Feature space management

Abstract

A method, an apparatus, and a computer program product for extending a feature space of a data sample are presented. In the method, a global representation of a feature of a plurality of features for a data sample is obtained. A local representation for the feature is obtained based on a classification criterion for classifying the data sample into one of a plurality of pre-defined domains. A representation of the feature for the data sample is generated based on the global representation and the local representation. With these implementations, a dedicated feature space dedicated to capture domain-specific knowledge and characteristics may be created for each domain identified by classification criteria.

Description

Technology Domain

[0001] The present disclosure relates generally to feature space management, and more particularly, to methods, apparatus, and computer program products for expanding a feature space of a data sample according to a domain associated with the data sample. Background Art

[0002] Currently, machine learning techniques have been widely used in data processing. In machine learning techniques, various features can be collected from data samples, and then these features can be converted into an embedding space to represent the data samples. Usually, a shared feature space is used to represent data samples from all domains, which may hinder its performance when processing different domains. At this time, it is expected to represent the features of data samples related to each domain in a more efficient way. Summary of the invention

[0003] In a first aspect of the present disclosure, a method for expanding a feature space of a data sample is provided. The method comprises: obtaining a global representation of a feature among multiple features of the data sample; obtaining a local representation of the feature based on a classification criterion for classifying the data sample into one of multiple predefined domains; and generating a representation of the feature of the data sample based on the global representation and the local representation.

[0004] In a second aspect of the present disclosure, an electronic device is provided. The electronic device comprises: a computer processor coupled to a computer readable memory unit, the memory unit comprising instructions for implementing the method according to the first aspect of the present disclosure when executed by the computer processor.

[0005] In a third aspect of the present disclosure, a computer program product is provided, which includes a computer-readable storage medium having program instructions embodied therein, and the program instructions can be executed by an electronic device to enable the electronic device to perform the method according to the first aspect of the present disclosure.

[0006] This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] The above and other objects, features and advantages of the present disclosure will become more apparent through a more detailed description of some implementations of the present disclosure in the accompanying drawings, in which like reference numerals generally refer to like parts in the implementations of the present disclosure.

[0008] Figure 1 An example environment for feature space management based on machine learning techniques is shown;

[0009] Figure 2 An example diagram for feature space management according to an implementation of the present disclosure is shown;

[0010] Figure 3 An example diagram for an expanded feature space according to an implementation of the present disclosure is shown;

[0011] Figure 4 An example diagram for initializing a local representation according to an implementation of the present disclosure is shown;

[0012] Figure 5 An example diagram for updating a local representation according to an implementation of the present disclosure is shown;

[0013] Figure 6 An example diagram for updating a local representation according to an implementation of the present disclosure is shown;

[0014] Figure 7 An example diagram for determining a sample representation for a data sample according to an implementation of the present disclosure is shown;

[0015] Figure 8 An example diagram for obtaining output from a machine learning model based on sample representations according to an implementation of the present disclosure is shown;

[0016] Fig. 9 An example flow chart showing a method for feature space management according to an implementation of the present disclosure; and

[0017] Fig.10 A block diagram of a computing device is shown in which various implementations of the present disclosure may be implemented. DETAILED DESCRIPTION

[0018] The principles of the present disclosure will now be described with reference to some implementations. It should be understood that these implementations are described only for illustrative purposes and are helpful for those skilled in the art to understand and implement the present disclosure without implying any limitation on the scope of the present disclosure. The disclosure described herein can be implemented in various ways different from the ways described below.

[0019] In the following description and claims, unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.

[0020] References in this disclosure to "one implementation," "implementation," "example implementation," etc. indicate that the implementation being described may include a particular feature, structure, or characteristic, but not every implementation necessarily includes the particular feature, structure, or characteristic. Furthermore, such phrases do not necessarily refer to the same implementation. Furthermore, when a particular feature, structure, or characteristic is described in conjunction with an example implementation, it is considered to be within the knowledge of one skilled in the art to implement such feature, structure, or characteristic in conjunction with other implementations, whether or not explicitly described.

[0021] It should be understood that although the terms "first" and "second" etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, without departing from the scope of the example implementation, the first element may be referred to as the second element, and similarly, the second element may be referred to as the first element. As used herein, the term "and / or" includes any and all combinations of one or more of the listed items.

[0022] The terms used herein are only used for the purpose of describing specific implementations and are not intended to limit example implementations. As used herein, the singular forms "a", "an" and "the" are also intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the terms "including", "having" and / or "comprising" when used herein specify the presence of the features, elements and / or components, etc., but do not exclude the presence or addition of one or more other features, elements, components and / or combinations thereof.

[0023] The principles of the present disclosure will now be described with reference to some implementations. It should be understood that these implementations are described for illustrative purposes only and are helpful for those skilled in the art to understand and implement the present disclosure without implying any limitation on the scope of the present disclosure. The disclosure described herein can be implemented in various ways different from the ways described below. In the following description and claims, unless otherwise defined, all technical and scientific terms used herein have the same meanings as those commonly understood by those of ordinary skill in the field to which the present disclosure belongs.

[0024] It is understandable that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) should comply with the requirements of relevant laws and regulations and relevant rules.

[0025] It is understandable that before using the technical solutions disclosed in various implementations of the present disclosure, the type, scope of use and usage scenarios of the personal information involved in the present disclosure should be notified to the user in an appropriate manner in accordance with relevant laws and regulations, and the user's authorization should be obtained.

[0026] For example, in response to receiving an active request from a user, a prompt message is sent to the user to clearly inform the user that the requested operation will require obtaining and using the user's personal information. Therefore, the user can independently choose whether to provide personal information to software or hardware such as an electronic device, application, server, or storage medium that performs the operation of the technical solution of the present disclosure according to the prompt message.

[0027] As an optional but non-limiting implementation, in response to receiving an active request from the user, the method of sending the prompt information to the user may include, for example, a pop-up window, and the prompt information may be presented in the form of text in the pop-up window. In addition, the pop-up window may also carry a selection control for the user to select "agree" or "disagree" to provide personal information to the electronic device.

[0028] It is understandable that the above notification and the process of obtaining user authorization are only illustrative and do not limit the implementation of the present disclosure. Other methods that meet relevant laws and regulations are also applicable to the implementation of the present disclosure.

[0029] At present, machine learning technology has been widely used in data processing. In machine learning technology, various features can be collected from data samples, and then these features can be converted into an embedding space to represent the data samples. Figure 1 A brief description of feature space management is given here Figure 1 An example environment 100 for feature space management based on machine learning techniques is shown. Figure 1 As shown, the data sample 110 may include a plurality of features 112, 114, etc. Here, the data sample 110 may be a data entry to be processed by a machine learning model, and the data sample 110 may have various features depending on the purpose of the machine learning model.

[0030] For the purpose of description, the recommendation model can be used as an example machine learning model. For example, the recommendation model can receive user data and video data, and then output the probability level of whether the user is interested in the video. At this time, the features of the data sample can include user-related features (e.g., user ID, user's device type, device OS type, etc.) and video-related features (e.g., video ID, video resolution, video length, etc.). Here, each feature can be indicated by a corresponding representation, for example, feature 112 can be indicated by representation 120, and feature 114 can be indicated by representation 122, and so on.

[0031] Typically, representations have a feature space with a fixed dimension size (e.g., 16 bits or another size) to represent data samples from all domains, which may hinder its performance when processing different domains. Using a single feature space shared by all domains, machine learning models cannot effectively capture domain-specific knowledge. For example, in the case where a video includes shopping and gaming data, user behaviors and preferences can differ significantly from each other. A single feature space fails to adequately distinguish and represent these different user behaviors and preferences, which results in suboptimal distinctions for shopping videos and gaming videos. At this point, it is desirable to represent the features of data samples associated with each domain in a more efficient manner.

[0032] In view of this, the present disclosure proposes a feature space management solution based on the domain of data samples. Generally, data samples can be classified into specific domains according to predefined classification criteria. Then, local representations are obtained based on the specific domains, and then the local representations are used when generating representations for the data samples. In other words, domain information is added to the extended feature space for the data samples.

[0033] Using these implementations of the present disclosure, for each domain identified from the classification criteria, a dedicated feature space can be created that is dedicated to capturing domain-specific knowledge and characteristics. By expanding the feature space, the machine learning model can tailor its predictions to accommodate the unique preferences and behaviors exhibited in different domains.

[0034] refer to Figure 2 A brief description of the present disclosure is provided here Figure 2 An example diagram 200 for feature space management according to an implementation of the present disclosure is shown. Figure 2 As shown on the left side of , classification criteria 210 may be predefined to classify data samples into a plurality of predetermined domains. For example, classification criteria 210 may relate to domain 212, domain 214, and the like. For example, classification criteria 210 may relate to the type of device of the user, in which case the predetermined domains may include: a personal computer (PC) domain and a mobile domain. Alternatively and / or additionally, classification criteria 210 may relate to the OS type of the user's device, in which case the predetermined domains may include: an IOS domain and an Android domain. Alternatively and / or additionally, classification criteria 210 may relate to video resolution, in which case the predetermined domains may include: a high-resolution domain, a medium-resolution domain, and a low-resolution domain.

[0035] refer to Figure 2On the right side of , the representation 220 can be divided into multiple fields. For example, the representation 220 can include a global representation 240 and a local representation 230. Here, the global representation 240 can be determined based on an existing solution. Here, the local representation 230 can be further divided into multiple fields. For example, it can have multiple parts corresponding to multiple domains, respectively. Here, part 232 can correspond to domain 212, and part 234 can correspond to domain 214, and so on. At this point, the conventional feature space is expanded to further include the local representation 230. Using these implementations of the present disclosure, the corresponding features in the data sample 110 can be expressed as follows: Figure 2 The new structure shown is used to represent it.

[0036] In implementations of the present disclosure, using the extended feature space, a global representation 230 of a feature among multiple features of a data sample can be obtained, and a local representation 240 for the feature can be obtained based on a classification criterion 210 for classifying the data sample into one of a plurality of predetermined domains 212, 214, etc. Then, a representation 220 of the feature of the data sample can be generated based on the global representation 230 and the local representation 240. Using these implementations, the global representation 240 can be used as a feature space shared by all domains, which is designed to capture common patterns and knowledge shared across all domains. In addition, the local representation 230 can be used as a feature space dedicated to its own domain, which is designed to capture specific patterns and knowledge specific to a particular domain. By learning from collective knowledge, the machine learning model can gain a broader understanding of general user preferences and video content relevance.

[0037] Having provided a general description of the present disclosure, the following paragraphs will provide more details on feature space management.In implementations of the present disclosure, each field in representation 220 may have its own size. Figure 3 An example diagram 300 of an expanded feature space according to an implementation of the present disclosure is shown. Figure 3 As shown, representation 220 may have dimension size 310. With respect to two fields in representation 220, local representation 230 may have local dimension size 320, and global representation 240 may have global dimension size 330. Additionally, with respect to multiple fields in local representation 230, portion 232 may have portion size 340, and portion 234 may have portion size 342, and so on.

[0038] In implementations of the present disclosure, the local dimension size 320 may indicate the influence of the domain information on the machine learning model. Therefore, the larger the local dimension size 320, the more the influence of the domain information is considered in the machine learning model. At this time, the local dimension size 320 may be determined based on the importance of the domain information. Using these implementations of the present disclosure, the local dimension size 320 may be determined in a more efficient manner for extracting more knowledge within various domains. Since the global representation 240 involves all domains and the local representation involves a specific domain, the local dimension size 320 may generally be smaller than the global dimension size 330. For example, the local dimension size 320 may be set to half (or another ratio) of the global dimension size 330.

[0039] In implementations of the present disclosure, a global dimension size 330 for a global representation 330 may be determined based on a predetermined dimension size for the representation 220. In implementations of the present disclosure, a global dimension size 330 may be determined based on a predetermined dimension size. Assuming that the predetermined dimension size is 16 bits, the global representation 240 may have a length of 16 bits. In addition, a local representation 230 may be added, and the local dimension size 320 may be determined based on the above process. For example, if the domain is important, the local dimension size 320 may be set to 8 bits (or another value determined based on the importance), and then the dimension size 310 may be set to: 16+8=24 bits. With these implementations of the present disclosure, the representation 220 is extended by a local representation 230 having 8 bits. Therefore, more bits are used to represent features, and therefore the performance of the machine learning model may be increased in terms of capturing specific patterns and knowledge for a specific domain.

[0040] In implementations of the present disclosure, the local representation size 320 may be determined based on the difference between the predetermined dimension size and the local dimension size 320. For example, if the local dimension size 320 is set to 8 bits and the dimension size 310 is 16 bits, the global dimension size 330 may be set to: 16-8=8 bits. With these implementations of the present disclosure, all 16 bits are divided into the local representation 230 (8 bits) and the global representation 240 (8 bits). At this point, more knowledge for a specific domain is obtained without increasing the dimension size 310 of the representation 220.

[0041] In an implementation of the present disclosure, the local representation may include multiple parts corresponding to multiple predetermined domains, respectively, and the sum of multiple dimension sizes for the multiple parts may be equal to the local dimension size 320. In other words, multiple parts share the local dimension size 320. In an implementation of the present disclosure, each part may have the same size. At this time, the dimension size for the first part of the multiple parts may be equal to the dimension size for the second part of the multiple parts. In other words, the local dimension size 320 is evenly divided into multiple parts. Continuing with the above example, if there are two domains and the local dimension size 320 is 8 bits, the part size 340 for part 232 and the part size 342 for part 234 can be equivalently set to 8 / 2=4 bits. With these implementations, the dimension sizes for various fields can be determined in an easy and efficient manner.

[0042] In addition, the number of data samples belonging to a specific domain also affects the knowledge distribution between the data samples. For example, the larger the number of samples involved in a specific domain, the more knowledge can be extracted from the specific domain. At this time, the local dimension size 320 can also be determined based on the number of data samples belonging to the domain. Specifically, if there are more data samples belonging to the first domain than to the second domain, the portion size 340 can be greater than the portion size 342. For example, 5 bits from 8 bits can be allocated to portion 232, and 8-5=3 bits can be allocated to portion 234.

[0043] In implementations of the present disclosure, once the dimension sizes 340 and 342 are determined, these portions 232, 234, etc. within the local representation 230 may be initialized. Figure 4 More details about the initialization phase are described here. Figure 4 An example diagram 400 for initializing a local representation according to an implementation of the present disclosure is shown. Figure 4 As shown, an initialization vector having a length corresponding to the size of the portion can be loaded into each portion. Continuing the above example, where portions 232 and 234 have the same size of 4 bits, the initialization vector can be represented as having 4 bits (0000). Alternatively and / or additionally, if the portions have different lengths, the initialization vector can have another structure.

[0044] In an implementation of the present disclosure, the local representation may be determined based on the domain of the data sample. Here, the domain of the data sample may be determined based on the classification criteria, and then the local representation 230 may be obtained based on the domain of the data sample. Specifically, if it is determined that the domain of the data sample belongs to a target domain among multiple predetermined domains, a target portion corresponding to the target domain among the multiple portions may be updated. Figure 5 More details about the update process are described here. Figure 5An example graph 500 for updating a local representation according to an implementation of the present disclosure is shown.

[0045] like Figure 5 As shown, the domain of the data sample 510 can be determined according to the classification criteria. Assuming that the classification criteria involves the user's device type, the predetermined domains include: PC domain and mobile domain. At this time, part 232 can indicate the PC domain, and part 234 can indicate the mobile domain. Based on the classification criteria, if the domain of the data sample 510 belongs to domain 212 (also referred to as the first domain in multiple predetermined domains), part 232 (i.e., the first part in the local representation 230) can be updated according to the domain (as shown by arrow 520); at the same time, the initialization vector in part 234 (i.e., the second part in the local representation 230) remains unchanged. In addition, the value in the global representation 240 (as shown by arrow 530) can be determined according to the existing solution.

[0046] Figure 6 An example diagram 600 for updating a local representation according to an implementation of the present disclosure is shown. Based on the classification criteria, if the domain of the data sample 610 belongs to the domain 214 (also referred to as the second domain of the plurality of predetermined domains), the portion 234 (i.e., the second portion in the local representation 230) may be updated according to the domain (as indicated by arrow 620); meanwhile, the initialization vector in the portion 232 (i.e., the first portion in the local representation 230) remains unchanged. In addition, the value in the global representation 240 (as indicated by arrow 630) may be determined according to existing solutions.

[0047] Although the above paragraphs have provided details for determining the local representation by way of example, in this example the local representation includes two parts corresponding to two domains, respectively. Alternatively and / or additionally, the local representation may include more parts for more domains. For example, if the classification criteria involve video resolution, the three domains may include: a high-resolution domain, a medium-resolution domain, and a low-resolution domain. At this point, the local representation may include three parts corresponding to the three domains, respectively. In addition, the dimensional sizes of the three parts may be determined based on the ratio of the number of data samples associated with the three domains, respectively.

[0048] Specifically, if there are 200 videos of high resolution, 100 videos of medium resolution, and 100 videos of low resolution, the dimensional size of these three parts can be set to 2:1:1. Assuming that the total dimensional size of the local representation is 12 bits, the three parts can be set to 6 bits, 3 bits, and 3 bits, respectively. Alternatively and / or additionally, the three parts can be equally set to 4 bits, 4 bits, and 4 bits.

[0049] In an implementation of the present disclosure, local representation 230 and global representation 240 may be joined to form representation 220. With the joined structure, representation 220 may reflect knowledge related to all domains as well as specific domains. Thus, a machine learning model may tailor its predictions to accommodate unique knowledge presenting different domains.

[0050] Although the above paragraphs have provided the determination of a representation for a feature. One or more features of the data sample can be processed according to the above scheme. For example, another representation for another feature of the multiple features of the data sample can be generated. Then, a sample representation for the data sample can be generated based on the representation for the feature and the another representation for the other feature. For more details, refer to Figure 7 ,here, Figure 7 An example diagram 700 for determining a sample representation for a data sample is shown in accordance with an implementation of the present disclosure.

[0051] like Figure 7 As shown, data sample 110 includes feature 112, feature 114, etc. A representation 710 for feature 112 can be generated according to the above process, wherein representation 710 can include a local representation and a global representation. Similarly, a representation 720 for feature 114 can be generated according to the above process. Next, representations 710 and 720 can be joined to form a sample representation 730. With these implementations, sample representation 730 can reflect more domain information related to the data sample, and thus can help improve the performance of the machine learning model.

[0052] In an implementation of the present disclosure, the sample representation 730 may be input into a machine learning model for further processing. For example, a task model may represent an association relationship between a first feature and a second feature of a reference sample representation of a reference data sample. At this point, the sample representation 730 may be input into the task model to obtain an association relationship between a feature and another feature.

[0053] Figure 8 An example diagram 800 for obtaining output from a machine learning model based on a sample representation according to an implementation of the present disclosure is shown. Continuing with the example of the recommendation model 810 described above, the recommendation model 810 can be trained and describes the association relationship between user features and video features, for example, the association relationship can indicate the probability level of whether the user is interested in the video. At this point, the sample representation 730 can be input into the recommendation model 810, and an output 820 can be received.

[0054] With these implementations of the present disclosure, a sample representation 730 including a dedicated feature space dedicated to capturing domain-specific knowledge and characteristics can be created for each domain identified by the classification criteria. At this point, the recommendation model 810 can learn more about the domain information, and thus the performance of the recommendation model 810 can be increased.

[0055] The above paragraphs have described the details of feature space management. According to the implementation of the present disclosure, a method for expanding the feature space of a data sample is provided. Fig. 9 More details about this method are described, among which, Fig. 9 An example flow chart of a method 900 for expanding a feature space according to an implementation of the present disclosure is shown. At block 910, a global representation of a feature in a plurality of features for a data sample is obtained. At block 920, a local representation for the feature is obtained based on a classification criterion for classifying the data sample into one of a plurality of predetermined domains. At block 930, a representation of the feature for the data sample is generated based on the global representation and the local representation.

[0056] In an implementation of the present disclosure, obtaining a local representation for a feature includes: determining a domain of a data sample based on a classification criterion; and obtaining the local representation based on the domain of the data sample.

[0057] In an implementation of the present disclosure, a global dimension size of a global representation for a feature is determined based on a predetermined dimension size of a representation for the feature.

[0058] In an implementation of the present disclosure, the global dimension size is determined based on any of: a difference between a predetermined dimension size and a local dimension size for a local representation of a feature; or a predetermined dimension size.

[0059] In an implementation of the present disclosure, the local representation includes a plurality of parts respectively corresponding to a plurality of predetermined domains, and a sum of a plurality of dimensional sizes of the plurality of parts is equal to the local dimensional size.

[0060] In an implementation of the present disclosure, the dimension size of a part among the plurality of parts is determined based on any one of the following: the importance of a domain corresponding to the part, or the number of data samples belonging to the domain corresponding to the part.

[0061] In an implementation of the present disclosure, a dimension size for a first portion of the plurality of portions is equal to a dimension size for a second portion of the plurality of portions.

[0062] In an implementation of the present disclosure, obtaining a local representation includes: respectively initializing multiple parts in the local representation; and in response to determining that the domain of the data sample belongs to a target domain in multiple predetermined domains, updating a target part in the multiple parts corresponding to the target domain.

[0063] In an implementation of the present disclosure, the method further includes: generating another representation for another feature of the plurality of features of the data sample; and generating a sample representation for the data sample based on the representation for the feature and another representation for another feature.

[0064] In an implementation of the present disclosure, the method further includes: obtaining an association relationship between a feature and another feature according to a task model and a sample representation, wherein the task model represents an association relationship between a first feature and a second feature of a reference sample representation of a reference data sample.

[0065] According to an implementation of the present disclosure, a device for expanding a feature space of a data sample is provided. The device includes: a global obtaining unit configured to obtain a global representation of a feature among multiple features of a data sample; a local unit configured to obtain a local representation of the feature based on a classification standard for classifying the data sample into one of multiple predetermined domains; and a generating unit configured to generate a representation of the feature of the data sample based on the global representation and the local representation. In addition, the device may include other units for implementing other steps in method 800.

[0066] According to an implementation of the present disclosure, an electronic device for implementing method 800 is provided. The electronic device includes: a computer processor, the computer processor is coupled to a computer-readable memory unit, the memory unit includes instructions, and the instructions implement a method for expanding the feature space of a data sample when executed by the computer processor. The method includes: obtaining a global representation of a feature among a plurality of features of the data sample; obtaining a local representation of the feature based on a classification criterion for classifying the data sample into one of a plurality of predetermined domains; and generating a representation of the feature of the data sample based on the global representation and the local representation.

[0067] In an implementation of the present disclosure, obtaining a local representation for a feature includes: determining a domain of a data sample based on a classification criterion; and obtaining the local representation based on the domain of the data sample.

[0068] In an implementation of the present disclosure, a global dimension size of a global representation for a feature is determined based on a predetermined dimension size of a representation for the feature.

[0069] In an implementation of the present disclosure, the global dimension size is determined based on any of: a difference between a predetermined dimension size and a local dimension size for a local representation of a feature; or a predetermined dimension size.

[0070] In an implementation of the present disclosure, the local representation includes a plurality of parts respectively corresponding to a plurality of predetermined domains, and a sum of a plurality of dimensional sizes of the plurality of parts is equal to the local dimensional size.

[0071] In an implementation of the present disclosure, the dimension size of a part among the plurality of parts is determined based on any one of the following: the importance of a domain corresponding to the part, or the number of data samples belonging to the domain corresponding to the part.

[0072] In an implementation of the present disclosure, a dimension size for a first portion of the plurality of portions is equal to a dimension size for a second portion of the plurality of portions.

[0073] In an implementation of the present disclosure, obtaining a local representation includes: respectively initializing multiple parts in the local representation; and in response to determining that the domain of the data sample belongs to a target domain in multiple predetermined domains, updating a target part in the multiple parts corresponding to the target domain.

[0074] In an implementation of the present disclosure, the method further includes: generating another representation for another feature of the plurality of features of the data sample; and generating a sample representation for the data sample based on the representation for the feature and another representation for another feature.

[0075] In an implementation of the present disclosure, the method further includes: obtaining an association relationship between a feature and another feature according to a task model and a sample representation, wherein the task model represents an association relationship between a first feature and a second feature of a reference sample representation of a reference data sample.

[0076] According to an implementation of the present disclosure, a computer program product includes a computer-readable storage medium having program instructions embodied therewith, and the program instructions can be executed by an electronic device to enable the electronic device to perform method 900.

[0077] Fig.10 900 in which various implementations of the present disclosure may be implemented. It should be understood that Fig.10 The computing device 1000 shown in the figure is for illustration purposes only and does not in any way imply any limitation on the functionality and scope of the present disclosure. The computing device 1000 can be used to implement the above method 900 in the implementation of the present application. Fig.10 As shown, computing device 1000 may be a general computing device. Computing device 1000 may include at least one or more processors or processing units 1010, memory 1020, storage unit 1030, one or more communication units 1040, one or more input devices 1050, and one or more output devices 1060.

[0078] Processing unit 1010 may be a physical or virtual processor and may implement various processes based on programs stored in memory 1020. In a multi-processor system, multiple processing units execute computer-executable instructions in parallel to increase the parallel processing capabilities of computing device 1000. Processing unit 1010 may also be referred to as a central processing unit (CPU), a microprocessor, a controller, or a microcontroller.

[0079] The computing device 1000 typically includes various computer storage media. Such media can be any media accessible to the computing device 1000, including but not limited to volatile and non-volatile media, or removable and inseparable media. The memory 1020 can be a volatile memory (e.g., a register, a cache, a random access memory (RAM)), a non-volatile memory (such as a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM) or flash memory), or any combination thereof. The storage unit 1030 can be any removable or inseparable medium, and can include a machine-readable medium, such as a memory, a flash drive, a disk, or another other medium that can be used to store information and / or data and can be accessed in the computing device 1000.

[0080] The computing device 1000 may also include additional removable / inseparable volatile / nonvolatile storage media. Fig.10 Not shown, but a disk drive for reading from and / or writing to a removable and nonvolatile disk and an optical drive for reading from and / or writing to a removable nonvolatile optical disk may be provided. In this case, each drive may be connected to a bus (not shown) via one or more data medium interfaces.

[0081] The communication unit 1040 communicates with another computing device via a communication medium. In addition, the functionality of the components in the computing device 1000 can be implemented by a single computing cluster or multiple computing machines that can communicate via a communication connection. Therefore, the computing device 1000 can operate in a networked environment using logical connections to one or more other servers, networked personal computers (PCs), or other general network nodes.

[0082] The input device 1050 may be one or more of various input devices, such as a mouse, keyboard, trackball, voice input device, etc. The output device 1060 may be one or more of various output devices, such as a display, speaker, printer, etc. With the aid of the communication unit 1040, the computing device 1000 may also communicate with one or more external devices (not shown) (such as storage devices and display devices), wherein one or more devices enable a user to interact with the computing device 1000 or any device (such as a network card, modem, etc.) that enables the computing device 1000 to communicate 1070, 1080 with one or more other computing devices (if necessary). Such communication may be performed via an input / output (I / O) interface (not shown).

[0083] In some implementations, some or all components of the computing device 1000 may also be arranged in a cloud computing architecture rather than being integrated in a single device. In a cloud computing architecture, components may be provided remotely and work together to implement the functions described in the present disclosure. In some implementations, cloud computing provides computing, software, data access, and storage services, which will not require the end user to know the physical location or configuration of the system or hardware that provides these services. In various implementations, cloud computing provides services via a wide area network (such as the Internet) using a suitable protocol. For example, a cloud computing provider provides an application via a wide area network that can be accessed through a web browser or any other computing component. The software or components of the cloud computing architecture and the corresponding data may be stored on a server at a remote location. The computing resources in a cloud computing environment may be merged or distributed at a location in a remote data center. Cloud computing infrastructure may provide services through a shared data center, although they appear as a single access point for the user. Therefore, the cloud computing architecture may be used to provide the components and functions described herein from a service provider at a remote location. Alternatively, they may be provided from a conventional server or directly installed or otherwise installed on a client device.

[0084] The functions described herein may be performed at least in part by one or more hardware logic components. For example, but not limited to, illustrative types of hardware logic components that may be used include field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), etc.

[0085] The program code for executing the method of the subject matter described herein can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device so that the program code enables the functions / operations specified in the flow chart and / or block diagram to be implemented when executed by the processor or controller. The program code can be executed in whole or in part on the machine, in part on a remote machine as a stand-alone software package, or in whole on a remote machine or server.

[0086] In the context of the present disclosure, a machine-readable medium may be any tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or apparatus. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media would include 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 foregoing.

[0087] In addition, although operations are shown in a particular order, this should not be understood as requiring such operations to be performed in the particular order shown or in a sequential order, or performing all the operations shown to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the above discussion, these details should not be interpreted as limitations on the scope of the subject matter described herein, but should be interpreted as descriptions of features that may be specific to a particular implementation. Certain features described in the context of a separate implementation may also be implemented in combination in a single implementation. On the contrary, the various features described in a single implementation may also be implemented in multiple implementations individually or in any suitable subcombination.

[0088] Although the subject matter has been described in language specific to structural features and / or methodological acts, it should be understood that the subject matter specified in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as example forms of implementing the claims.

[0089] It can be understood from the foregoing that the specific implementation of the presently disclosed technology has been described herein for illustrative purposes, but various modifications may be made without departing from the scope of the present disclosure. Therefore, the technology of the present disclosure is not limited to the appended claims.

[0090] The implementation of the subject matter and functional operations described in the present disclosure can be implemented in various systems, digital electronic circuits or computer software, firmware or hardware, including the structures disclosed in this specification and their structural equivalents, or a combination of one or more of them. The implementation of the subject matter described in this specification can be implemented as one or more computer program products, that is, one or more modules of computer program instructions encoded on a tangible and non-transitory computer-readable medium for a data processing device to execute or control the operation of the data processing device. The computer-readable medium can be a machine-readable storage device, a machine-readable storage substrate, a memory device, a material composition that implements a machine-readable propagation signal, or a combination of one or more of them. The term "data processing unit" or "data processing device" includes all devices, equipment and machines for processing data, such as a programmable processor, a computer or multiple processors or computers. In addition to hardware, the device may also include code to create an execution environment for the computer program in question, for example, code constituting a processor firmware, a protocol stack, a database management system, an operating system, or a combination of one or more of them.

[0091] A computer program (also referred to as a program, software, software application, script, or code) may be written in any form of programming language, including compiled or interpreted languages, and it may be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. A computer program does not necessarily correspond to a file in a file system. A program may be stored in a portion of a file that holds other programs or data (e.g., one or more scripts stored in a markup language document), in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files storing one or more modules, subroutines, or code portions). A computer program may be deployed to execute on one computer or on multiple computers located at one site or distributed across multiple sites and interconnected by a communications network.

[0092] Processors suitable for executing computer programs include, for example, general-purpose and special-purpose microprocessors and any one or more processors of any kind of digital computer. Typically, the processor will receive instructions and data from a read-only memory or a random access memory or both. The basic elements of a computer are a processor for executing instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include or be operably coupled to receive data from one or more mass storage devices, or to transfer data to one or more mass storage devices, or both, one or more mass storage devices for storing data, such as magnetic, magneto-optical disks or optical disks. However, a computer need not have such a device. Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media and memory devices, including, for example, semiconductor memory devices, such as EPROM, EEPROM and flash memory devices. The processor and memory can be supplemented by or incorporated into a dedicated logic circuit.

[0093] The specification and drawings are intended to be considered exemplary only, in which the exemplary devices are examples.As used herein, the use of "or" is intended to include "and / or" unless the context clearly dictates otherwise.

[0094] Although the present disclosure contains many details, these details should not be interpreted as limitations on the scope of any disclosure or claimed content, but should be interpreted as descriptions of features that may be specific to a particular implementation of a particular disclosure. Certain features described in the present disclosure in the context of separate implementations may also be implemented in combination in a single implementation. Conversely, various features described in the context of a single implementation may also be implemented in multiple implementations individually or in any suitable sub-combination. In addition, although features may be described above as working in certain combinations and even initially claimed as such, in some cases, one or more features from the claimed combination may be cut out from the combination, and the claimed combination may be directed to a sub-combination or a variation of the sub-combination.

[0095] Similarly, although operations are shown in a particular order in the accompanying drawings, this should not be understood as requiring that such operations be performed in the particular order shown or in a sequential order, or that all of the illustrated operations be performed to achieve the desired results. In addition, the separation of various system components in the implementations described in this disclosure should not be understood as requiring such separation in all implementations. Only some implementations and examples are described, and other implementations, enhancements, and variations may be made based on what is described and illustrated in this disclosure.

Claims

1. A method for expanding a feature space of a data sample, comprising: Obtaining a global representation of a feature among a plurality of features for the data sample; obtaining a local representation for the feature based on a classification criterion for classifying the data sample into one of a plurality of predefined domains; as well as A representation of the feature for the data sample is generated based on the global representation and the local representation.

2. The method of claim 1 , wherein obtaining the local representation for the feature comprises: determining a domain of the data sample based on the classification criteria; as well as Based on the domain of the data samples, the local representation is obtained. 3 . The method of claim 1 , wherein a global dimension size of the global representation for the feature is determined based on a predetermined dimension size of the representation for the feature.

4. The method of claim 3, wherein the global dimension size is determined based on any of the following: a difference between the predetermined dimension size and the local dimension size of the local representation of the feature; or The predetermined dimension size. 5 . The method according to claim 3 , wherein the local representation comprises a plurality of parts respectively corresponding to the plurality of predefined domains, and a sum of a plurality of dimensional sizes of the plurality of parts is equal to the local dimensional size.

6. The method of claim 5, wherein the size of a dimension for a portion of the plurality of portions is determined based on any one of the following: the importance of the domain corresponding to the part, or The number of data samples belonging to the domain corresponding to the portion.

7. The method of claim 6, wherein a dimension size for a first portion of the plurality of portions is equal to a dimension size for a second portion of the plurality of portions.

8. The method of claim 5, wherein obtaining the local representation comprises: Initializing the plurality of parts in the local representation respectively; as well as In response to determining that the domain of the data sample belongs to a target domain among the plurality of predetermined domains, a target portion among the plurality of portions corresponding to the target domain is updated.

9. The method according to claim 1, further comprising: generating another representation of another feature of the plurality of features for the data sample; as well as A sample representation for the data sample is generated based on the representation for the feature and the another representation for the another feature.

10. The method according to claim 9, further comprising: An association relationship between the feature and the other feature is obtained according to a task model and the sample representation, wherein the task model represents an association relationship between a first feature and a second feature of a reference sample representation of a reference data sample.

11. An electronic device comprising a computer processor coupled to a computer readable memory unit, the memory unit comprising instructions which, when executed by the computer processor, implement a method for expanding a feature space of a data sample, the method comprising: Obtaining a global representation of a feature among a plurality of features for the data sample; obtaining a local representation for the feature based on a classification criterion for classifying the data sample into one of a plurality of predefined domains; as well as A representation of the feature for the data sample is generated based on the global representation and the local representation.

12. The apparatus of claim 11, wherein obtaining the local representation for the feature comprises: determining a domain of the data sample based on the classification criteria; as well as Based on the domain of the data samples, the local representation is obtained.

13. The apparatus of claim 11, wherein a global dimension size of the global representation for the feature is determined based on a predetermined dimension size of the representation for the feature.

14. The apparatus of claim 13, wherein the global dimension size is determined based on any of: a difference between the predetermined dimension size and the local dimension size of the local representation of the feature; or The predetermined dimension size.

15. The apparatus according to claim 13, wherein the local representation comprises a plurality of parts respectively corresponding to the plurality of predefined domains, and a sum of a plurality of dimensional sizes for the plurality of parts is equal to the local dimensional size.

16. The apparatus of claim 15, wherein the size of a dimension for a portion of the plurality of portions is determined based on any one of: the importance of the domain corresponding to the part, or The number of data samples belonging to the domain corresponding to the portion.

17. The apparatus of claim 15, wherein obtaining the local representation comprises: Initializing the plurality of parts in the local representation respectively; as well as In response to determining that the domain of the data sample belongs to a target domain among the plurality of predetermined domains, a target portion among the plurality of portions corresponding to the target domain is updated.

18. The apparatus of claim 11, wherein the method further comprises: generating another representation of another feature of the plurality of features for the data sample; as well as A sample representation for the data sample is generated based on the representation for the feature and the another representation for the another feature.

19. The apparatus of claim 18, wherein the method further comprises: An association relationship between the feature and the other feature is obtained according to a task model and the sample representation, wherein the task model represents an association relationship between a first feature and a second feature of a reference sample representation of a reference data sample.

20. A non-transitory computer program product, the non-transitory computer program product comprising a computer-readable storage medium having program instructions embodied therewith, the program instructions being executable by an electronic device to cause the electronic device to perform a method for expanding a feature space of a data sample, the method comprising: Obtaining a global representation of a feature among a plurality of features for the data sample; obtaining a local representation for the feature based on a classification criterion for classifying the data sample into one of a plurality of predefined domains; as well as A representation of the feature for the data sample is generated based on the global representation and the local representation.

Images