A telescope observation planning method based on a large language model

Abstract

This invention discloses a telescope observation planning method based on a large language model, including constructing a neural network model for telescope observation planning, constructing a knowledge graph for telescope observation planning, applying the large language model to construct a knowledge graph of telescope observation application information, and combining a deep learning model to achieve intelligent telescope observation planning. This invention combines artificial intelligence technology with the needs of telescope observation planning, leverages the interactive advantages of the large language model, uses deep learning algorithms to construct a neural network model, applies the large language model to construct a knowledge graph of telescope observation application information, and correlates expert knowledge, big data sources, and the large language model to analyze combinations of numerous observation tasks to find the optimal observation planning scheme, maximizing the telescope's observation capabilities.

Description

Technical Field

[0001] This invention belongs to the field of astronomical telescope technology, specifically relating to a method for telescope observation planning based on a large language model. Background Technology

[0002] Telescope observation planning refers to the process of developing an observation plan and tasks based on the needs and objectives of astronomy before conducting telescope observations. This includes determining the observation time, the astronomical characteristics of the target, the observation method, and the selection of observation sites. Existing telescope observation planning methods are generally based on the experience and expertise of astronomical specialists, employing manual or semi-automated approaches. However, with the continuous development of telescope technology and the explosive growth of observational data, traditional planning methods suffer from inefficiency, lack of precision, and inability to cope with changes. Therefore, researching more intelligent telescope observation planning methods is of great significance for promoting the development of astronomy and other scientific fields.

[0003] Artificial intelligence (AI) technology has been widely applied in the field of astronomy, and new AI technologies and application scenarios are constantly being introduced. Deep learning is a machine learning technique based on artificial neural networks. Its main advantage lies in its ability to automatically discover complex nonlinear relationships in input data without manual feature engineering, thus greatly improving the accuracy and generalization ability of algorithms. Deep learning is based on neural networks, and it achieves efficient data processing and pattern recognition through deep neural networks. A deep neural network consists of multiple neurons and multiple layers. Each neuron is responsible for processing a portion of the input data and passing it to the next layer, thereby achieving high-level abstraction and extraction of information. The input layer receives raw data about the applicant, and the output layer produces a prediction of the probability of acceptance. Several hidden layers pass information between these two layers. Backpropagation is used to adjust network parameters and train the model to minimize prediction errors, thereby improving the model's accuracy. A knowledge graph is a knowledge representation method based on a graph structure that describes the relationships between entities. It is a semantic network model used to express and organize real-world knowledge, aiming to transform various information, entities, and concepts in the real world into a computer-readable form. Knowledge graphs typically consist of three elements: entities, relationships, and attributes. Entities represent things in the real world, relationships describe the connections between entities, and attributes describe the characteristics or properties of entities.

[0004] Large language models refer to natural language processing models with billions or trillions of parameters. They are deep learning-based models used for tasks such as natural language processing, text generation, translation, summarization, and question answering. Large language models learn language rules and patterns from massive amounts of text data by training large-scale neural network models, thereby achieving the understanding and generation of natural language. These models typically have a multi-layered Transformer architecture and employ techniques such as self-attention mechanisms and positional encoding to process text sequences, enabling them to handle longer text sequences and more linguistic context. Summary of the Invention

[0005] The purpose of this invention is to combine artificial intelligence technology with the needs of telescope observation planning, leverage the interactive advantages of large language models, use deep learning algorithms to construct neural network models, apply large language models to construct a knowledge graph of telescope observation application information, and combine expert knowledge, big data sources, and large language models to analyze a large number of observation tasks in order to find the optimal observation planning scheme and maximize the telescope's observation capabilities.

[0006] Specifically, the present invention is implemented using the following technical solutions.

[0007] 1. Constructing a Neural Network Model for Telescope Observation Planning

[0008] This invention provides a telescope observation planning method based on a large language model, including the construction of a telescope observation planning neural network model based on deep learning.

[0009] 1.1 Data Collection and Preprocessing: Collect data related to the telescope observation plan to build a deep learning model, and then preprocess the data, including outlier correction, missing value handling, data normalization, etc.; further, after dividing the dataset into training dataset, validation dataset and test dataset, save these datasets to the database.

[0010] 1.2 Neural Network Model Construction: This section describes the writing of neural network code and the construction of a neural network model based on the deep learning framework TensorFlow. The structure of this neural network model includes the number of neurons in the input layer, hidden layers, and output layer, as well as the activation function and loss function.

[0011] 1.3 Neural Network Model Training: The mean squared error is used as the loss function, and stochastic gradient descent (SGD) is used as the optimization algorithm. The telescope observation application data is input into the neural network model, and the model parameters are continuously adjusted and the loss function is minimized through the optimization algorithm, so that the model gradually approaches the optimal solution.

[0012] 1.4 Model Evaluation and Tuning: The model's performance is evaluated using a validation set, including metrics such as accuracy, recall, and F1-score. Finally, based on the model evaluation results, the model's parameters and architecture are adjusted, and the model is optimized through continuous iteration to bring its performance to the best.

[0013] 2. Constructing a knowledge graph for telescope observation planning

[0014] This invention provides a telescope observation planning method based on a large language model, which includes inferring the relationships between features of related entity information based on Grakn fusion of expert knowledge, large language model and big data source, and generating and storing them in a knowledge graph.

[0015] 2.1 Construct a knowledge graph framework. The knowledge graph library in the model base stores telescope observation planning information, including relevant information such as observation priority, application time for observation, and observation targets;

[0016] 2.2 Acquire relevant knowledge through interaction with expert knowledge and big data sources to construct a telescope observation map;

[0017] 2.3 Improve telescope observation maps through large language model interaction, such as intelligent information service sources like CHATGPT, BARD, and Wenxin Yiyan. Perform text mining, text understanding and analysis on the responses from the large language model, generate structured feedback opinions and send them back to the telescope observation map library.

[0018] 2.4 Perform text mining, text understanding and analysis on the response content of the large language model, generate structured feedback opinions and send them back to the telescope observation planning knowledge graph database;

[0019] 3. Implementation of a Telescope Observation Planning Based on a Large Language Model

[0020] This invention provides a telescope observation planning method based on a large language model, which includes constructing a knowledge graph of telescope observation application information using a large language model, combining it with a deep learning model to realize intelligent telescope observation planning, leveraging the interactive advantages of the large language model, and maximizing the telescope's observation capabilities.

[0021] 3.1 Preprocess the telescope observation planning application and extract the features of the telescope observation planning information, including the telescope applicant information, the telescope observation application objectives, the telescope observation application time, the telescope observation applicant's published articles, and the scientific significance of the telescope observation.

[0022] 3.2 Input the preprocessed features containing telescope observation planning information into the telescope observation planning knowledge graph, which is composed of a large language model, a large data source, and expert knowledge;

[0023] 3.3 The knowledge graph uses relevant expert knowledge, large language models, and deep learning models to infer the relationships between the features of related entity information, and stores them in the knowledge graph library;

[0024] 3.4 Search and match the telescope observation planning feature attributes in the telescope observation planning knowledge graph to obtain the historical interaction data of the telescope observation planning;

[0025] 3.5 If there are matching historical telescope observation plan feature attributes in the telescope observation plan knowledge graph, compare and evaluate the current telescope observation plan feature attributes with the previous observation plan feature attributes, and update the telescope observation plan feature attributes.

[0026] 3.6 If there is no matching historical telescope observation plan feature attribute in the telescope observation plan knowledge graph, send the current telescope observation plan feature attribute to the large language model for further processing and update the telescope observation plan knowledge graph to obtain the updated telescope observation plan feature attribute.

[0027] 3.7 The updated telescope observation planning feature attributes, including the telescope observation applicant, the telescope observation application unit, the telescope observation application target, the number of telescope observation applications, and the telescope observation application results, are converted into feature vectors and input into the telescope observation planning neural network model;

[0028] 3.8 The telescope observation map is converted into feature vectors and input into the neural network model of the course. The telescope observation plan is generated by combining the matching results and success rate.

[0029] 3.9 The telescope observation plan is generated by combining the matching results and success rate of the telescope observation planning neural network model.

[0030] The advantages of this invention's telescope observation planning based on a large language model are as follows:

[0031] This invention provides a telescope observation planning method based on a large language model. Combining artificial intelligence technology with the needs of telescope observation planning, it leverages the interactive advantages of the large language model, uses deep learning algorithms to construct a neural network model for telescope observation planning, and connects the large language model, expert knowledge, and big data sources to construct a knowledge graph related to telescope observation planning. By organically combining deep learning, knowledge graphs, and the large language model, this invention realizes the research on a telescope observation planning method based on a large language model.

[0032] This invention formulates the optimal observation plan based on the information characteristics of the telescope observation plan and the limitations of the telescope itself, thus scientifically maximizing the telescope's observation capabilities. By introducing technologies such as large language model intelligent algorithms and machine learning, it not only improves the efficiency, accuracy, and reliability of telescope observations but also optimizes resource utilization, enhances scientific research results, and protects the long-term operation and use of the telescope. This is of great significance for promoting the development of astronomy and other scientific fields. Attached Figure Description

[0033] Figure 1 This invention provides an overall roadmap for observation planning based on a large language model.

[0034] Figure 2 This is a flowchart illustrating the process of constructing a neural network model for telescope observation planning using deep learning in this embodiment.

[0035] Figure 3 This is a flowchart illustrating the process of constructing a knowledge graph for telescope observation planning in this embodiment. Detailed Implementation

[0036] The present invention will now be described in further detail with reference to the embodiments and the accompanying drawings.

[0037] This invention was supported by the National Natural Science Foundation of China projects “Intelligent Research on Control Systems of Large Astronomical Optical Telescopes” (U1931207), “Research on Intelligent Monitoring and Diagnostic Methods for Imaging Quality of Optical Astronomical Telescopes” (12203079), and “Research and Evaluation Methods for Unexpected State Evolution of Direct-Drive Systems of Extremely Large Telescopes under Extreme Environments” (11973065). It was also supported by the Jiangsu Provincial Natural Science Foundation project “Research on Key Technologies for Intelligent Optimization of Dome Seeing Based on LAMOST” (BK20221156). Furthermore, it was supported by the China Scholarship Council (CSC 201904910254) and the Overseas Scholarship Fund of the Chinese Academy of Sciences (Xu Lingzhe).

[0038] This invention combines artificial intelligence technology with the needs of telescope observation planning, leveraging the interactive advantages of large language models to formulate observation plans based on the information characteristics of the telescope observation plan and the limitations of the telescope itself. The telescope conducts observations according to the pre-determined plan, which not only improves the efficiency, accuracy, and reliability of telescope observations but also optimizes resource utilization, enhances scientific research results, and protects the long-term operation and use of the telescope. This is of great significance for promoting the development of astronomy and other scientific fields.

[0039] One embodiment of the present invention is a telescope observation planning method based on a large language model. For example... Figure 1 As shown, this is the overall roadmap for an intelligent optimization technique for observation planning based on a large language model.

[0040] This study utilizes big language technology in artificial intelligence to research the formulation of telescope observation plans. It combines deep learning to construct a neural network model for a telescope observation plan, and integrates expert knowledge, big language models, and big data sources to create a knowledge graph for telescope observation plans. Based on the big language model, it provides a theoretical foundation and technical support for the formulation of next-generation telescope observation plans.

[0041] The first step is to preprocess the telescope observation planning application and extract the features of the telescope observation planning information, including the telescope applicant information, the telescope observation application target, the telescope observation application time, and the scientific significance of the telescope observation.

[0042] The second step involves inputting the preprocessed features containing telescope observation planning information into a telescope observation planning knowledge graph composed of a large language model, a large data source, and expert knowledge, and then matching them according to the applicant's information.

[0043] The third step is to obtain more information about the applicant if a matching node exists in the telescope observation planning knowledge graph, including attributes such as partners, influence in the research field, whether previous observation applications were approved, and research results achieved.

[0044] The fourth step is to send the current application information to the large language model for further processing if there is no matching node in the telescope observation planning knowledge graph. This will obtain information such as the applicant's partners, the influence of the research field, whether previous observation applications were approved, and the scientific research results obtained. The telescope observation planning knowledge graph will then be updated accordingly.

[0045] The fifth step involves constructing a feature vector based on the output of the knowledge graph and the information from the observation application. This feature vector is then output to the trained neural network model, which determines whether the application is approved and the timeline.

[0046] The process of building a neural network model for telescope observation planning using deep learning is as follows: Figure 2 As shown. Includes:

[0047] Deep neural networks consist of multiple neurons and multiple layers. Each neuron is responsible for processing a portion of the input data and passing it to the next layer, thereby achieving a high-level abstraction and extraction of telescope observation planning information.

[0048] The first step is to build a deep learning model to collect telescope observation plans and preprocess the data, including outlier correction, missing value handling, and data normalization.

[0049] The second step is to divide the dataset into training dataset, validation dataset, and test dataset, and then save these datasets to the database.

[0050] The third step involves writing neural network code and building a neural network model based on the deep learning framework TensorFlow. The structure of this neural network model includes the number of neurons in the input layer, hidden layers, and output layer, as well as the activation function and loss function.

[0051] The fourth step uses the mean squared error as the loss function and stochastic gradient descent (SGD) as the optimization algorithm. Data is input into the neural network model, and the optimization algorithm continuously adjusts the model's parameters and minimizes the loss function, gradually bringing the model closer to the optimal solution.

[0052] The fifth step is to evaluate the model's performance using a validation set, including calculating metrics such as accuracy, recall, and F1-score. Based on the evaluation results, adjust the model's parameters and architecture, iterating continuously to optimize the model and bring its performance to its best.

[0053] The process of constructing the knowledge graph for telescope observation planning is as follows: Figure 3 As shown. Includes:

[0054] Based on Grakn, the relationships between the features of related entities are inferred by integrating expert knowledge, large language models and big data sources, and generated and stored in a knowledge graph.

[0055] The first step is to store telescope observation planning information in the knowledge graph database of the model library, including information such as observation priority, application time for observation, and observation targets;

[0056] The second step is to acquire relevant knowledge and construct a telescope observation map by interacting with expert knowledge and big data sources.

[0057] The third step is to improve the telescope's observation map through large language model interaction, such as intelligent information service sources like CHATGPT, BARD, and Wenxin Yiyan.

[0058] The fourth step is to perform text mining, text understanding and analysis on the responses from the large language model, generate structured feedback opinions and send them back to the telescope observation map library;

[0059] This invention discloses a telescope observation planning method based on a large language model. It organically combines multiple technologies, including deep learning, knowledge graphs, and large language models, to generate telescope observation plans. This includes constructing a neural network model based on deep learning to generate a telescope observation plan and building a knowledge graph related to telescope observations. The method also utilizes the large language model for analysis and processing to study the formulation of telescope observation plans.

[0060] This invention provides a telescope observation planning method based on a large language model. It combines artificial intelligence technology with the needs of telescope observation planning, leverages the interactive advantages of the large language model, uses deep learning algorithms to construct a neural network model, generates a knowledge graph by linking expert knowledge, the large language model, and big data sources, and analyzes a large number of observation tasks based on the large language model to find the optimal observation planning scheme.

[0061] This invention can scientifically maximize the observation capabilities of a telescope, allocate observation time for long-term observation tasks over the next few months, select valuable observation applications from numerous applications, and then formulate an observation plan based on the information characteristics of the telescope's observation plan and the limitations of the telescope itself; at the same time, it can formulate short-term plans and adjust the observation plan in case of sudden astronomical phenomena or telescope malfunctions.

[0062] Please note that not all activities or elements described in the general description above are essential, a particular activity or part of the apparatus may not be essential, and one or more further activities or included elements may be performed in addition to those described. Furthermore, the order in which the activities are listed does not necessarily represent the order in which they are performed. Moreover, these concepts have been described with reference to specific embodiments. However, those skilled in the art will recognize that various modifications and changes can be made without departing from the scope of this disclosure as set forth in the following claims. Therefore, the specification and drawings are to be considered illustrative rather than restrictive, and all such modifications are included within the scope of this disclosure.

[0063] The benefits, other advantages, and solutions to problems have been described above with respect to specific embodiments. However, any benefits, advantages, solutions to problems, and any features that may lead to or make any benefit, advantage, or solution more apparent should not be construed as critical, essential, or essential features of any or all claims. Furthermore, the specific embodiments disclosed above are merely illustrative, as the disclosed subject matter can be modified and implemented in different but equivalent ways that would be apparent to those skilled in the art benefiting from the teachings herein. There is no intention to limit the details of the constructions or designs shown herein other than those described in the claims. Therefore, it is apparent that the specific embodiments disclosed above can be altered or modified, and all such changes are considered to be within the scope of the disclosed subject matter.

Claims

1. A telescope observation planning method based on a large language model, characterized in that, include: Step 1: Construct a neural network model for telescope observation planning; specifically including the following steps: Step 1.1, Data Collection and Preprocessing: Collect telescope observation planning data to build a deep learning model, and then preprocess the data; divide the dataset into training dataset, validation dataset, and test dataset, and save these datasets to the database; Step 1.2, Neural Network Model Construction: Write neural network code and build a neural network model based on the deep learning framework TensorFlow; Step 1.3, Neural Network Model Training: The mean squared error is used as the loss function, and stochastic gradient descent (SGD) is used as the optimization algorithm. The telescope observation application data is input into the neural network model, and the model parameters are continuously adjusted and the loss function is minimized through the optimization algorithm, so that the model gradually approaches the optimal solution. Step 1.4, Model Evaluation and Tuning: Use the validation set to evaluate the model's performance; finally, based on the model evaluation results, adjust the model's parameters and architecture, and optimize the model through continuous iteration until the model's performance is optimal. Step 2: Construct a knowledge graph for telescope observation planning. This includes inferring relationships between entity information features based on Grakn, integrating expert knowledge, large language models, and big data sources, generating and storing the relationships in the knowledge graph. Specifically, this includes the following steps: Step 2.1: Construct a knowledge graph framework. The knowledge graph database in the model library stores telescope observation planning information. Step 2.2: Acquire the necessary knowledge through interaction with expert knowledge and big data sources to construct a telescope observation map; Step 2.3: Improve the telescope observation map through large language model interaction. Perform text mining, text understanding and analysis on the response content of the large language model, generate structured feedback opinions and send them back to the telescope observation map library. Step 2.4: Perform text mining, text understanding and analysis on the response content of the large language model, generate structured feedback opinions and send them back to the telescope observation planning knowledge graph database; Step 3: Implementation of Telescope Observation Planning Based on a Large Language Model: This includes constructing a knowledge graph of telescope observation application information using a large language model, and combining it with a deep learning model to achieve intelligent telescope observation planning; specifically, it includes the following steps: Step 3.1: Preprocess the telescope observation planning application and extract the telescope observation planning information features; the extracted telescope observation planning information features include telescope applicant information, telescope observation application objectives, telescope observation application time, telescope observation applicant's published articles, and the scientific significance of telescope observation; Step 3.2: Input the preprocessed features containing telescope observation planning information into a telescope observation planning knowledge graph composed of a large language model, a large data source, and expert knowledge; Step 3.3: The knowledge graph uses associated expert knowledge, large language models, and deep learning models to infer the relationships between the features of related entity information and stores them in the knowledge graph library; Step 3.4: Search and match the telescope observation planning feature attributes in the telescope observation planning knowledge graph to obtain the historical interaction data of the telescope observation planning; Step 3.5: If there are matching historical telescope observation plan feature attributes in the telescope observation plan knowledge graph, compare and evaluate the current telescope observation plan feature attributes with the previous observation plan feature attributes, and update the telescope observation plan feature attributes. Step 3.6: If there is no matching historical telescope observation plan feature attribute in the telescope observation plan knowledge graph, send the current telescope observation plan feature attribute to the large language model for further processing and update the telescope observation plan knowledge graph to obtain the updated telescope observation plan feature attribute. Step 3.7: Convert the updated telescope observation planning feature attribute information into feature vectors and input them into the telescope observation planning neural network model; wherein, the updated telescope observation planning feature attributes include telescope observation applicant, telescope observation application unit, telescope observation application target, number of telescope observation applications, and telescope observation application results; Step 3.8: Convert the telescope observation map into feature vectors and input them into the neural network model. Generate the telescope observation plan by combining the comprehensive matching results and success rate of the telescope observation planning neural network model.

2. The telescope observation planning method based on a large language model according to claim 1, characterized in that, In step 1.1, the data preprocessing includes outlier correction, missing value handling, and data normalization.

3. The telescope observation planning method based on a large language model according to claim 1, characterized in that, In step 1.2, the structure of the neural network model includes an input layer, a hidden layer, and an output layer. The output layer is configured with the number of neurons, an activation function, and a loss function.

4. The telescope observation planning method based on a large language model according to claim 1, characterized in that, In step 1.4, the performance of the model includes calculating accuracy, recall, and F1-score.

5. The telescope observation planning method based on a large language model according to claim 1, characterized in that, In step 2.1, the telescope observation planning information includes the observation priority, the application time for observation, and the observation target.

6. The telescope observation planning method based on a large language model according to claim 1, characterized in that, In step 2.3, the large language model includes CHATGPT, BARD, and Wenxin Yiyan.

Images