Crystal structure prediction method and device, electronic equipment and storage medium

Abstract

The invention discloses a crystal structure prediction method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring crystal structure initial parameters of a to-be-predicted material; determining the spatial position of each atom in the to-be-predicted material through a trained crystal structure prediction model according to the crystal structure initial parameters of the to-be-predicted material; wherein the crystal structure prediction model is a depth deterministic strategy gradient model trained according to state information and action information, the state information comprises crystal structure parameters of the crystal, and the action information comprises a spatial position adjustment strategy of each atom in the crystal. According to the technical scheme, crystal atom arrangement prediction is carried out through the depth deterministic strategy gradient model, and the accuracy of crystal structure prediction is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a crystal structure prediction method, device, electronic device and storage medium. Background Art

[0002] In recent years, the continued demand for new functional materials requires more effective strategies to accelerate the discovery of new materials, and crystal structure prediction is one of the basic tasks in this direction. As a long-standing challenge in condensed matter physics, crystal structure prediction aims to generate reasonable crystalline materials given elements and ratios.

[0003] In the existing technology, the most stable structure of a crystal is often predicted through models such as generative adversarial networks or graph neural networks, which involves optimizing the atomic arrangement of the crystal to minimize the energy of the crystal.

[0004] In the process of implementing the present invention, it is found that there are at least the following technical problems in the prior art: the existing crystal structure prediction scheme has the problem of low prediction accuracy. Summary of the invention

[0005] The present invention provides a crystal structure prediction method, device, electronic device and storage medium to improve the accuracy of crystal structure prediction.

[0006] According to one aspect of the present invention, a crystal structure prediction method is provided, comprising:

[0007] Obtaining initial crystal structure parameters of the material to be predicted;

[0008] Determine the spatial position of each atom in the material to be predicted according to the initial crystal structure parameters of the material to be predicted by the trained crystal structure prediction model;

[0009] Among them, the crystal structure prediction model is a deep deterministic policy gradient model trained according to state information and action information, the state information includes the crystal structure parameters of the crystal, and the action information includes the spatial position adjustment strategy of each atom in the crystal.

[0010] According to another aspect of the present invention, there is provided a crystal structure prediction device, comprising:

[0011] A crystal structure initial parameter acquisition module is used to obtain the crystal structure initial parameters of the material to be predicted;

[0012] A crystal atom spatial position prediction module, used to determine the spatial position of each atom in the material to be predicted according to the initial crystal structure parameters of the material to be predicted by using the trained crystal structure prediction model;

[0013] Among them, the crystal structure prediction model is a deep deterministic policy gradient model trained according to state information and action information, the state information includes the crystal structure parameters of the crystal, and the action information includes the spatial position adjustment strategy of each atom in the crystal.

[0014] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0015] at least one processor;

[0016] and a memory communicatively coupled to the at least one processor;

[0017] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the crystal structure prediction method described in any embodiment of the present invention.

[0018] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the crystal structure prediction method described in any embodiment of the present invention when executed.

[0019] The technical solution of the embodiment of the present invention obtains the initial parameters of the crystal structure of the material to be predicted, and then determines the spatial position of each atom in the material to be predicted based on the initial parameters of the crystal structure of the material to be predicted through the trained crystal structure prediction model; wherein the crystal structure prediction model is a deep deterministic policy gradient model trained based on state information and action information, the state information includes the crystal structure parameters of the crystal, and the action information includes the spatial position adjustment strategy of each atom in the crystal. The above technical solution predicts the arrangement of crystal atoms through the deep deterministic policy gradient model, which effectively improves the accuracy of crystal structure prediction.

[0020] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1is a flow chart of a crystal structure prediction method provided according to Embodiment 1 of the present invention;

[0023] Figure 2 is a flow chart of a crystal structure prediction method provided according to Embodiment 2 of the present invention;

[0024] Figure 3 is a schematic diagram of a crystal structure prediction model training provided according to an embodiment of the present invention;

[0025] Figure 4 is a schematic diagram of the structure of a crystal structure prediction device provided according to Embodiment 3 of the present invention;

[0026] Figure 5 It is a schematic diagram of the structure of an electronic device for implementing the crystal structure prediction method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0028] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. The acquisition, storage, use, processing, etc. of data in the technical solution of this application comply with the relevant provisions of national laws and regulations.

[0029] Embodiment 1

[0030] Figure 1This is a flow chart of a crystal structure prediction method provided in the first embodiment of the present invention. This embodiment is applicable to the case of predicting crystal structure through a reinforcement learning model. The method can be executed by a crystal structure prediction device. The crystal structure prediction device can be implemented in the form of hardware and / or software. The crystal structure prediction device can be configured in a terminal and / or server. Figure 1 As shown, the method includes:

[0031] S110, obtaining initial parameters of the crystal structure of the material to be predicted.

[0032] The material to be predicted refers to the crystal material to be predicted or searched for structure. The initial parameters of the crystal structure of the material to be predicted include at least one of the crystal space group, crystal constant, crystal axis angle, element type and element ratio. The crystal constants include a, b and c, and the crystal axis angles include α, β and γ.

[0033] S120, determining the spatial position of each atom in the material to be predicted according to the initial crystal structure parameters of the material to be predicted by the trained crystal structure prediction model.

[0034] Among them, the crystal structure prediction model is a deep deterministic policy gradient model (DDPG) trained based on state information and action information. The state information includes the crystal structure parameters of the crystal, and the action information includes the spatial position adjustment strategy of each atom in the crystal.

[0035] In an embodiment of the present invention, the deep deterministic policy gradient model is a deep reinforcement learning algorithm based on deterministic policy gradient. In reinforcement learning, state information refers to the state; in the crystal structure prediction scenario, state information refers to the crystal structure parameters of the crystal. In reinforcement learning, action information refers to the action selected by the agent from the continuous space under a given state; in the crystal structure prediction scenario, action information refers to the spatial position adjustment strategy of each atom in the crystal, that is, the change of the atomic position in the crystal.

[0036] The embodiment of the present invention selects a deep deterministic policy gradient algorithm to implement the training of the crystal structure prediction model. Specifically, the value (Critic) network takes state information and action information as input and outputs the expected cumulative reward (Q value). The strategy (Actor) network takes state information as input and action information as output.

[0037] The technical solution of the embodiment of the present invention obtains the initial parameters of the crystal structure of the material to be predicted, and then determines the spatial position of each atom in the material to be predicted based on the initial parameters of the crystal structure of the material to be predicted through the trained crystal structure prediction model; wherein the crystal structure prediction model is a deep deterministic policy gradient model trained based on state information and action information, the state information includes the crystal structure parameters of the crystal, and the action information includes the spatial position adjustment strategy of each atom in the crystal. The above technical solution predicts the arrangement of crystal atoms through the deep deterministic policy gradient model, which effectively improves the accuracy of crystal structure prediction.

[0038] Embodiment 2

[0039] Figure 2 A flow chart of a crystal structure prediction method provided in the second embodiment of the present invention, the method of this embodiment can be combined with each optional scheme in the crystal structure prediction method provided in the above embodiment. The crystal structure prediction method provided in this embodiment is further optimized. Optionally, before obtaining the initial parameters of the crystal structure of the material to be predicted, it also includes: inputting the current crystal structure parameters of the crystal into the online strategy network to obtain the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters; adding noise to the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters, and obtaining the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise; inputting the current crystal structure parameters of the crystal and the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise into the online value network to obtain the current expected cumulative reward; updating the parameters of the online strategy network according to the current expected cumulative reward through the strategy gradient, updating the parameters of the online value network according to the current expected cumulative reward through the preset loss function, and updating the target network based on the parameters of the updated online strategy network and the parameters of the updated online value network, until the model training stop condition is met, and a trained crystal structure prediction model is obtained.

[0040] like Figure 2 As shown, the method includes:

[0041] S210, inputting the current crystal structure parameters of the crystal into the online strategy network to obtain the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters.

[0042] Among them, the current crystal structure parameters can be used Indicates that the current state, the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters can be used It represents the action selected in the current state. The online policy network refers to the policy network updated by online training, which can be a multi-layer perceptron, whose input is the current crystal structure parameters, and whose output is the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters.

[0043] Optionally, the current crystal structure parameters of the crystal are input into the online strategy network to obtain the spatial position adjustment strategy for each atom in the crystal under the current crystal structure parameters, including: determining the current hidden state information based on the current crystal structure parameters and the target state of the crystal; determining the attention weight vector of the current action information based on the current hidden state information; determining the spatial position adjustment strategy for each atom in the crystal under the current crystal structure parameters based on the attention weight vector of the current hidden state information and the current action information.

[0044] It is understandable that when the current state of the crystal is close to the target state, the network needs to be fine-tuned. On the contrary, when the current state of the crystal deviates from the target state, a larger adjustment is required. This requires that the policy function of the online policy network can adjust the focus and amplitude at each step according to the specific task objectives. Therefore, the embodiment of the present invention redesigns the online policy network by deriving a hidden state, as shown in the following formula:

[0045] ;

[0046] ;

[0047] ;

[0048] in, represents the current crystal structure parameters of the crystal, Indicates the current hidden state information. The attention weight vector representing the current action information, represents the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters, W represents the weight of the online strategy network, and b represents the bias of the online strategy network.

[0049] S220, adding noise to the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters to obtain the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise.

[0050] It should be noted that the embodiment of the present invention adopts off-policy training, that is, during the training process, the strategy for generating behavior (i.e., behavior strategy) is different from the strategy for evaluating agent performance (i.e., target strategy). Specifically, the action of the agent at a certain moment is not generated by a deterministic strategy. In order to ensure sufficient exploration, the embodiment of the present invention introduces exploration noise in the action selection process, as shown in the following formula:

[0051] ;

[0052] ;

[0053] in, represents the exploration noise, represents the current crystal structure parameters of the crystal, represents the target state, represents the parameters of the online policy network, It represents the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters. Represents the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise.

[0054] S230, inputting the current crystal structure parameters of the crystal and the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise into the online value network to obtain the current expected cumulative reward.

[0055] The online value network refers to a value network updated by online training, which may be a multi-layer perceptron.

[0056] For example, the current crystal structure parameters of the crystal and the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise are spliced ​​and input into the multi-layer perceptron to obtain the current expected cumulative reward, as shown in the following formula:

[0057] ;

[0058] in, represents the parameters of the online value network, Represents the current expected cumulative reward.

[0059] S240, updating the parameters of the online policy network according to the current expected cumulative reward through policy gradient, updating the parameters of the online value network according to the current expected cumulative reward through a preset loss function, updating the target network based on the updated parameters of the online policy network and the updated parameters of the online value network, until the model training stop condition is met, and obtaining a trained crystal structure prediction model.

[0060] Among them, the target network is used to ensure the convergence of network parameters.

[0061] Optionally, the parameter update formula of the online policy network is:

[0062] ;

[0063] in, represents the parameters of the online policy network, represents the online policy network learning rate, represents the gradient calculated by the online value network, represents the gradient calculated by the online policy network, Indicates status information. Indicates action information. represents the parameters of the online value network, represents the current crystal structure parameters of the crystal, It represents the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise, and N represents the number of iterations.

[0064] Optionally, the parameter update formula of the online value network is:

[0065] ;

[0066] ;

[0067] ;

[0068] in, represents the parameters of the online value network, represents the online value network learning rate, represents the gradient calculated by the preset loss function, represents the preset loss function, represents the current crystal structure parameters of the crystal, represents the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise, represents the current expected cumulative reward, Indicates the target value corresponding to the current crystal structure parameter, Indicates the reward corresponding to the current crystal structure parameters, Indicates timely reward, represents the next crystal structure parameter of the crystal, represents the parameters of the target policy network, represents the spatial position adjustment strategy of each atom in the crystal under the next crystal structure parameters, represents the parameters of the target value network, represents the next expected cumulative reward, and N represents the number of iterations.

[0069] Optionally, the target network includes a target policy network and a target value network, and the formula for updating the target network is:

[0070] ;

[0071] ;

[0072] in, represents the parameters of the online value network, represents the parameters of the target value network, represents the parameters of the online policy network, represents the parameters of the target policy network, represents the update coefficient.

[0073] For example, Figure 3 is a schematic diagram of a crystal structure prediction model training provided according to an embodiment of the present invention. Specifically, the environment can be a crystal structure screening system. When interacting with the environment, the agent selects and executes an action according to the current state. After executing the action, the environment transitions to a new state and provides a reward signal to the agent. The agent updates the decision-making process through feedback and repeatedly selects subsequent actions until the maximum expected cumulative reward is achieved. Further, N-tuples are sampled from the experience replay pool. , the goal is to optimize To maximize the expected cumulative reward:

[0074] ;

[0075] Furthermore, the policy gradient is used to adjust the parameters of the online policy network Update to maximize the expected cumulative reward. The parameters of the online value network are adjusted by presetting the loss function. The goal of the online value network is to minimize the mean square error loss of the expected cumulative reward. The target value network and the target policy network are updated using a soft update method to avoid performance fluctuations or instability caused by excessive updates. It should be noted that the embodiment of the present invention introduces post-experience replay through the experience replay pool, which improves sample efficiency and accelerates learning.

[0076] S250, obtaining initial parameters of the crystal structure of the material to be predicted.

[0077] S260, determining the spatial position of each atom in the material to be predicted according to the initial crystal structure parameters of the material to be predicted by the trained crystal structure prediction model.

[0078] The technical solution of the embodiment of the present invention is to input the current crystal structure parameters of the crystal into the online policy network to obtain the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters; add noise to the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters to obtain the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise; input the current crystal structure parameters of the crystal and the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise into the online value network to obtain the current expected cumulative reward; update the parameters of the online policy network according to the current expected cumulative reward through the policy gradient, update the parameters of the online value network according to the current expected cumulative reward through the preset loss function, update the target network based on the parameters of the updated online policy network and the parameters of the updated online value network, until the model training stop condition is met, and obtain a trained crystal structure prediction model. The above technical solution trains the crystal structure prediction model through a deep deterministic policy gradient algorithm, which effectively improves the prediction accuracy of the crystal structure prediction model.

[0079] Embodiment 3

[0080] Figure 4 This is a schematic diagram of the structure of a crystal structure prediction device provided in Embodiment 3 of the present invention. Figure 4 As shown, the device comprises:

[0081] A crystal structure initial parameter acquisition module 310 is used to acquire the crystal structure initial parameters of the material to be predicted;

[0082] A crystal atom spatial position prediction module 320 is used to determine the spatial position of each atom in the material to be predicted according to the initial crystal structure parameters of the material to be predicted by using the trained crystal structure prediction model;

[0083] Among them, the crystal structure prediction model is a deep deterministic policy gradient model trained according to state information and action information, the state information includes the crystal structure parameters of the crystal, and the action information includes the spatial position adjustment strategy of each atom in the crystal.

[0084] The technical solution of the embodiment of the present invention obtains the initial parameters of the crystal structure of the material to be predicted, and then determines the spatial position of each atom in the material to be predicted based on the initial parameters of the crystal structure of the material to be predicted through the trained crystal structure prediction model; wherein the crystal structure prediction model is a deep deterministic policy gradient model trained based on state information and action information, the state information includes the crystal structure parameters of the crystal, and the action information includes the spatial position adjustment strategy of each atom in the crystal. The above technical solution predicts the arrangement of crystal atoms through the deep deterministic policy gradient model, which effectively improves the accuracy of crystal structure prediction.

[0085] In some optional embodiments, the crystal structure prediction apparatus further comprises:

[0086] An online strategy network prediction module is used to input the current crystal structure parameters of the crystal into the online strategy network to obtain a spatial position adjustment strategy for each atom in the crystal under the current crystal structure parameters;

[0087] A noise adding module is used to add noise to the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters to obtain the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise;

[0088] An online value network prediction module, used for inputting the current crystal structure parameters of the crystal and the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise into the online value network to obtain the current expected cumulative reward;

[0089] A network update module is used to update the parameters of the online policy network according to the current expected cumulative reward through policy gradient, update the parameters of the online value network according to the current expected cumulative reward through a preset loss function, and update the target network based on the updated parameters of the online policy network and the updated parameters of the online value network until the model training stop condition is met, thereby obtaining a trained crystal structure prediction model.

[0090] In some optional implementations, the online policy network prediction module may be specifically used for:

[0091] determining current hidden state information based on current crystal structure parameters of the crystal;

[0092] Determine an attention weight vector of current action information based on the current hidden state information;

[0093] Based on the attention weight vector of the current hidden state information and the current action information, a spatial position adjustment strategy for each atom in the crystal under the current crystal structure parameters is determined.

[0094] In some optional implementations, the parameter update formula of the online policy network is:

[0095] ;

[0096] in, represents the parameters of the online policy network, represents the online policy network learning rate, represents the gradient calculated by the online value network, represents the gradient calculated by the online policy network, Indicates status information. Indicates action information. represents the parameters of the online value network, represents the current crystal structure parameters of the crystal, It represents the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise, and N represents the number of iterations.

[0097] In some optional implementations, the parameter update formula of the online value network is:

[0098] ;

[0099] ;

[0100] ;

[0101] in, represents the parameters of the online value network, represents the online value network learning rate, represents the gradient calculated by the preset loss function, represents the preset loss function, represents the current crystal structure parameters of the crystal, represents the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise, represents the current expected cumulative reward, Indicates the target value corresponding to the current crystal structure parameter, Indicates the reward corresponding to the current crystal structure parameters, Indicates timely reward, represents the next crystal structure parameter of the crystal, represents the parameters of the target policy network, represents the spatial position adjustment strategy of each atom in the crystal under the next crystal structure parameters, represents the parameters of the target value network, represents the next expected cumulative reward, and N represents the number of iterations.

[0102] In some optional implementations, the target network includes a target policy network and a target value network, and the formula for updating the target network is:

[0103] ;

[0104] ;

[0105] in, represents the parameters of the online value network, represents the parameters of the target value network, represents the parameters of the online policy network, represents the parameters of the target policy network, represents the update coefficient.

[0106] In some optional embodiments, the initial crystal structure parameters of the material to be predicted include at least one of a crystal space group, a crystal constant, a crystal axis angle, an element type, and an element ratio.

[0107] The crystal structure prediction device provided in the embodiment of the present invention can execute the crystal structure prediction method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0108] Embodiment 4

[0109] Figure 5 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0110] like Figure 5 As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. The I / O interface 15 is also connected to the bus 14.

[0111] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0112] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 performs the various methods and processes described above, such as a crystal structure prediction method, which includes:

[0113] Obtaining initial crystal structure parameters of the material to be predicted;

[0114] Determine the spatial position of each atom in the material to be predicted according to the initial crystal structure parameters of the material to be predicted by the trained crystal structure prediction model;

[0115] Among them, the crystal structure prediction model is a deep deterministic policy gradient model trained according to state information and action information, the state information includes the crystal structure parameters of the crystal, and the action information includes the spatial position adjustment strategy of each atom in the crystal.

[0116] In some embodiments, the crystal structure prediction method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the crystal structure prediction method described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to perform the crystal structure prediction method in any other appropriate manner (e.g., by means of firmware).

[0117] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0118] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0119] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, device, or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0120] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0121] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0122] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0123] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0124] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A crystal structure prediction method, characterized in that: include: Obtaining initial crystal structure parameters of the material to be predicted; Determine the spatial position of each atom in the material to be predicted according to the initial crystal structure parameters of the material to be predicted by the trained crystal structure prediction model; Among them, the crystal structure prediction model is a deep deterministic policy gradient model trained according to state information and action information, the state information includes the crystal structure parameters of the crystal, and the action information includes the spatial position adjustment strategy of each atom in the crystal.

2. The method according to claim 1, characterized in that Before obtaining the initial parameters of the crystal structure of the material to be predicted, the method further includes: Inputting the current crystal structure parameters of the crystal into the online strategy network to obtain the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters; Adding noise to the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters to obtain the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise; Inputting the current crystal structure parameters of the crystal and the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise into the online value network to obtain the current expected cumulative reward; The parameters of the online policy network are updated according to the current expected cumulative reward through policy gradient, the parameters of the online value network are updated according to the current expected cumulative reward through a preset loss function, and the target network is updated based on the updated parameters of the online policy network and the updated parameters of the online value network until the model training stop condition is met, thereby obtaining a trained crystal structure prediction model.

3. The method according to claim 2, characterized in that The step of inputting the current crystal structure parameters of the crystal into the online strategy network to obtain the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters comprises: Determining current hidden state information based on current crystal structure parameters and a target state of the crystal; Determine an attention weight vector of current action information based on the current hidden state information; Based on the attention weight vector of the current hidden state information and the current action information, a spatial position adjustment strategy for each atom in the crystal under the current crystal structure parameters is determined.

4. The method according to claim 2, characterized in that: The parameter update formula of the online strategy network is: ； in, represents the parameters of the online policy network, represents the online policy network learning rate, represents the gradient calculated by the online value network, represents the gradient calculated by the online policy network, Indicates status information. Indicates action information. represents the parameters of the online value network, represents the current crystal structure parameters of the crystal, It represents the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise, and N represents the number of iterations.

5. The method according to claim 2, characterized in that: The parameter update formula of the online value network is: ； ； ； in, represents the parameters of the online value network, represents the online value network learning rate, represents the gradient calculated by the preset loss function, represents the preset loss function, represents the current crystal structure parameters of the crystal, represents the spatial position adjustment strategy of each atom in the crystal under the current crystal structure parameters after adding noise, represents the current expected cumulative reward, Indicates the target value corresponding to the current crystal structure parameter, Indicates the reward corresponding to the current crystal structure parameters, Indicates timely reward, represents the next crystal structure parameter of the crystal, represents the parameters of the target policy network, represents the spatial position adjustment strategy of each atom in the crystal under the next crystal structure parameters, represents the parameters of the target value network, represents the next expected cumulative reward, and N represents the number of iterations.

6. The method according to claim 2, characterized in that The target network includes a target strategy network and a target value network. The formula for updating the target network is: ； ； in, represents the parameters of the online value network, represents the parameters of the target value network, represents the parameters of the online policy network, represents the parameters of the target policy network, represents the update coefficient.

7. The method according to any one of claims 1 to 6, characterized in that: The initial crystal structure parameters of the material to be predicted include at least one of a crystal space group, a crystal constant, a crystal axis angle, an element type and an element ratio.

8. A crystal structure prediction device, characterized in that: include: A crystal structure initial parameter acquisition module is used to obtain the crystal structure initial parameters of the material to be predicted; A crystal atom spatial position prediction module, used to determine the spatial position of each atom in the material to be predicted according to the initial crystal structure parameters of the material to be predicted by using the trained crystal structure prediction model; Among them, the crystal structure prediction model is a deep deterministic policy gradient model trained according to state information and action information, the state information includes the crystal structure parameters of the crystal, and the action information includes the spatial position adjustment strategy of each atom in the crystal.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively coupled to the at least one processor; The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the crystal structure prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the crystal structure prediction method according to any one of claims 1 to 7 when executed.

Images