Time domain astronomical transient source identification method and device based on intelligent agent

Abstract

The invention provides a time domain astronomical transient source identification method and device based on an agent, relates to the field of agent data processing, and is suitable for astronomical information processing scenes based on a large model. The method comprises the following steps: inputting observation data into an intelligent agent, and obtaining a temporary source candidate body; the intelligent agent obtains an identification guidance strategy of the temporary source candidate from a preset temporary source knowledge base, determines identification steps according to the identification guidance strategy, and calls a corresponding identification tool in each step to obtain and process multiband data of the temporary source candidate; the intelligent agent analyzes the processing result of each identification tool to obtain an identification conclusion; the identification conclusion comprises a temporary source candidate body type, a reasoning process for determining the temporary source candidate body type and a marked key node; the reasoning process comprises feature data of each identification step. Therefore, the transient source analysis threshold is reduced, the credibility of the identification result is judged according to the reasoning process of each identification step, the error reasoning process is adjusted, the accuracy and the credibility are improved, and interaction with the user is realized.

Description

Technical Field

[0001] The present invention relates to the field of intelligent agent data processing, and is applicable to astronomical information processing scenarios based on large models. More specifically, it relates to a method, device, equipment, medium, and program product for identifying transient astronomical sources in the time domain based on intelligent agents. Background Technique

[0002] The Tian Guan satellite (Einstein Probe, hereinafter referred to as EP) is a representative time-domain astronomy project in China. Its main goal is to conduct rapid time-domain sky surveys of soft X-rays, aiming to discover and explore various high-energy transient and eruptive celestial bodies in the universe and monitor the X-ray variations of various celestial bodies. Identifying transient sources is one of the core steps in EP's data processing and analysis. The transient source identification process includes three parts, namely detecting transient source signals from the data obtained by EP, discriminating and classifying the detected transient source signals based on EP's observation data and reference data from different sources, and conducting follow-up observations on high-value transient sources and further judging the nature of high-value transient sources according to the observation results.

[0003] In the identification of transient sources in the EP satellite and its experimental satellite LEIA (Lobster Eye Imager for Astronomy), the EP Science Center has set up two positions, namely the Transient Advocate (TA) and the Duty Scientist (DS), to complete the identification of transient sources through the TA and the DS. The TA is responsible for checking the X-ray data downloaded by the satellite every day, analyzing the transient source candidates among them, and giving preliminary identification results. Then, the DS makes decisions on the transient source identification results and arranges follow-up observations for high-value transient sources. The most labor-intensive step in this process is for the TA to analyze the transient source candidates. For a high-confidence transient source candidate, the TA needs to call different tools and reference data for analysis. Although TDAIC (Time Domain Astronomy Information Center, a set of time-domain data analysis software) has provided an integrated data fusion interaction interface, this process still takes 10 - 30 minutes. For the volume of about 50 transient source candidates that need to be checked by the TA on average every day for the EP satellite, the EP Science Center needs to recruit at least 50 TAs to perform the daily transient source identification work. This makes the TA spend a great deal of energy on these relatively basic tasks, thus reducing the energy invested in solving more important scientific problems and lowering the scientific output rate.

[0004] There are currently some automated transient source identification methods to solve the problem of low efficiency in manually analyzing transient sources. For example, (1) an automated transient source identifier that can eliminate false sources caused by instruments and data processing to reduce the number of transient source candidates that need to be manually inspected; (2) the Asteroid Terrestrial impact Last Alert System (ATLAS) uses a candidate source detection algorithm to identify true and false sources and classify them, and at the same time uses a machine learning algorithm to identify the specific category of transient sources and submit them for manual inspection and confirmation; (3) the Zwicky Transient Facility (ZTF) uses a high-brightness transient target identifier based on a neural network to score transient source candidates, and at the same time identifies new transient sources from them and automatically submits follow-up observation requests; (4) the transient source search algorithm developed by the Swift space telescope can achieve automatic identification of transient sources with low latency, but ultimately, it is necessary to make a manual decision on whether to conduct follow-up observations. Although these methods have achieved a certain degree of automation to some extent, these automated operations are all discriminative operations, which can often only give a specific result and probability value, rather than clear data evidence and reasoning processes. For high-value transient source candidates, it is still necessary for humans to re-judge them based on relevant data, and it is impossible to fundamentally reduce the labor cost and improve the scientific output efficiency. Summary of the Invention

[0005] In view of the above problems, the present invention provides an agent-based time-domain astronomical transient source identification method, device, equipment, medium, and program product that reduce the manual secondary judgment of transient source candidates and improve the scientific output efficiency.

[0006] According to the first aspect of the present invention, an agent-based time-domain astronomical transient source identification method is provided, including: inputting observation data into an agent to obtain transient source candidates; using the agent to obtain a confirmation guidance strategy for the transient source candidates from a pre-constructed transient source knowledge base; according to the confirmation guidance strategy, using the agent to determine confirmation steps, and in each step, calling corresponding confirmation tools to obtain and process multi-band data of the transient source candidates; using the agent to analyze the processing results of each confirmation tool to obtain a confirmation conclusion; where the confirmation conclusion includes the type of transient source candidates, the reasoning process for determining the type of transient source candidates, and the marked key nodes; the reasoning process includes the characteristic data of each confirmation step.

[0007] According to an embodiment of the present invention, based on the identification guidance strategy, an agent is used to determine the identification steps, and corresponding identification tools are called in each step to obtain and process multi-band data of transient source candidates, including: based on the identification guidance strategy, the agent disassembles the identification task into multiple subtasks to generate identification steps; corresponding identification tools are called according to the identification steps to obtain and process multi-band data of transient source candidates.

[0008] According to an embodiment of the present invention, the agent is used to analyze the processing results of each identification tool to obtain the identification conclusion, including: the agent extracts the processing results of each identification tool respectively, and according to the processing results, obtains the characteristic data of each identification step; the confidence level of each identification step is calculated based on the characteristic data; the type of transient source candidate in each identification step is determined according to the confidence level of each identification step; the agent marks the nodes whose characteristic data in the identification step do not meet the preset conditions as key nodes.

[0009] According to an embodiment of the present invention, the reasoning process is obtained in the following manner:

[0010] Based on each identification step and its confidence level, and the type of transient source candidate in each identification step, a reasoning process is generated according to a preset reasoning method.

[0011] According to an embodiment of the present invention, the method further includes: constructing a visualization interface based on the agent operation log; displaying the identification conclusion and the historical record of identification tool calls through the visualization interface; if the user finds abnormal content from the identification conclusion and the historical record of identification tool calls, the agent is guided to optimize through natural language instructions or operation demonstrations.

[0012] According to an embodiment of the present invention, the method further includes: optimizing the transient source identification operation of the agent by using a dual-path identification method, a reinforcement learning method, and an evaluation tool.

[0013] Another aspect of the present invention provides an agent-based time-domain astronomical transient source identification device, including:

[0014] A transient source candidate acquisition module, configured to input observation data into the agent to obtain transient source candidates;

[0015] An identification guidance strategy acquisition module, configured to use the agent to obtain the identification guidance strategy of transient source candidates from a pre-constructed transient source knowledge base;

[0016] A transient source candidate processing module, configured to determine identification steps according to the identification guidance strategy by using the agent, and call corresponding identification tools in each step to obtain and process multi-band data of transient source candidates;

[0017] The identification conclusion acquisition module is used to utilize an intelligent agent to analyze the processing results of each identification tool and obtain an identification conclusion; wherein, the identification conclusion includes the type of transient source candidate, the reasoning process for determining the type of transient source candidate, and the marked key nodes; the reasoning process includes the characteristic data of each identification step.

[0018] Another aspect of the embodiments of the present invention provides an electronic device, including: one or more processors; a memory 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 are caused to execute the method as described above.

[0019] Another aspect of the embodiments of the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the processor is caused to execute the method as described above.

[0020] Another aspect of the embodiments of the present invention provides a computer program product, including a computer program, and when the computer program is executed by a processor, the method as described above is implemented.

[0021] The above one or more embodiments have the following beneficial effects: It can at least partially solve the problem that the existing transient source identification process lacks reasoning basis and confidence, and thus can realize that while the intelligent agent gives an identification conclusion, it can also give the reasoning process, identification decision, and the confidence corresponding to the decision in each step of the identification process. In this way, researchers can judge whether there are places with incorrect reasoning based on the reasoning process, identification decision, and the confidence corresponding to the decision in each step given, and adjust the incorrect reasoning process in time to optimize the intelligent agent and improve the identification accuracy;

[0022] It can also at least partially solve the problem of interaction between the intelligent agent and scientific researchers, and thus can realize guiding the intelligent agent to answer scientific researchers' questions, query / process / analyze data through natural language. In this way, the data analysis threshold is reduced, enabling scientific researchers to devote more energy to scientific achievement output and improving scientific productivity. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Through the following description of the embodiments of the present invention with reference to the drawings, the above content and other objects, features, and advantages of the present invention will become clearer. In the drawings:

[0024] Figure 1 Schematically shows an exemplary system architecture diagram in which the various methods and devices described herein can be implemented according to an embodiment of the present invention;

[0025] Figure 2 Schematically shows a flowchart of a method for identifying transient astronomical sources based on an intelligent agent according to an embodiment of the present invention;

[0026] Figure 3 Schematically shows a schematic diagram of the agent function according to an embodiment of the present invention;

[0027] Figure 4 Schematically shows a structural block diagram of an agent-based time-domain astronomical transient source identification device according to an embodiment of the present invention;

[0028] Figure 5 Schematically shows a block diagram of an electronic device suitable for implementing an agent-based time-domain astronomical transient source identification method according to an embodiment of the present invention.

[0029] It should be noted that, for clarity, in the drawings used to describe the embodiments of the present invention, the dimensions of the overall / local structure or the overall / local area may be enlarged or reduced, that is, these drawings are not drawn according to the actual scale. Detailed implementation manners

[0030] Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. In the following detailed description, for the sake of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present invention. However, obviously, one or more embodiments can also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present invention.

[0031] The terms used herein are merely for describing specific embodiments and are not intended to limit the present invention. The terms "including", "comprising", etc. used herein indicate the presence of the described features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0032] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0033] In the case of using expressions such as "at least one of A, B, and C, etc.", generally, it should be interpreted according to the meaning commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having only A, only B, only C, having A and B, having A and C, having B and C, and / or having A, B, and C, etc.).

[0034] Such as Figure 1As shown, the system architecture 100 according to this embodiment may include a terminal device 102, a network 103, and a server 104. The network 103 is used to provide a medium for the communication link between the terminal device 102 and the server 104. The network 103 may include various connection types, such as wired and / or wireless communication links, etc.

[0035] Users can use the terminal device 102 to interact with the server 104 through the network 103. Users can send observation data and / or problem descriptions 101 through the interaction interface provided by the terminal device 102. The terminal device 102 can send the observation data and / or problem descriptions 101 to the server 104 through the network 103, so that the server 104 can call the large model and output the identification conclusion reply content 105. The server sends the identification conclusion / reply content 105 to the terminal device 102, so that the terminal device 102 can display the identification conclusion / reply content 105 to the user.

[0036] Various communication client applications can be installed on the terminal device 102, such as intelligent assistant applications, knowledge reading applications, web browser applications, search applications, instant messaging tools, email clients, and / or social platform software, etc. (only as examples). Users can input observation data and / or problem descriptions 101 in the interaction interfaces of these client applications, and these client applications can display the generated identification conclusion / reply content 105 to the users.

[0037] In one embodiment, the server 104 can use the large model to generate the identification conclusion / reply content 105 to display the identification conclusion / reply content 105 on the terminal device 102.

[0038] The terminal device 102 can be configured with various electronic devices having a display screen and supporting web browsing, including but not limited to smartphones, tablets, laptop portable computers, and desktop computers, etc.

[0039] The server 104 can be a server that provides various services, such as a background management server that provides support for the content browsed by users using the interaction interface of the terminal device 102 (only as an example). The background management server can call the large model to perform identification operations or question answers on the received observation data and / or problem descriptions, and feedback the identification conclusion / reply content to the terminal device 102 and display it through the interaction interface. The server 104 can also be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system, solving the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or simply "VPS"). The server 104 can also be a server of a distributed system, or a server combined with a blockchain.

[0040] It should be noted that the method for identifying transient astronomical sources in the time domain based on an agent provided by the embodiments of the present invention can generally be executed by the server 104. Correspondingly, the device for identifying transient astronomical sources in the time domain based on an agent provided by the embodiments of the present invention can also be set in the server 104. The method for identifying transient astronomical sources in the time domain based on an agent provided by the embodiments of the present invention can also be executed by a server or a server cluster different from the server 104 and capable of communicating with the terminal device 102 and / or the server 104. Correspondingly, the device for identifying transient astronomical sources in the time domain based on an agent provided by the embodiments of the present invention can also be set in a server or a server cluster different from the server 104 and capable of communicating with the terminal device 102 and / or the server 104.

[0041] Alternatively, the method for identifying transient astronomical sources in the time domain based on an agent provided by the embodiments of the present invention can generally also be executed by the terminal device 102. Correspondingly, the device for identifying transient astronomical sources in the time domain based on an agent provided by the embodiments of the present invention can generally be set in the terminal device 102.

[0042] It should be understood that Figure 1 the numbers of the terminal devices and servers in are merely illustrative. According to the implementation requirements, there can be any number of terminal devices and servers.

[0043] In the technical solution of the present invention, the collection, storage, use, processing, transmission, provision, disclosure, and application of the user's personal information and other processes all comply with the provisions of relevant laws and regulations, take necessary confidentiality measures, and do not violate public order and good customs.

[0044] In the technical solution of the present invention, before obtaining or collecting the user's personal information, the authorization or consent of the user is obtained.

[0045] In view of this, the embodiments of the present invention provide a method for identifying transient astronomical sources in the time domain based on an agent. By constructing a transient source knowledge base to enhance and train the agent, the enhanced and trained agent can identify transient sources, give identification conclusions, and can also interact with users, answer users' questions or re-analyze data according to users' requirements. Users can also judge whether the reasoning in the identification process is incorrect based on the provided identification evidence, modify the incorrect reasoning process, and optimize the agent at the same time, so as to improve the identification accuracy and reduce the labor cost.

[0046] Figure 2 is a flowchart of a method for identifying transient astronomical sources in the time domain based on an agent according to an embodiment of the present invention.

[0047] As Figure 2 shown, the method 200 includes:

[0048] In operation S210, observation data is input into the agent to obtain transient source candidates.

[0049] In some embodiments, an agent includes a system or entity capable of autonomously perceiving the environment, making decisions, and performing actions to complete specific tasks. The large model can provide decision support for the agent, providing the agent with the ability of reasoning analysis and task planning. The agent utilizes the analysis results of the large model to execute or optimize its decision-making process. The agent can integrate multiple large models to handle different types of tasks.

[0050] A large model may refer to a deep learning model with a large number of model parameters. A large model usually contains hundreds of millions, tens of billions, hundreds of billions, trillions or even more than one hundred trillion model parameters. Large models can include large language models, vision large models, multi-modal large models, and so on. The large model involved in the embodiments of the present invention can be a general large model, or can also be an expert large model obtained after fine-tuning based on requirements. The embodiments of the present invention do not limit this.

[0051] In some embodiments, after the agent performs preprocessing operations such as calibration and denoising on the observation data, the image difference method is used to identify new or disappeared sources from the preprocessed observation data, or detect mutation points in the preprocessed observation data. The pulse search method can also be used to detect isolated pulses, and the sources corresponding to the identified new sources, mutation points or isolated pulses are used as transient source candidates.

[0052] In operation S220, the agent is used to obtain the confirmation guidance strategy of the transient source candidates from the pre-constructed transient source knowledge base.

[0053] In some embodiments, the transient source knowledge base includes various materials in the field of transient sources, the transient source confirmation process, existing transient source confirmation records, and a set of question-and-answer pairs; among them, the set of question-and-answer pairs is constructed based on various materials in the field of transient sources, the transient source confirmation process, and existing transient source confirmation records;

[0054] Furthermore, various materials in the field of transient sources can be stored in the transient source knowledge base in the form of vector data; the transient source confirmation process, existing transient source confirmation records, and the set of question-and-answer pairs can be stored in the transient source knowledge base in the form of a knowledge graph.

[0055] In some embodiments, the agent is fine-tuned according to the set of question-and-answer pairs so that the fine-tuned agent is more professional and accurate in answering user questions.

[0056] In some embodiments, a retrieval library is constructed based on various materials in the transient source field, and the intelligent agent is enhanced using a retrieval enhancement method, enabling the enhanced intelligent agent to query content related to the transient source candidate from the retrieval library and provide responses based on the queried content during interaction with the user. This can not only improve the professionalism of the responses but also enhance the confirmation accuracy;

[0057] For example, after the user inputs observation data or a problem description to the intelligent agent, the intelligent agent filters out transient source candidates based on the observation data, generates vector data based on the transient source candidates or converts the problem description into vector data, and searches for content related to the transient source candidates or the problem description through vector indexing; among them, the vector index can be constructed in the following manner:

[0058] The named entity recognition method is used to extract the key terms, key contents and conclusions in the abstracts of each material in the retrieval library, parse the meta-information such as the titles, authors, and publication years of each material, convert the extracted key terms, key contents and conclusions in the abstracts and the meta-information into vector data, and construct a vector index;

[0059] For example, based on the transient source candidate, the infrared band color index of the transient source candidate can be obtained, and through the vector index, content related to the infrared band color index is found in the retrieval library. This related content includes the infrared band color indices of different celestial body types. By comparing the infrared band color index of the transient source candidate with the found infrared band color indices of different celestial body types, the celestial body type of the transient source candidate can be preliminarily determined. During the subsequent confirmation process, the found infrared band color indices of different celestial body types can be used as subsequent confirmation guidance strategies; for another example, the problem description input by the user is "What are the light variation characteristics of cosmic rays?" This question focuses on the light variation characteristics of cosmic rays. By searching through the vector index in the retrieval library for content related to the light variation characteristics of cosmic rays, it may be found that the light variation analysis result of cosmic rays shows a single-frame bulge. Then the intelligent agent will display the found relevant content to the user as the response content, that is, display "The light variation analysis result of cosmic rays shows a single-frame bulge" as the response content to the user.

[0060] In some embodiments, the identification guidance strategy includes determining the steps to be performed, the data to be called, and how to analyze the data for transient source candidate types. The data to be called may include the astronomical features of existing transient sources. For example, various materials in the field of transient sources record that steps such as multi-band cross-validation and celestial source table matching are required to determine the transient source candidate type. The transient source identification process and the existing transient source identification records record which data to call and how to analyze the data. For example, data such as transient source candidates, multi-band reference data, spectral data, and astrometric data of existing transient sources are required to further determine the type of transient source candidates. For another example, the data analysis method recorded in the transient source identification process and the existing transient source identification records is to judge whether there is a single-frame bulge in the transient source candidate through the light curve. If so, it can be preliminarily judged that the transient source candidate is a cosmic ray.

[0061] In operation S230, according to the identification guidance strategy, an agent is used to determine the identification steps, and in each step, the corresponding identification tool is called to obtain and process the multi-band data of the transient source candidate.

[0062] In some embodiments, before using the agent to call the corresponding identification tool to process the multi-band data of the transient source candidate, the agent is trained according to the transient source knowledge base so that the trained agent can call the corresponding identification tool according to the identification steps and learn how to identify the observation data to determine the type of the transient source candidate.

[0063] Specifically, the agent is trained according to the transient source identification process. Each step of the transient source identification process is converted into a regularized description, and the identification tools required for each step are structurally extracted to obtain structural parameters. The regularized description can be: if the identification parameters of the current step meet the preset execution conditions, then execute the next step. The regularized description and the structural parameters are used as the first training sample to train the agent.

[0064] Since the transient source identification process is natural language text, these natural language texts need to be converted into identification tool call commands, and then the corresponding identification tools are called through the API. To implement the call of the identification tools, the corpus of calling each identification tool using natural language and the call commands of each identification tool are collected to obtain the identification tool call data set and used as the second training sample to train the agent so that the agent can call the corresponding identification tool. The API is encapsulated according to the model context protocol by each identification tool (such as the long-term light curve generation tool, the multi-band cross-validation tool, the celestial source table matching tool), multi-band reference data, etc. Therefore, the multi-band reference data can be obtained through the API.

[0065] Use the observation data in the existing transient source identification records as the third training sample and the existing identification conclusions as the labels, and train the agent based on the third training sample and the labels so that the agent learns to analyze the observation data and obtain the identification conclusions.

[0066] In operation S240, use the agent to analyze the processing results of each identification tool to obtain the identification conclusions; among them, the identification conclusions include the types of transient source candidates, the reasoning process for determining the types of transient source candidates, and the key nodes of the labels; the reasoning process includes the characteristic data of each identification step.

[0067] In the embodiments of the present invention, the agent based on retrieval enhancement realizes transient source identification by calling identification tools and adopting a multi-modal data analysis method. It can not only obtain the types of transient sources, but also obtain the reasoning process of identification, which is beneficial for researchers to check the identification process according to the identification conclusions, correct the identification conclusions with abnormal content, and optimize the agent based on the correction operation, improve the identification accuracy, reduce the labor cost. The agent interacts with the user, and the user can guide the agent to re-analyze the data, reducing the threshold of data analysis, and thus improving the scientific output rate.

[0068] In some embodiments, in operation S230, according to the identification guidance strategy, use the agent to determine the identification steps, and in each step, call the corresponding identification tool to obtain and process the multi-band data of the transient source candidate, including:

[0069] According to the identification guidance strategy, use the agent to decompose the identification task into multiple subtasks and generate the identification steps;

[0070] Call the corresponding identification tool according to the identification steps to obtain and process the multi-band data of the transient source candidate.

[0071] For example, it is obtained from the transient source knowledge base that to perform the identification task (determine the type of transient source candidate), it is necessary to call multi-band reference data, as well as spectral data, X-ray band images, long-term light curve data of existing transient sources, etc. It is necessary to perform steps such as multi-band cross-validation, light curve analysis, spectral fitting, and follow-up observation. According to these identification guidance strategies, the agent decomposes the identification task into 4 subtasks, generates the identification steps according to these 4 subtasks, and the agent performs the identification operation step by step according to the generated identification steps. During the execution of multi-band cross-validation, light curve analysis, and spectral fitting, the corresponding identification tools are called to perform the corresponding operations. According to the data analysis methods learned during the training process, comprehensive analysis is performed on multi-modal data such as the processing results of each identification tool, the identification guidance strategy, and the multi-band reference data, and the type of transient source candidate can be obtained. Finally, the follow-up observation task is performed to mark the transient source candidate of interest for subsequent observation.

[0072] In some embodiments, in operation S240, an agent is used to analyze the processing results of each identification tool to obtain an identification conclusion, including:

[0073] The agent extracts the processing results of each identification tool respectively, and obtains the characteristic data of each identification step according to the processing results;

[0074] Calculate the confidence level of each identification step based on the characteristic data;

[0075] Determine the type of transient source candidate in each identification step according to the confidence level of each identification step;

[0076] The agent labels the nodes whose characteristic data in the identification step do not meet the preset conditions as key nodes.

[0077] For example, the identification tool called by the agent in the first identification step is a long-term light curve generation tool, and the identification tool called in the second identification step is an X-ray spectrum analysis tool. The characteristic data output by the long-term light curve generation tool and the X-ray spectrum analysis tool are extracted respectively to obtain the light variation characteristics and spectral characteristics of the transient source candidate. In the first identification step, assuming that the light variation characteristic of the transient source candidate shows a single-frame bulge, and in the extracted identification guidance strategy, the light variation characteristic of the cosmic ray in the light variation characteristic of the transient source shows a single-frame bulge, the light variation characteristic of the transient source candidate showing a single-frame bulge can be used as an evidence data. Based on this evidence data, calculate the confidence level that the transient source candidate obtained in the first identification step is a cosmic ray. In this embodiment, the light variation analysis result of the transient source candidate shows a single-frame bulge, and it is judged to be a cosmic ray, and the confidence level of this judgment result is 98%; in the second identification step, assuming that the spectrum of the transient source candidate is a power-law spectrum, according to the characteristic data of the multi-band reference data, a transient source with a power-law spectrum is a gamma-ray burst. The spectrum of the transient source candidate being a power-law spectrum can be used as a new evidence data. According to this new evidence data and the confidence level of the previous step, calculate the confidence level of the second identification step. If the evidence data of the second identification step is that the spectrum of the transient source candidate is a power-law spectrum, and the confidence level that the transient source candidate is a gamma-ray burst is the highest, then it can be considered that the type of transient source obtained in this identification step is a gamma-ray burst. Finally, combine the conclusions of the first identification step and the second identification step to finally obtain the type of the transient source;

[0078] In the above identification process, the processing result of the long-term light curve generation tool is subjected to feature extraction to obtain the brightness change amplitude. If the brightness change amplitude is very small or almost none, it indicates that there may be a lack of optical counterpart. The brightness change amplitude does not meet the preset conditions (brightness change amplitude threshold), so this node (lack of optical counterpart) is labeled as a key node for the user to guide the agent to re-analyze the data according to the key node.

[0079] In some embodiments, the confidence level of the authentication step is calculated as follows:

[0080]

[0081] in, Indicates that the current step temporary source candidate is Class Source The probability of Indicates Evidence data, Indicates the current step The marginal likelihood probability of the evidence data is used as a normalization factor. Indicates that the current step temporarily presents the source candidate as Class Source When The probability of Indicates that the next step evidence data is When Class Source The posterior probability (i.e., confidence level) of the evidence data can be the data obtained by extracting features from the data obtained by the authentication tool;

[0082] When calculating the confidence of the first authentication step, Indicates that the transient source candidate is Class Source The prior probability is related to the observation facilities, such as the observation depth, the coverage area, the sensitivity of the observation instrument, and the occurrence density of various transient sources in different bands. In the absence of evidence data, As an uninformative prior probability, that is, among all transient source types, the transient source candidate is the first Class Source The probability of is considered to be uniformly distributed.

[0083] In some embodiments, Calculated by the following formula:

[0084]

[0085] in, Indicates that the current step temporary source candidate is Class Source The probability of Indicates that the current step temporarily presents the source candidate as Class Source When probability.

[0086] In some embodiments, the feature data may further include infrared band color index, X-ray band image features, astrometric features, optical band image features, and single light variation features.

[0087] In some embodiments, in operation S240, the inference process is obtained according to the following manner:

[0088] According to each identification step and its confidence level, and the types of transient source candidates in each identification step, an inference process is generated according to a preset inference method.

[0089] For example, the first identification step is to process the transient source candidate using a long-term light curve generation tool. In this step, the confidence level that the transient source candidate is a cosmic ray is the highest, so it is considered that the transient source candidate is a cosmic ray; in the second identification step, the transient source candidate is processed using an X-ray spectral analysis tool. In this step, the confidence level that the transient source candidate is a gamma-ray burst is the highest, so it is considered that the transient source candidate is a gamma-ray burst; in the third identification step, the transient source candidate is processed using an astrometric tool. In this step, the confidence level that the transient source candidate is a high-energy gamma-ray burst is the highest, so it is considered that the transient source candidate is a high-energy gamma-ray burst; using inference methods such as inductive reasoning and deductive reasoning, an inference process is generated. This inference process can be described as follows: in the first identification step, feature data is extracted from the processing result of the long-term light curve generation tool. This feature data is the same as the feature data of cosmic rays in the identification guidance strategy or multi-band reference data, and the confidence level that the transient source candidate is a cosmic ray is the highest, so the transient source candidate is determined to be a cosmic ray; in the second identification step, feature data is extracted from the processing result of the X-ray spectral analysis tool. This feature data is the same as the feature data of gamma-ray bursts in the identification guidance strategy or multi-band reference data, and the confidence level that the transient source candidate is a gamma-ray burst is the highest, so the transient source candidate is determined to be a gamma-ray burst; in the third identification step, feature data is extracted from the processing result of the astrometric tool. This feature data is the same as the feature data of high-energy gamma-ray bursts in the identification guidance strategy or multi-band reference data, and the confidence level that the transient source candidate is a high-energy gamma-ray burst is the highest, so the transient source candidate is determined to be a high-energy gamma-ray burst; combining these three identification steps, the type of the transient source candidate is finally obtained as a high-energy gamma-ray burst.

[0090] In some embodiments, method 200 further includes:

[0091] Construct a visualization interface based on the agent operation log;

[0092] Display the identification conclusion and the historical record of identification tool calls through the visualization interface;

[0093] If the user finds abnormal content from the identification conclusion and the historical record of identification tool calls, the agent is guided to be optimized through natural language instructions or operation demonstrations.

[0094] For example, the user guides the agent to modify improper operations through natural language instructions (such as preferentially identifying transient source candidates that do not match the optical counterparts) or operation demonstrations (for example, marking the abnormal interval of the light curve in the identification evidence on the agent display interface). The agent records the modification operation and converts the modification operation into structured data, and adjusts its own model parameters according to the structured data to achieve optimization; wherein, the structured data includes the type of modification operation and the parameters that need to be adjusted for executing the modification operation.

[0095] In some embodiments, the visualization interface also shows the confidence level of each identification step to the user in the form of heat map annotation;

[0096] The agent uses the decision path visualization tool to display the call logic of the identification tool in real time through the visualization interface to assist the user in review and reduce the number of iterations during the training process of the agent.

[0097] In some embodiments, the identification tool call history record includes the X-ray flux data of the transient source candidates obtained by the identification tool and the fitting parameter adjustment record. By showing the historical state of the identification tool call to the user, it is possible to judge in real time whether the information between each identification step is coherent, thereby improving the identification accuracy.

[0098] In some embodiments, method 200 further includes:

[0099] Optimizing the transient source identification operation of the agent by using the dual-path identification method, reinforcement learning method, and evaluation tool;

[0100] Furthermore, the dual-path identification method means that the agent also constructs a virtual environment, uses the transient source identification algorithm to identify the transient source, compares the identification conclusion obtained by this method with the identification conclusion obtained by method 200, evaluates the identification performance of these two methods. For example, according to the preset scoring criteria, calculate the scores of these two methods in terms of the confidence level of the identification step, identification conclusion, etc., and improve the part with a low score in method 200. The dual-path identification method also verifies the end-to-end reliability of the identification tool call chain;

[0101] Using the evaluation tool to record in real time the resource consumption, processing time, identification accuracy, efficiency, adaptability and other indicators during the agent's identification process, and improving the part with poor indicator performance;

[0102] Optimizing the agent based on the reinforcement learning method is to set the reward function and optimization goal. Based on the reward function and optimization goal, select a suitable reinforcement learning method to optimize the agent until the optimization goal is achieved.

[0103] Figure 3Schematically shows a schematic diagram of the agent function according to an embodiment of the present invention.

[0104] As Figure 3 shown, the agent 310 is trained, enhanced, and fine-tuned according to the constructed transient source knowledge base. When the observation data and / or problem description are input into the trained, enhanced, and fine-tuned agent, the agent calls the identification tool according to the identification steps. According to the processing results of the identification tool, combined with multi-modal data such as multi-band parameter data and feature data of existing transient sources, multi-modal data analysis is performed to obtain evidence data. Based on the evidence data, the confidence of each identification step is calculated. According to the confidence of each identification step, the final type of transient source candidate is determined. According to each identification step and its confidence, and the type of transient source candidate in each identification step, the reasoning process is obtained. During the process of the agent performing transient source identification, according to the processing results of the identification tool, multi-modal data such as multi-band parameter data and feature data of existing transient sources, key nodes are marked, and the final type of transient source candidate, the reasoning process, and the marked key nodes are output as the identification conclusion through the visualization interface; the agent retrieves content related to the problem description from the transient source knowledge base, generates a reply content and outputs it. The user judges whether the reasoning process is correct according to the output identification conclusion, makes modifications to the incorrect reasoning process, and the agent adjusts its own model parameters according to the user's modification operation, so as to realize the optimization of the agent and improve the accuracy and reliability of identification.

[0105] Based on the above time-domain astronomical transient source identification method, the present invention also provides an agent-based time-domain astronomical transient source identification device. The following will be combined with Figure 4 to describe this device in detail.

[0106] Figure 4 Schematically shows a structural block diagram of an agent-based time-domain astronomical transient source identification device according to an embodiment of the present invention.

[0107] As Figure 4 shown, the device 400 includes a transient source candidate acquisition module 410, an identification guidance strategy acquisition module 420, a transient source candidate processing module 430, and an identification conclusion acquisition module 440.

[0108] The transient source candidate acquisition module 410 is used to input the observation data into the agent to obtain transient source candidates. In one embodiment, the transient source candidate acquisition module 410 can be used to perform the operation S210 described above, which will not be elaborated here.

[0109] The identification guidance strategy acquisition module 420 is used to use the agent to obtain the identification guidance strategy of the transient source candidate from the pre-constructed transient source knowledge base. In one embodiment, the identification guidance strategy acquisition module 420 can be used to perform the operation S220 described above, which will not be elaborated here.

[0110] The transient source candidate processing module 430 is configured to determine the identification steps by using an agent according to the identification guidance strategy, and call corresponding identification tools in each step to obtain and process the multi-band data of the transient source candidates. In one embodiment, the transient source candidate processing module 430 may be configured to perform the operation S230 described above, which will not be elaborated herein.

[0111] The identification conclusion obtaining module 440 is configured to analyze the processing results of the respective identification tools by using an agent to obtain an identification conclusion; wherein, the identification conclusion includes the type of the transient source candidate, the reasoning process for determining the type of the transient source candidate, and the marked key nodes; the reasoning process includes the characteristic data of each identification step. In one embodiment, the identification conclusion obtaining module 440 may be configured to perform the operation S240 described above, which will not be elaborated herein.

[0112] For parts not mentioned in the device section, reference may be made to the respective embodiments of the above method for understanding. That is, the device section includes modules respectively configured to perform the respective steps of any one of the method embodiments described above. Moreover, the implementation manners, the technical problems solved, the functions achieved, and the technical effects achieved by the respective modules / units / sub-units, etc. in the device section embodiment are respectively the same as or similar to those of the corresponding steps in the method section embodiment, which will not be elaborated herein.

[0113] According to the embodiments of the present invention, any plurality of modules among the transient source candidate obtaining module 410, the identification guidance strategy obtaining module 420, the transient source candidate processing module 430, and the identification conclusion obtaining module 440 may be combined and implemented in one module, or any one of the modules may be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules may be combined with at least part of the functions of other modules and implemented in one module.

[0114] According to an embodiment of the present invention, at least one of the transient source candidate acquisition module 410, the identification guidance strategy acquisition module 420, the transient source candidate processing module 430, and the identification conclusion acquisition module 440 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on a substrate, a system in a package, an application specific integrated circuit (ASIC), or any other reasonable manner of integrating or packaging circuits, etc., implemented by hardware or firmware, or implemented in any one of the three implementation manners of software, hardware, and firmware, or in an appropriate combination of any several of them. Alternatively, at least one of the transient source candidate acquisition module 410, the identification guidance strategy acquisition module 420, the transient source candidate processing module 430, and the identification conclusion acquisition module 440 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0115] Figure 5 FIG. schematically shows a block diagram of an electronic device suitable for implementing the method for identifying transient astronomical sources in the time domain based on an agent according to an embodiment of the present invention.

[0116] As Figure 5 shown, the electronic device 500 according to an embodiment of the present invention includes a processor 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 section 508 into a random access memory (RAM) 503. The processor 501 can include, for example, a general microprocessor (such as a CPU), an instruction set processor, and / or a related chipset, and / or a dedicated microprocessor (such as an application specific integrated circuit (ASIC)), etc. The processor 501 can also include on-board memory for caching purposes. The processor 501 can include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0117] In the RAM 503, various programs and data required for the operation of the electronic device 500 are stored. The processor 501, the ROM 502, and the RAM 503 are connected to each other through a bus 504. The processor 501 performs various operations of the method flow according to an embodiment of the present invention by executing the programs in the ROM 502 and / or the RAM 503. It should be noted that the programs can also be stored in one or more memories other than the ROM 502 and the RAM 503. The processor 501 can also perform various operations of the method flow according to an embodiment of the present invention by executing the programs stored in the one or more memories.

[0118] According to an embodiment of the present invention, the electronic device 500 may further include an input / output (I / O) interface 505, and the input / output (I / O) interface 505 is also connected to the bus 504. The electronic device 500 may further include one or more of the following components connected to the input / output (I / O) interface 505: an input portion 506 including a keyboard, a mouse, etc.; an output portion 507 including, for example, a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 508 including a hard disk, etc.; and a communication portion 509 including a network interface card such as a LAN card, a modem, etc. The communication portion 509 performs communication processing via a network such as the Internet. The drive 510 is also connected to the input / output (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 mounted on the drive 510 as needed so that a computer program read therefrom is installed into the storage portion 508 as needed.

[0119] The present invention also provides a computer-readable storage medium, which may be included in the device / device / system described in the above embodiment; or may exist separately without being assembled into the device / device / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present invention is implemented.

[0120] According to an embodiment of the present invention, the computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, it may include but is not limited to: 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), 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, the computer-readable storage medium may be any tangible medium that contains or stores a program, and the program may be used by or in combination with an instruction execution system, device, or device. For example, according to an embodiment of the present invention, the computer-readable storage medium may include the above-described ROM 502 and / or RAM 503 and / or one or more memories other than ROM 502 and RAM 503.

[0121] An embodiment of the present invention also includes a computer program product, which includes a computer program, and the computer program includes program codes for executing the method shown in the flowchart. When the computer program product runs on a computer system, the program codes are used to cause the computer system to implement the method provided by the embodiment of the present invention.

[0122] When the computer program is executed by the processor 501, the above functions defined in the system / apparatus of the embodiments of the present invention are executed. According to an embodiment of the present invention, the above-described systems, apparatuses, modules, units, etc. can be implemented by computer program modules.

[0123] In one embodiment, the computer program can rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program can also be transmitted and distributed in the form of signals on a network medium, and be downloaded and installed through the communication part 509, and / or be installed from the removable medium 511. The program code included in the computer program can be transmitted by any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0124] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 509, and / or be installed from the removable medium 511. When the computer program is executed by the processor 501, the above functions defined in the system of the embodiments of the present invention are executed. According to an embodiment of the present invention, the above-described systems, devices, apparatuses, modules, units, etc. can be implemented by computer program modules.

[0125] According to an embodiment of the present invention, the program code for executing the computer program provided in the embodiments of the present invention can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages include but are not limited to, such as Java, C++, python, the "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, by using an Internet service provider to connect through the Internet).

[0126] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, a segment of a program, or a part of code, and the above-mentioned module, segment of a program, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order from that marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, as well as the combinations of blocks in the block diagram or flowchart, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0127] Those skilled in the art can understand that the features described in the various embodiments of the present invention can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present invention. In particular, without departing from the spirit and teachings of the present invention, the features described in the various embodiments of the present invention can be combined and / or combined in various ways. All such combinations and / or combinations fall within the scope of the present invention.

[0128] The embodiments of the present invention have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of the present invention. Although the embodiments have been described separately above, this does not mean that the measures in each embodiment cannot be used advantageously in combination. Without departing from the scope of the present invention, those skilled in the art can make various substitutions and modifications, and all such substitutions and modifications should fall within the scope of the present invention.

Claims

1. An agent-based method for identifying transient astronomical sources in the time domain, characterized in that The method includes: Inputting the observation data into an agent to obtain transient source candidate bodies; Using the agent to obtain the identification guidance strategy of the transient source candidate bodies from a pre-constructed transient source knowledge base; According to the identification guidance strategy, using the agent to determine the identification steps, and calling corresponding identification tools in each step to obtain and process multi-band data of the transient source candidate bodies; Using the agent to analyze the processing results of each identification tool to obtain an identification conclusion; wherein, the identification conclusion includes the type of the transient source candidate body, the reasoning process for determining the type of the transient source candidate body, and the marked key nodes; the reasoning process includes the characteristic data of each identification step.

2. The method according to claim 1, characterized in that, The step of using the agent to determine the identification steps according to the identification guidance strategy and calling corresponding identification tools in each step to obtain and process multi-band data of the transient source candidate bodies includes: According to the identification guidance strategy, using the agent to decompose the identification task into multiple subtasks and generate identification steps; Calling corresponding identification tools according to the identification steps to obtain and process multi-band data of the transient source candidate bodies.

3. The method according to claim 2, characterized in that, The step of using the agent to analyze the processing results of each identification tool to obtain an identification conclusion includes: Using the agent to separately extract the processing results of each identification tool, and according to the processing results, obtaining the characteristic data of each identification step; Calculating the confidence level of each identification step based on the characteristic data; Determining the type of the transient source candidate body in each identification step according to the confidence level of each identification step; Using the agent to mark the nodes whose characteristic data in the identification steps do not meet the preset conditions as key nodes.

4. The method according to claim 3, wherein The reasoning process is obtained according to the following method: According to each identification step and its confidence level, and the type of the transient source candidate body in each identification step, generating a reasoning process according to a preset reasoning method.

5. The method according to claim 1, characterized in that, The method further includes: Based on the operation log of the agent, constructing a visualization interface; Displaying the identification conclusion and the historical record of calling identification tools through the visualization interface; If the user finds abnormal content from the identification conclusion and the historical record of calling identification tools, guiding the agent to optimize through natural language instructions or operation demonstrations.

6. The method according to claim 1, wherein The method further includes: Adopting a dual-path identification method, a reinforcement learning method, and an evaluation tool to optimize the transient source identification operation of the agent.

7. An agent-based time-domain astronomical transient source identification device, characterized in that The device includes: A transient source candidate body acquisition module, configured to input the observation data into an agent to obtain transient source candidate bodies; An identification guidance strategy acquisition module, configured to use the agent to obtain the identification guidance strategy of the transient source candidate bodies from a pre-constructed transient source knowledge base; A transient source candidate body processing module, configured to use the agent to determine the identification steps according to the identification guidance strategy, and call corresponding identification tools in each step to obtain and process multi-band data of the transient source candidate bodies; An identification conclusion acquisition module, configured to use the agent to analyze the processing results of each identification tool to obtain an identification conclusion; wherein, the identification conclusion includes the type of the transient source candidate body, the reasoning process for determining the type of the transient source candidate body, and the marked key nodes; the reasoning process includes the characteristic data of each identification step.

8. An electronic device, including: One or more processors; A memory, configured to store one or more computer programs, Wherein, the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 6.

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

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

Images