removable medium 511. When the computer program is executed by the central processing unit (CPU) 501, the above-mentioned functions defined in the system of the present invention are executed.

[0093] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0094] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0095] The modules involved in the embodiments of the present invention may be implemented by software or hardware. The modules described may also be set in a processor, for example, it may be described as: a processor includes an acquisition module, an input module, an adjustment module and a generation module. The names of these modules do not constitute a limitation on the modules themselves in some cases, for example, the acquisition module may also be described as a "module for acquiring feature data of a target object and a feature score of a target object".

[0096] As another aspect, the present invention further provides a computer-readable medium, which may be included in the device described in the above embodiment; or may exist independently without being assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device includes:

[0097] Acquire feature data of a target object and a feature score of the target object;

[0098] Inputting feature data of the target object into the first neural network to generate an initial vector of the target object;

[0099] Inputting the initial vector of the target object into the second neural network to obtain a predicted score of the target object;

[0100] Repeating the process of adjusting the parameters of the first neural network and the second neural network based on the prediction score and the feature score until a preset stop condition is met, and using the first neural network after the parameter adjustment as a feature extraction network;

[0101] A feature vector of the target object is generated based on the feature data of the target object and a feature extraction network.

[0102] As another aspect, the present invention further provides a computer program product, including a computer program, which implements the method of the embodiment of the present invention when the computer program is executed by a processor.

[0103] According to the technical solution of the embodiment of the present invention, it is possible to generate an initial embedding vector and a predicted score of the target object based on the neural network and the feature data of the target object, and to adjust the parameters of the neural network based on the predicted score and the acquired feature score of the target object, so that the adjusted neural network can output the feature vector of the target object, thereby realizing accurate generation of the N-dimensional feature vector of the AOI through the neural network, converting the multi-dimensional feature data of the AOI into an embedding vector, which can better characterize the characteristics of the AOI and facilitate subsequent machine learning tasks.

[0104] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for generating a feature vector, characterized in that: include: Acquire feature data of a target object and a feature score of the target object; Inputting the characteristic data of the target object into a first neural network to generate an initial vector of the target object; Inputting the initial vector of the target object into a second neural network to obtain a predicted score of the target object; Repeating the process of adjusting the parameters of the first neural network and the second neural network based on the prediction score and the feature score until a preset stop condition is met, and using the first neural network after parameter adjustment as a feature extraction network; A feature vector of the target object is generated based on the feature data of the target object and the feature extraction network.

2. The method according to claim 1, characterized in that The adjusting the parameters of the first neural network and the second neural network based on the prediction score and the feature score comprises: Calculating a loss value based on the prediction score and the feature score; Parameters of the first neural network and the second neural network are adjusted based on the loss value.

3. The method according to claim 2, characterized in that The preset stop condition is that the loss value is not greater than a preset loss threshold, or the number of parameter adjustments is greater than or equal to a preset number threshold.

4. The method according to any one of claims 1 to 3, characterized in that: The obtaining of the feature score of the target object includes: The characteristic data of the target object is input into a preset scoring system to obtain the characteristic score of the target object.

5. The method according to claim 4, characterized in that The step of inputting the characteristic data of the target object into a preset scoring system to obtain the characteristic score of the target object includes: Determining scoring information of each feature of the target object by using the preset scoring system and the feature data of the target object; The feature score of the target object is calculated based on the score information of each feature of the target object.

6. The method according to claim 5, characterized in that The step of determining the scoring information of each feature of the target object by using the preset scoring system and the feature data of the target object includes: Determining a scoring rule corresponding to the target object according to the type feature in the feature data of the target object; The scoring information of each feature of the target object is calculated by the preset scoring system according to the scoring rules corresponding to the target object.

7. A device for generating a feature vector, characterized in that: include: An acquisition module, used to acquire feature data of a target object and a feature score of the target object; An input module, used to input the characteristic data of the target object into a first neural network to generate an initial vector of the target object; input the initial vector of the target object into a second neural network to obtain a predicted score of the target object; an adjustment module, configured to repeatedly adjust the parameters of the first neural network and the second neural network based on the prediction score and the feature score until a preset stop condition is met, and use the first neural network after parameter adjustment as a feature extraction network; A generating module is used to generate a feature vector of the target object based on the feature data of the target object and the feature extraction network.

8. An electronic device, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 6.

9. A computer readable medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images