System and method for predicting physical properties

Abstract

A system for predicting physical properties of a material of the present invention may: extract a graph embedding by inputting material information to a first AI model; extract a text embedding by inputting a text description of a crystal structure of the material to a second AI model; divide the text embedding into a plurality of structural information embeddings; and combine the text embedding with at least one of the graph embedding or the plurality of structural information embeddings, wherein the structural information embeddings may be divided into global information, semi-global information, and local information about the crystal structure.

Description

System and method for predicting physical properties

[0001] The present invention relates to a system and method for predicting material properties, and more particularly, to a system and method for predicting material properties by separating interactions between atoms constituting a material into multiple scales and processing and integrating each piece of information.

[0002] Artificial intelligence (AI) technology has recently been garnering attention across society as it demonstrates cutting-edge developments. AI encompasses "a computer brain capable of executing tasks previously reserved for human intelligence," "the engineering and science of creating intelligent machines," and "a set of algorithmic systems designed to think, perceive, and act like humans," enabling computers to perform highly advanced, human-like intellectual abilities.

[0003] AI, combined with augmented reality, the Internet of Things, edge computing, and digital twins, is being touted as a key new technology that will lead the Fourth Industrial Revolution, promising highly integrated smart spaces. Furthermore, AI is gaining attention as a next-generation growth engine capable of evolving industrial ecosystems beyond simply solving standardized problems. It is being actively applied across knowledge service industries, including IT, healthcare, agriculture, energy, automobiles, and robotics, as well as distribution, finance, law, education, real estate, advertising, and communications. In other words, AI is integrating with all existing systems, from those that seek to improve the convenience and quality of daily life to the broader cultural and artistic landscape of our society, preparing for a new era.

[0004] With the recent diversification of product development methods, the development of new materials for use in product manufacturing is actively underway. These materials significantly impact product characteristics, and their physical properties can even determine the characteristics of the final product. Therefore, predicting and analyzing the properties of these materials is essential for more efficient development and mass production.

[0005] Traditionally, to determine the properties of a product based on the material's characteristics, multiple materials were developed, each with its own characteristics verified, and then tested on the final product to confirm the properties. However, this conventional method required significant costs and time for material and material characteristic development, and it was difficult to identify the material with optimal properties.

[0006] To improve upon these traditional methods, active research is underway on methods for predicting material properties using artificial intelligence (Republic of Korea Patent Publication No. 10-2024-0011349).

[0007] One embodiment of the present invention aims to provide a material property prediction system and method that divides interatomic interactions into multi-scales of local, semi-global, and global, and processes information in a manner optimized for each scale to maximize the accuracy of material property prediction.

[0008] A system for predicting the properties of a material according to one embodiment of the present invention comprises: at least one processor; and at least one memory for storing commands, information, or artificial intelligence models executed by the at least one processor; wherein the commands, information, or artificial intelligence models executed by the at least one processor may include: a preprocessing module for receiving atomic structure data of a material and generating first input data for a local interaction operation, second input data for a quasi-global interaction operation, and third input data for a global interaction operation; a local interaction operation unit for calculating local interactions between a specific atom and its surrounding neighboring atoms based on the first input data to extract local interaction features; a quasi-global interaction operation unit for calculating interactions between all atoms in a unit cell of a crystal based on the second input data to extract quasi-global interaction features; and a global interaction operation unit for calculating long-range interactions considering periodicity between unit cells based on the third input data to extract global interaction features.

[0009] Here, the local interaction operation unit includes a graph neural network, and the first input data input to the graph neural network may be a graph structure in which atoms of the material are nodes and pairs of atoms located within a specific cutting radius of the material are edges.

[0010] In addition, the quasi-global interaction operation unit may include a transformer, and the second input data input to the transformer may include an input sequence that sequentially lists feature vectors of atoms in the unit cell.

[0011] In addition, the quasi-global interaction operation unit can extract the quasi-global interaction features by using the relative distance matrix between all atom pairs within the unit cell as an attention bias of the self-attention operation.

[0012] Additionally, the global interaction operation unit may include another graph neural network, and the third input data input to the other graph neural network may be an extended graph structure that extends beyond the boundary of the unit cell and connects to neighboring atoms.

[0013] Additionally, the system further includes a property prediction unit, wherein the property prediction unit can generate an integrated feature vector by integrating feature vectors of each of the local interaction feature, the semi-global interaction feature, and the global interaction feature.

[0014] In addition, the above-mentioned property prediction unit can predict target properties by inputting the integrated feature vector into a multilayer perceptron.

[0015] Additionally, the preprocessing module can receive a crystallographic information file (CIF) and obtain the atomic structure data.

[0016] According to one embodiment of the present invention, a computerized method for predicting the properties of a material may include the steps of: receiving a crystallographic information file (CIF) as atomic structure data of the material, and generating input data including first input data for a local interaction calculation, second input data for a quasi-global interaction calculation, and third input data for a global interaction calculation; calculating local interactions between a specific atom and its surrounding neighboring atoms based on the first input data to extract local interaction features; calculating interactions between all atoms in a unit cell of a crystal based on the second input data to extract quasi-global interaction features; calculating long-range interactions considering periodicity between unit cells based on the third input data to extract global interaction features; and predicting target properties of the material by integrating the local interaction features, the quasi-global interaction features, and the global interaction features.

[0017] Here, the input data generation step can generate the first input data in a graph structure in which atoms of the material are nodes and pairs of atoms within a specific cutting radius are edges through a preprocessing module.

[0018] In addition, the local interaction feature extraction step can extract the local interaction feature through a message passing operation of a graph neural network for the first input data.

[0019] In addition, the input data generation step can generate the second input data including an input sequential sequence listing feature vectors of atoms in the unit cell in order and a relative distance matrix calculating distances between all atom pairs.

[0020] In addition, the semi-global interaction feature extraction step extracts the semi-global interaction feature by applying the self-attention mechanism of the transformer architecture to the second input data, and the relative distance matrix can be used as an attention bias.

[0021] Additionally, the input data generation step can generate the third input data in an extended graph structure that is connected to neighboring atoms beyond the boundary of the unit cell.

[0022] In addition, the global interaction feature extraction step can extract the global interaction features including interactions between unit cells by applying a graph neural network to the extended graph structure of the third input data.

[0023] In addition, the target property prediction step can predict the target property by combining vectors representing the local interaction feature, the semi-global interaction feature, and the global interaction feature, respectively, to generate an integrated feature vector, and passing the integrated feature vector through a fully connected neural network.

[0024] A service system for predicting the properties of a material according to one embodiment of the present invention may include a server computing system including: a user computing device that transmits an analysis request for atomic structure data of a material and receives a predicted property value; and a preprocessing module that is communicatively connected to the user computing device and receives the atomic structure data to generate first input data for a local interaction operation, second input data for a quasi-global interaction operation, and third input data for a global interaction operation; a multi-scale interaction operation unit that extracts local interaction features based on the first input data, quasi-global interaction features based on the second input data, and global interaction features based on the third input data; and a property prediction unit that predicts a target property of the material by integrating each of the extracted interaction features.

[0025] Here, the multi-scale interaction operation unit may include a local interaction operation unit that applies a message passing operation of a graph neural network to the first input data to extract the local interaction features; a quasi-global interaction operation unit that applies a self-attention of a transformer to the second input data to extract the quasi-global interaction features; and a global interaction operation unit that applies a message passing operation of a graph neural network to the third input data to extract the global interaction features.

[0026] In addition, the property prediction unit can predict the target property by combining each feature vector representing the local, semi-global, and global interaction features to generate an integrated feature vector and inputting the integrated feature vector into a multilayer perceptron.

[0027] In an application-specific integrated circuit for predicting the properties of a material, the application-specific integrated circuit comprises a memory in which information, commands, and an artificial intelligence model are stored, and a functional block including at least one processor requesting access to the memory, wherein the memory receives a crystallographic information file as atomic structure data of the material and generates input data including first input data for a local interaction operation, second input data for a quasi-global interaction operation, and third input data for a global interaction operation, and extracts local interaction features by calculating local interactions between specific atoms and their surrounding neighboring atoms based on the first input data, and extracts quasi-global interaction features by calculating interactions between all atoms in a unit cell of the crystal based on the second input data, and extracts global interaction features by calculating long-range interactions considering periodicity between unit cells based on the third input data, and an artificial intelligence model, commands, or information that predicts target properties of the material by integrating the local interaction features, the quasi-global interaction features, and the global interaction features may be stored.

[0028] According to one embodiment of the present invention, by separating and learning interatomic interactions within a material into three physical scales—local, quasi-global, and global—it is possible to comprehensively consider multi-scale physical phenomena that are difficult to capture with a single model, thereby enabling more accurate and reliable property predictions. Furthermore, by individually applying architectures that best capture the characteristics of each scale and synthesizing the results, the accuracy of material property predictions can be dramatically improved compared to existing methods, contributing to significant time and cost savings in the new material development process.

[0029] Figure 1 is a schematic diagram of an electronic device according to one embodiment of the present invention.

[0030] Figure 2 is a schematic diagram of a property prediction model of one embodiment of the present invention.

[0031] Figure 3 is a schematic diagram of a property prediction method of one embodiment of the present invention.

[0032] FIG. 4 is a schematic diagram illustrating the concepts of local, quasi-global, global interactions and unit cells of one embodiment of the present invention.

[0033] Figure 5 is a schematic diagram of a material property prediction service system utilizing a material property prediction model of one embodiment of the present invention.

[0034] To clarify the technical idea of ​​the present disclosure, embodiments of the present disclosure will be described in detail with reference to the attached drawings. In describing the present disclosure, if a detailed description of a related known function or component is determined to unnecessarily obscure the gist of the present disclosure, the detailed description will be omitted. In the drawings, components having substantially the same functional configuration are given the same reference numbers and symbols as possible even if they are shown in different drawings. For convenience of explanation, devices and methods are described together when necessary. Each operation of the present disclosure does not necessarily have to be performed in the described order and may be performed in parallel, selectively, or individually.

[0035] The terms used in the embodiments of this disclosure have been selected from widely used, current terms, taking into account the functions of the present disclosure. However, these terms may vary depending on the intentions of those skilled in the art, precedents, the emergence of new technologies, etc. Furthermore, in certain cases, the applicant may arbitrarily select terms, and in such cases, their meanings will be described in detail in the description of the relevant embodiments. Therefore, the terms used in this specification should not be defined simply as names of terms, but rather based on their meanings and the overall content of the present disclosure.

[0036] Throughout this disclosure, singular expressions may include plural expressions unless the context clearly dictates otherwise. Terms such as "comprise" or "have" should be understood to indicate the presence of a feature, number, step, operation, component, part, or combination thereof, but do not preclude the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof. In other words, when it is said throughout this disclosure that a part "comprises" a certain component, unless specifically stated otherwise, this does not mean that other components may be included, but rather that other components may be excluded.

[0037] Expressions such as "at least one" modify the entire list of elements, not individual elements of the list. For example, "at least one of A, B, and C" and "at least one of A, B, or C" refer to only A, only B, only C, both A and B, both B and C, both A and C, all of A, B, and C, or any combination thereof.

[0038] In addition, terms such as “...unit”, “...module”, etc. described in the present disclosure mean a unit that processes at least one function or operation, which may be implemented as hardware or software, or a combination of hardware and software.

[0039] Throughout this disclosure, when a part is said to be "connected" to another part, this includes not only cases where the parts are "directly connected," but also cases where the parts are "electrically connected" with other elements intervening. Furthermore, when a part is said to "include" a component, this does not exclude other components, but rather includes other components, unless otherwise specifically stated.

[0040] The expression “configured to” as used throughout this disclosure can be used interchangeably with, for example, “suitable for,” “having the capacity to,” “designed to,” “adapted to,” “made to,” or “capable of.” The term “configured to” does not necessarily mean something that is “specifically designed to” in terms of hardware. Instead, in some contexts, the expression “a system configured to” can mean that the system, together with other devices or components, is “capable of.” For example, the phrase “a processor configured (or set) to perform A, B, and C” may mean a dedicated processor (e.g., an embedded processor) for performing those operations, or a generic-purpose processor (e.g., a CPU or application processor) that can perform those operations by executing one or more software programs stored in memory.

[0041] Artificial intelligence (AI) is a field of computer engineering and information technology that studies how to enable computers to perform human-like tasks, such as thinking, learning, and self-improvement. It aims to enable computers to mimic human intelligent behavior. Furthermore, AI does not exist in isolation; rather, it is closely linked, both directly and indirectly, to other fields of computer science. In particular, efforts are actively underway to incorporate AI elements into various fields of information technology and utilize them to solve problems in those fields.

[0042] Machine learning is a branch of artificial intelligence that empowers computers to learn without explicit programming. Specifically, machine learning is the study and development of algorithms and systems that learn from empirical data, make predictions, and improve their own performance. Rather than executing strictly defined, static program instructions, machine learning algorithms build specific models based on input data to derive predictions or decisions. The term "machine learning" can be used interchangeably with "machine learning."

[0043] Many machine learning algorithms have been developed to classify data. Representative examples include decision trees, Bayesian networks, support vector machines (SVMs), and artificial neural networks (ANNs). Decision trees are an analytical method that performs classification and prediction by diagramming decision rules in a tree-like structure. Bayesian networks are models that represent probabilistic relationships (conditional independence) between multiple variables in a graph structure. Bayesian networks are suitable for data mining through unsupervised learning. Support vector machines are supervised learning models for pattern recognition and data analysis, primarily used for classification and regression analysis. Artificial neural networks model the operating principles and interconnected relationships of biological neurons. They are information processing systems in which numerous neurons, called nodes or processing elements, are connected in layers.

[0044] An artificial neural network (ANN) is a model used in machine learning. It is a statistical learning algorithm inspired by biological neural networks (especially the brain, the central nervous system of animals) in machine learning and cognitive science. Specifically, an ANN can refer to a general model in which artificial neurons (nodes) form a network through the connection of synapses, which change the strength of the synaptic connections through learning, thereby achieving problem-solving capabilities. The term "ANN" can be used interchangeably with the term "neural network."

[0045] An artificial neural network can include multiple layers, each of which can include multiple neurons. Furthermore, an artificial neural network can include synapses, which connect neurons. An artificial neural network can generally be defined by three factors: a) the connection pattern between neurons in different layers, b) a learning process that updates the weights of the connections, and c) an activation function that generates an output value from a weighted sum of the inputs received from the previous layer.

[0046] Artificial neural networks may include, but are not limited to, network models such as Deep Neural Networks (DNNs), Recurrent Neural Networks (RNNs), Bidirectional Recurrent Deep Neural Networks (BRDNNs), Multilayer Perceptrons (MLPs), and Convolutional Neural Networks (CNNs). In this specification, the term "layer" may be used interchangeably with the term "layer."

[0047] Artificial neural networks are categorized into single-layer neural networks and multi-layer neural networks based on the number of layers. A typical single-layer neural network consists of an input layer and an output layer. A typical multi-layer neural network consists of an input layer, one or more hidden layers, and an output layer.

[0048] The input layer is the layer that receives external data, and the number of neurons in the input layer is the same as the number of input variables, and the hidden layer is located between the input layer and the output layer. It receives signals from the input layer, extracts characteristics, and transmits them to the output layer. The output layer receives signals from the hidden layer and outputs output values ​​based on the received signals. The input signals between neurons are multiplied by each connection strength (weight) and then added, and if this sum is greater than the threshold of the neuron, the neuron is activated and outputs the output value obtained through the activation function.

[0049] Meanwhile, deep neural networks, which include multiple hidden layers between the input and output layers, are representative artificial neural networks that implement deep learning, a type of machine learning technique. The term "deep learning" can be used interchangeably with "deep learning," and the term "learning" can be used interchangeably with "training."

[0050] The workflow of machine learning consists of a series of steps: collecting data for learning and validation, modeling, and then training the model. This can include the processes of collecting training data, inspecting and exploring the data, preprocessing and cleaning the data, modeling, and training.

[0051] 1. Collect training data

[0052] The training data applied to the learning model of this specification can be generated using data collected from multiple samples. In this specification, at least one or more different types of training data sets can be used to train the learning model, and each training data set can further include one or more experimental results used as feature labels. At least a portion of the training data set can be used to train the learning model, and another portion can be used to validate the learned learning model.

[0053] The data used in the graph model of one embodiment of the present invention may be expressed in a graph format represented by nodes and edges, and may be represented as an adjacency matrix or an adjacency list. The adjacency matrix and adjacency list represent the connection relationships of the graph as a two-dimensional array and a list, respectively.

[0054] The data used for text embedding may include a text description of the crystal structure. In one embodiment, the text description data may be obtained through the Robocrystallographer package, which generates text descriptions similar to how a real crystallographer would analyze a structure. The Robocrystallographer package indicates symmetry, local environment, and extended connectivity when generating text descriptions of crystal structures, and the package may include utilities for identifying molecule names, component orientations, and heterostructure information. Alternatively, the data used for text embedding may be derived from a CIF file.

[0055] 2. Checking and exploring data

[0056] Once training data for learning a learning model is collected, the collected training data can be inspected and explored for data structure, noisy data, and data cleaning methods for applying machine learning.

[0057] This data review and exploration phase is called Exploratory Data Analysis (EDA), and EDA can be defined as the process of observing and understanding the collected data from various perspectives. Before data learning, visualizations such as graphs and statistical tests are used to examine independent and dependent variables, variable types, and their data types, allowing for preliminary identification of data characteristics and inherent structural relationships. Through EDA, data distribution and values ​​can be examined to better understand the phenomena expressed by the data and identify potential problems. Furthermore, through the process of examining data from various perspectives, various patterns that might not have been detected during the problem definition phase can be discovered, allowing for modification of existing hypotheses or the development of new ones. Exploratory data analysis can broadly include the process of searching for outliers and analyzing the relationships between data attributes.

[0058] The process of detecting outliers involves determining whether data contains outliers. This process can involve sampling, statistical methods, and visualization methods. Sampling methods extract a random sample from the data to identify overall trends and anomalies in the data values. Statistical methods can utilize summary statistics such as the mean, median, and mode to determine the center of the data, or the range and variance to determine the distribution of the data. Visualization methods can utilize probability density functions, histograms, dotplots, word clouds, time series charts, and maps to determine which statistical indicators are appropriate for each attribute of the collected data. However, when using statistical indicators, it is important to note that the mean reflects all data values ​​in the set, so outliers can affect the value, whereas the median uses the single value in the middle, so it can produce representative results even with outliers.

[0059] The process of analyzing the relationship between data attributes is to find combinations of attributes that have meaningful correlations within the data. The relationship analysis can be performed differently depending on the combination of attributes between qualitative attributes (Categorical Variable; Qualitative) that cannot be expressed numerically but can be arbitrarily quantified and quantitative attributes (Numeric Variable; Quantitative). The qualitative-qualitative relationship (Categorical - Categorical) can be displayed by using cross tables and mosaic plots to count the number of values ​​corresponding to each pair of attribute values. The quantitative-qualitative relationship (Numeric-Categorical) can be visually expressed by observing statistical values ​​(mean, median, etc.) by category or using box plots. The quantitative-quantitative relationship (Numeric-Numeric) can be analyzed for the association between two attributes using correlation coefficients. A correlation coefficient of -1 indicates a negative correlation where the two attributes change in opposite directions, 0 indicates no correlation, and 1 indicates a positive correlation where the two attributes always change in the same direction. The relationship between two attributes with a correlation coefficient can take many forms, and this can be visually represented using a scatter plot.

[0060] 3. Data preprocessing and cleaning

[0061] Once the data has been inspected and explored, data preprocessing is performed to transform it into a format suitable for machine learning training models. Data preprocessing involves refining data and transforming it into a form understandable by the model. Data preprocessing typically includes handling missing data, removing outliers, scaling, categorical data encoding, feature selection and extraction, and data transformation. The detailed data preprocessing steps can be performed in whole or in part, and a separate machine learning model may be used for data preprocessing.

[0062] Handling missing data involves handling missing values ​​in data. Missing values ​​can be displayed as NaN (Not a Number) or blank, or they can be deleted. Filling in or deleting missing values ​​within the data improves data completeness. When filling in missing values, values ​​such as the mean, median, or mode can be used.

[0063] Outlier removal is the process of removing outliers, values ​​that deviate from the normal data pattern. Outliers can degrade model performance and should therefore be removed or replaced. Identifying outliers can be accomplished by deleting the corresponding rows or columns or replacing them with different values.

[0064] Data scaling is the process of adjusting the size of data. Through data scaling, the range of data can be adjusted, and the performance of the model or the convergence speed can be improved. Through data scaling, the characteristics of the data can be adjusted to a similar range, and data scaling can generally be applied with standardization and normalization. Standardization is a method of converting data into a distribution with a mean of 0 and a standard deviation of 1, and is mainly converted using the mean and standard deviation, and the standardized value z is It can be expressed as (x is the original value, μ is the mean, σ is the standard deviation). Normalization is a method to convert the range of data to [0,1] or [-1,1], and mainly converts data using the minimum and maximum values, and the normalized value x norm silver can be expressed as (x is the original value, x min is the minimum, x max is the maximum value).

[0065] Categorical data encoding is the process of converting categorical variables, represented as strings or integers that cannot be directly input into a model, into numerical data types that can be input into the model. Typically, one-hot encoding or label encoding is used to convert categorical variables into numerical data types.

[0066] Feature selection and extraction is a process to improve model performance by selecting the most useful features for model learning or extracting new features. This process can reduce model complexity and prevent overfitting.

[0067] Data transformation is the process of transforming data to extract new information or to improve model understanding. This can include tokenizing text data or preprocessing image data. Data transformation can extract useful features from source data or transform data into an appropriate format, improving model performance.

[0068] Through data preprocessing as described above, the performance of machine learning models can be improved and stability can be secured.

[0069] Meanwhile, when training a learning model according to one embodiment of the present invention, a process of preprocessing information expressed in natural language and a process of learning a language model based on the preprocessed data may be performed.

[0070] 3-1. Text Preprocessing for Large-Scale Language Models

[0071] If the collected data has not been preprocessed to suit your needs, tokenization, cleaning, and normalization can be performed to suit the intended use of the data.

[0072] Tokenization refers to the process of dividing given data into units called tokens, which can be broadly defined as meaningful units. Tokenization can broadly include word tokenization and sentence tokenization.

[0073] Tokenization refers to the process of dividing given data into units called tokens, which can be broadly defined as meaningful units. Tokenization can broadly include word tokenization and sentence tokenization.

[0074] Word tokenization refers to cases where tokens are based on words, and in this case, words can include not only individual words but also word phrases and meaningful strings. Word tokenization separates words based on spaces or punctuation marks, such as periods, commas, question marks, semicolons, and exclamation marks. However, removing all punctuation or special characters during tokenization can sometimes result in tokens losing their meaning, necessitating a more precise tokenization algorithm. For example, if a word itself contains punctuation or uses special characters with meaning, simply removing them may not be enough. Therefore, tokenization rules such as the Penn Treebank Tokenization Rules can be applied during tokenization.

[0075] Sentence tokenization refers to the process of dividing text into sentences. Typically, if the data is unrefined, the corpus is not segmented into sentences, and thus sentence tokenization may be necessary to meet the intended use. Various rules for sentence tokenization can be defined depending on the language being used and how special characters are used within the corpus.

[0076] Tokenization is the process of classifying tokens according to their intended use. Before and after tokenization, cleaning and normalization are performed on text data to suit the intended use. Cleaning removes noise, while normalization integrates words with different representations and transforms them into a single, consistent word.

[0077] Refinement can occur before tokenization to eliminate any interference and facilitate tokenization. However, it can also be performed continuously and iteratively after tokenization to remove any remaining noise. The noise data removed during refinement are meaningless characters. Methods for removing unnecessary words include removing stopwords, low-frequency words, and short words.

[0078] Normalization work includes unifying words with different spellings based on rules, unifying uppercase and lowercase letters, etc. Unifying uppercase and lowercase letters is a normalization method that can reduce the number of words in English-speaking languages. In English-speaking languages, uppercase letters are only used in certain situations such as the beginning of a sentence, and most texts are written in lowercase letters, so unifying uppercase and lowercase letters can mostly be done by converting uppercase letters to lowercase letters.

[0079] Processing natural language in computing systems requires preprocessing, which involves digitizing text. This involves mapping each word in the text to a unique integer. This mapping process can utilize techniques such as integer encoding, padding, and one-hot encoding.

[0080] Integer encoding is a method of assigning integers to words. It creates a vocabulary by sorting words in order of frequency, and assigns integers in order of frequency, starting with the lowest number. Integer encoding performs sentence tokenization on text data containing multiple sentences, and performs word tokenization through parallel refinement and normalization. During this process, words are lowercase to unify the number of words, and stopwords and word length can be deleted. Through this, words can be recorded as keys and the frequency of each word as values. Integer encoding can be performed by sorting the text in order of frequency and assigning integers to words with high frequencies.

[0081] Padding is the process of randomly adjusting the length of sentences of different lengths within a text to the same length. Computing systems can perform parallel computations by grouping sentences of the same length into a single matrix. Specifically, to perform parallel computations, the lengths of sentences of different lengths within a text can be randomly padded with "0" to equalize the integer encoding results. Specifically, the longest sentence in a set of integer-encoded words can be identified, and a "0" can be added to the integer matrix corresponding to the length of the longest sentence. The computing system can then process sentences of the same length as a single matrix, allowing it to perform parallel processing. At this point, the computing system can ignore the "0" word, which is perceived as meaningless. This process of adjusting the size (shape) of data by filling it with a specific value is called padding. Using the number "0" to adjust the length is called zero padding.

[0082] One-hot encoding is a vector representation of words that uses the size of the vector as the dimension of the word set, assigning a value of 1 to the index of the word to be expressed, and 0 to all other indices. The vector expressed in this way is called a one-hot vector. One-hot encoding consists of integer encoding and an index assignment process. After integer encoding is performed and a unique integer is assigned to each word, the unique integer of the word to be expressed is regarded as an index, and a "1" is assigned to the corresponding position, and a "0" is assigned to the index positions of other words. However, one-hot encoding has the disadvantage that the space required to store the vector increases as the number of words increases (the dimensionality of the vector increases), and the similarity between words cannot be identified. To address these shortcomings, techniques that reflect the latent meaning of words and vectorize them into a multidimensional space include LSA (Latent Semantic Analysis), a count-based vectorization method; NNLM, RNNLM, Word2Vec, and FastText, which vectorize based on prediction; and the GloVe method, which uses both count-based and prediction-based methods.

[0083] Meanwhile, for computers to understand and process text, it must be appropriately converted into numbers. Because the performance of natural language processing can vary significantly depending on how words are represented, numerous techniques have been proposed to quantify words. Currently, the most widely used method is word embedding, which vectorizes each word through artificial neural network training.

[0084] Word embedding is a method of representing words as vectors, converting them into dense representations. The resulting word embedding is called a dense vector, or embedding vector. Word embedding methods include LSA, Word2Vec, FastText, and Glove.

[0085] 4. Modeling and Training

[0086] Artificial neural networks can be trained using training data. Here, "training" refers to the process of determining the parameters of an artificial neural network using training data to achieve objectives such as classification, regression analysis, or clustering of input data. Representative examples of artificial neural network parameters include the weights assigned to synapses and the biases applied to neurons.

[0087] An artificial neural network trained using training data can classify or cluster input data based on its patterns. Meanwhile, an artificial neural network trained using training data is referred to herein as a "trained model."

[0088] The following explains the learning methods of artificial neural networks. Learning methods of artificial neural networks can be broadly categorized into supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning.

[0089] Supervised learning is a machine learning method that infers a function from training data. Among these inferred functions, regression analysis is the process of outputting continuous values, while classification is the process of predicting and outputting the class of an input vector.

[0090] In supervised learning, an artificial neural network is trained with labels for training data. Here, the label can mean the correct answer (or result value) that the artificial neural network should infer when training data is input to the artificial neural network. In this specification, the correct answer (or result value) that the artificial neural network should infer when training data is input is called a label or labeling data. In addition, in this specification, setting a label on training data for learning of the artificial neural network is called labeling the training data. In this case, the training data and the label corresponding to the training data constitute a single training set, and can be input to the artificial neural network in the form of a training set.

[0091] Meanwhile, training data represents multiple features, and labeling the training data can mean that the features represented by the training data are labeled. In this case, the training data can represent the features of the input object in vector form. An artificial neural network can use the training data and labeled data to infer a function regarding the relationship between the training data and the labeled data. Furthermore, the parameters of the artificial neural network can be determined (optimized) by evaluating the function inferred by the artificial neural network.

[0092] Unsupervised learning is a type of machine learning in which training data is not labeled. Specifically, unsupervised learning can be a learning method that trains an artificial neural network to find and classify patterns in the training data itself, rather than the relationship between the training data and the corresponding labels. Examples of unsupervised learning include clustering and independent component analysis (ICA). In this specification, the term "clustering" may be used interchangeably with the term "clustering."

[0093] Examples of artificial neural networks that utilize unsupervised learning include generative adversarial networks (GANs) and autoencoders (AEs).

[0094] Generative adversarial networks (GANs) are a machine learning method in which two different AI components, a generator and a discriminator, compete to improve performance. In this case, the generator is a model that creates new data, capable of generating new data based on original data. The discriminator, a model that recognizes data patterns, can determine whether the input data is original or new data generated by the generator. The generator learns from data that fails to fool the discriminator, while the discriminator learns from data that the generator deceives. Accordingly, the generator can evolve to fool the discriminator as effectively as possible, while the discriminator can evolve to effectively distinguish between original data and data generated by the generator.

[0095] An autoencoder is a neural network that aims to reproduce the input itself as an output. An autoencoder comprises an input layer, at least one hidden layer, and an output layer. In this case, since the number of nodes in the hidden layer is smaller than that in the input layer, the data dimensionality is reduced, leading to compression or encoding. Furthermore, data output from the hidden layer is fed into the output layer. In this case, since the number of nodes in the output layer is larger than that in the hidden layer, the data dimensionality increases, leading to decompression or decoding.

[0096] Meanwhile, autoencoders express input data as hidden layer data by adjusting the connection strengths of neurons through learning. The hidden layer expresses information with a smaller number of neurons than the input layer. The ability to reproduce input data as output implies that the hidden layer has discovered and expressed hidden patterns in the input data.

[0097] Semi-supervised learning is a type of machine learning that utilizes both labeled and unlabeled training data. One technique for semi-supervised learning is to infer labels for unlabeled training data and then use these inferred labels to perform training. This technique can be useful in situations where labeling is expensive.

[0098] Reinforcement learning is the theory that, if an agent is given an environment in which it can determine the optimal action at any given moment, it can find the optimal path through experience without data. Reinforcement learning is primarily implemented using a Markov Decision Process (MDP). A Markov Decision Process is described as follows: first, an environment containing the information necessary for the agent to take the next action is provided; second, how the agent will act in that environment is defined; third, what rewards the agent will receive for performing well and what penalties will be imposed for performing poorly is defined; and fourth, the optimal policy is derived through repeated experience until the future reward reaches its maximum.

[0099] The structure of an artificial neural network is specified by the model configuration, activation function, loss function or cost function, learning algorithm, optimization algorithm, etc., and the hyperparameters are set in advance before learning, and the model parameters are set through learning afterwards, so that the content can be specified.

[0100] For example, factors that determine the structure of an artificial neural network may include the number of hidden layers, the number of hidden nodes included in each hidden layer, the input feature vector, and the target feature vector.

[0101] Hyperparameters include various parameters that must be initially set for learning, such as initial values ​​for model parameters. Furthermore, model parameters include various parameters to be determined through learning. For example, hyperparameters may include initial values ​​for inter-node weights, initial values ​​for inter-node biases, mini-batch size, number of learning iterations, and learning rates. Furthermore, model parameters may include inter-node weights, inter-node biases, and more.

[0102] The loss function can be used as an indicator (standard) to determine the optimal model parameters during the learning process of an artificial neural network. In an artificial neural network, learning refers to the process of manipulating model parameters to reduce the loss function, and the purpose of learning can be seen as determining the model parameters that minimize the loss function. The loss function can mainly use the mean squared error (MSE) or the cross entropy error (CEE), but the present invention is not limited thereto. The cross entropy error can be used when the correct answer label is one-hot encoded. One-hot encoding is an encoding method that sets the correct answer label value to 1 only for neurons corresponding to the correct answer, and sets the correct answer label value to 0 for neurons that are not the correct answer.

[0103] In machine learning or deep learning, learning optimization algorithms can be used to minimize the loss function. Learning optimization algorithms include gradient descent (GD), stochastic gradient descent (SGD), momentum, Nesterov Accelerate Gradient (NAG), Adagrad, AdaDelta, RMSProp, Adam, and Nadam.

[0104] Gradient descent is a technique that adjusts model parameters in a direction that reduces the loss function value by considering the gradient of the loss function at the current state. The direction of model parameter adjustment is called the step direction, and the size of the adjustment is called the step size. Here, the step size can represent the learning rate. Gradient descent obtains the gradient by partially differentiating the loss function with respect to each model parameter, and updates the model parameters by changing the learning rate in the direction of the obtained gradient.

[0105] Stochastic gradient descent is a technique that divides learning data into mini-batches and performs gradient descent on each mini-batch to increase the frequency of gradient descent.

[0106] Adagrad, AdaDelta, and RMSProp are techniques for improving optimization accuracy by adjusting the step size in SGD. In SGD, momentum and NAG are techniques for improving optimization accuracy by adjusting the step direction. Adam combines momentum and RMSProp to improve optimization accuracy by adjusting the step size and step direction. Nadam combines NAG and RMSProp to improve optimization accuracy by adjusting the step size and step direction.

[0107] The learning speed and accuracy of artificial neural networks are significantly influenced by not only the network structure and the type of learning optimization algorithm, but also hyperparameters. Therefore, to obtain a good learning model, it is crucial not only to determine an appropriate artificial neural network structure and learning algorithm, but also to set appropriate hyperparameters.

[0108] Typically, hyperparameters are experimentally set to various values ​​while training an artificial neural network, and the learning results are set to the optimal values ​​that provide stable learning speed and accuracy.

[0109] FIG. 1 is a schematic diagram of an electronic device according to one embodiment of the present invention, FIG. 2 is a schematic diagram of a material property prediction model according to one embodiment of the present invention, FIG. 3 is a schematic diagram of a material property prediction method according to one embodiment of the present invention, FIG. 4 is a schematic diagram for explaining the concepts of local, quasi-global, global interaction and unit cell according to one embodiment of the present invention, and FIG. 5 is a schematic diagram of a material property prediction service system utilizing a material property prediction model according to one embodiment of the present invention.

[0110] As illustrated in FIG. 1, an electronic device (100) according to embodiments of the present invention may include a processor (110), a memory (120), and a communication unit (130). The electronic device (100) is a basic configuration for performing a computing environment, and in other embodiments, the electronic device (100) may be implemented by additionally or alternatively including some other components, may be implemented as a single or multiple entities, or may be implemented as only some of the disclosed components. Components or at least some of the components inside or outside the electronic device (100) may be connected to each other through a BUS, a GPIO (General Purpose Input / Output), an SPI (Serial Peripheral Interface), or a MIPI (Mobile Industry Processor Interface), thereby transmitting and receiving data or signals.

[0111] Unless the context clearly indicates otherwise, the processor (110) may refer to a set of one or more processors, and may control components of the processor (110) and the electronic device (100) by executing software (e.g., commands, programs, etc.) stored in at least the memory (120). In addition, the processor (110) may perform various operations such as calculations, processing, data generation or processing, and may read data from or store data in the memory (120). The processor (110) may be composed of at least one core and may include a processor for data analysis, machine learning (ML), or deep learning (DL), such as a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), or a tensor processing unit (TPU). The processor (110) may read software stored in the memory (120) to perform data processing for machine learning (or deep learning) of the present invention. According to one embodiment of the present disclosure, the processor (110) can perform operations for learning a neural network. The processor (110) can perform calculations for learning a neural network, such as processing input data for learning in deep learning, extracting features from the input data, calculating errors, and updating weights of the neural network using backpropagation. At least one of the CPU, GPGPU, and TPU of the processor (110) can process learning of a neural network model. For example, the CPU and GPGPU can together process learning of a neural network model and data classification using a neural network model. In addition, in one embodiment of the present disclosure, at least one processor (110) of the electronic device (100) can be used together to process learning of a neural network model and data classification using a neural network model.

[0112] The memory (120) is for storing various data, and the data is data acquired, processed, or used by at least one component of the electronic device (100), and may include software (e.g., commands, programs, etc.). Unless explicitly expressed otherwise in the context, the memory (120) may refer to a set of one or more memories, and may include at least one type of storage medium among a flash memory type, a hard disk type, a multimedia card micro type, a card type memory (e.g., SD or XD memory, etc.), a RAM (Random Access Memory), a SRAM (Static Random Access Memory), a ROM (Read Only Memory), an EEPROM (Electrically Erasable Programmable Read-Only Memory), a PROM (Programmable Read-Only Memory), a magnetic memory, a magnetic disk, an optical disk, and a web storage that performs a storage function on the Internet. The instructions or programs or software stored in the memory (120) may be used to refer to an operating system, an application for controlling components of the electronic device (100), or middleware that provides various functions to the application so that the application can utilize the components of the electronic device (100). In one embodiment, when the processor (110) performs a specific operation, the memory (120) may store instructions that are performed by the processor (110) and correspond to the specific operation.

[0113] The communication unit (130) performs wireless or wired communication between the electronic device (100) and another device (e.g., a user terminal or another server), and the communication unit (130) can use wireless communication systems according to methods such as eMBB, URLLC, MMTC, LTE, LTE-A, NR, UMTS, GSM, CDMA, WCDMA, TDMA, FDMA, OFDMA, SCFDMA, WiBro, WiFi, Bluetooth, NFC, GPS, or GNSS. In addition, the communication unit (130) can use various wired communication systems such as USB, HDMI, RS-232 (Recommended Standard-232), POTS (Plain Old Telephone Service), PSTN (Public Switched Telephone Network), xDSL (x Digital Subscriber Line), RADSL (Rate Adaptive DSL), MDSL (Multi Rate DSL), VDSL (Very High Speed ​​DSL), UADSL (Universal Asymmetric DSL), HDSL (High Bit Rate DSL), and local area network (LAN). In one embodiment of the present invention, the communication unit (130) can be configured regardless of the communication mode such as wired or wireless, and can be configured with various communication networks such as a personal area network (PAN), a wide area network (WAN), etc. Additionally, the communication network may be the well-known World Wide Web (WWW), and may also utilize a wireless transmission technology used for short-distance communication, such as infrared (IrDA: Infrared Data Association) or Bluetooth.

[0114] An electronic device (100) according to one embodiment of the present invention may further include an input / output unit. The input / output unit may be configured as a separate input unit and an output unit, but alternatively, the input / output unit may have an integrated configuration. The input unit, as a means for inputting data, may be configured in various ways. For example, the input unit may be configured to receive user input. The input unit may be configured to receive user input from a user terminal. Here, “receiving input” may mean receiving an input signal (or selection signal) corresponding to a user input based on an input made by a user through an input unit configuration provided in a user terminal. The input unit may also be referred to as a user interface module. The input unit may include a touch screen, a computer mouse, a keyboard, a keypad, a touch pad, a trackball, a joystick, a voice recognition module, or other similar devices. However, the present invention does not place any limitations on the type of input unit. Furthermore, the input unit in the present invention does not necessarily mean a hardware means, but may be understood as a channel for receiving input from a user. Here, user input may include documents, text, images (or videos), etc. Next, the output unit may output information via an output unit configuration (e.g., a display unit, a touch screen, etc.) provided on a user terminal or computing device. The output unit does not necessarily refer to a hardware device, but can be understood as a conduit for outputting results to the user.

[0115] An electronic device (100) according to an embodiment of the present invention can execute software that configures a property prediction system or a property prediction method.

[0116] One embodiment of the present invention utilizes both the local interaction representation capability of GNNs and the long-range interaction representation capability of Transformers, while subdividing these into physically meaningful multi-scale models to maximize modeling accuracy. Specifically, the material property prediction system of one embodiment of the present invention divides the atomic interactions within a material into three scales, applies an architecture optimized for each scale to extract features, and then integrates these features to predict material properties. Here, the three types of material atomic interactions may include local interactions, quasi-global interactions, and global interactions. Referring to Fig. 4, local interaction refers to a strong chemical bond and interaction between a specific atom and its surrounding neighboring atoms, semi-global interaction refers to an interaction between all atoms within a unit cell, which is the basic unit of a crystal structure, and global interaction may refer to a long-range interaction between unit cells or between atoms within an extended lattice, considering the characteristics of a crystal structure in which unit cells are periodically repeated. Here, a unit cell is the smallest repeating unit that constitutes a crystal structure, and when these unit cells are stacked in a three-dimensional space, the entire crystal structure can be completed. In other words, the unit cell functions as a core blueprint that contains all information of the entire crystal.

[0117] As illustrated in FIG. 2, the property prediction model (200) of the property prediction system of one embodiment of the present invention may include a preprocessing module (210), a local interaction operation unit (220), a quasi-global interaction operation unit (230), a global interaction operation unit (240), and a property prediction unit (250).

[0118] The preprocessing module (210) is a processing module that receives atomic structure data of a material to be studied or predicted and converts it into a format that can be processed by a system or learning model. The material property prediction model (200) according to one embodiment of the present invention can receive data in a format that describes the atomic structure of a material in three-dimensional coordinates, for example, CIF (Crystallographic Information File) data.

[0119] CIF is a standard text file format for exchanging and storing crystallographic data. It is designed to standardize data in the field of crystallography, enabling easy interpretation of the atomic structure of materials by computing devices. A CIF file is a simple text-based format structured according to specific rules. Its main components can be data blocks, data names, data values, and loop structures. Data blocks specify multiple crystal structures contained within a single CIF file. A single CIF file can contain one or more data blocks, each of which can use the name of a specific compound. This means that multiple crystal structures can be contained within a single file. Data names (or tags) are standardized names that identify each data item. For example, a data name indicating the a-axis length of a unit cell could be written as _cell_length_a. Data values ​​are the actual values ​​corresponding to the data names, and values ​​can be separated from the data names by spaces. Loop structures are designed to efficiently represent repetitive data, such as multiple atomic coordinates. A loop structure uses, for example, the loop_ directive, and lists multiple data names starting from the line following loop_, and below that, lists the values ​​corresponding to each name in order.

[0120] CIF input data in this format can include the type of each atom, three-dimensional information, and information on the unit cell containing them, and the preprocessing module (210) can extract this information from this input data.

[0121] Alternatively, in another embodiment of the present invention, input data for crystal structures can be obtained via the Robocrystallographer package, which generates text descriptions similar to how a real crystallographer would analyze a structure. The Robocrystallographer package indicates symmetry, local environment, and extended connectivity when generating text descriptions for crystal structures, and the package may include utilities for identifying molecule names, component orientations, heterostructure information, etc. For example, the Robocrystallographer used in one embodiment, given SnO2 as input, reports "SnO2 is Rutile structured and crystallizes in the tetragonal P4_2 / mnm space group. The structure is three-dimensional. Sn(1) is bonded to six equivalent O(1) atoms to form a mixture of edge and corner-sharing SnO6 octahedra. The corner-sharing octahedral tilt angles are 51°. All Sn(1)-O(1) bond lengths are 2.09 . O(1) is bonded in a trigonal planar geometry to three equivalent Sn(1) atoms." Such text descriptions can contain a wide range of information, including global properties (e.g., space group and crystal type), local details (e.g., bond lengths and coordination environments), and semi-global properties (e.g., connectivity and structural arrangement).

[0122] In one embodiment of the present invention, the preprocessing module (210) can preprocess CIF input data so that it can be processed by each of the respective operation units. That is, the preprocessing module (210) can generate different types of input data that can be processed by each of the local interaction operation unit (220), the quasi-global interaction operation unit (230), and the global interaction operation unit (240).

[0123] First, the preprocessing module (210) can convert (generate) input data of a graph data structure for GNN. In one embodiment of the present invention, the local interaction operation unit (220) and the global interaction operation unit (240) can be configured as GNNs, but unit cell information can be further added to the data input to the global interaction operation unit (240) so that an overall graph including the periodicity of the crystal structure can be generated.

[0124] The information extracted from the preprocessing module (210) can be converted into a graph data structure in which atoms are nodes and distances between atoms or bonds between atoms are edges. The graph data converted by the preprocessing module (210) can then be transferred to subsequent computation units. Here, each node can have an initial feature vector representing the unique characteristics of the element, such as atomic number, atomic weight, and electronegativity. This initial feature vector can be converted into a high-dimensional vector by passing through an atom embedding layer. In addition, edges, which are relationships between atoms, can be generated for a pair of atoms within a specific cutoff radius. That is, edges can have geometric information, such as distances and directions, between two atoms as their features.

[0125] Meanwhile, unit cell information may be added to the data input to the global interaction operation unit (240) composed of GNN so that an overall graph including the periodicity of the crystal structure can be generated.

[0126] To this end, the preprocessing module (210) can generate a graph representing the overall structure including the periodicity of the crystal when converting input data for the GNN of the global interaction operation unit (240). The preprocessing module (210) can connect neighboring atoms based on a long distance, for example, a large cutoff of 8 to 10. At this time, periodic boundary conditions (PBC) can be applied to create an edge index that includes all neighbors beyond the unit cell boundary. In addition, the preprocessing module (210) can include information indicating which lattice the neighbors are located in along with the inter-atom distance. Here, the periodic boundary conditions are a method of expressing the entire crystal structure that is infinitely repeated with only one unit cell, which is the smallest repeating unit.

[0127] In addition, the preprocessing module (210) can convert input data for the transformer of the quasi-global interaction operation unit (230). The preprocessing module (210) can generate a sequence of atoms and all distance relationship information between atoms from a CIF file. The input sequence is a tensor that sequentially lists the initial atomic feature vectors, and the distance matrix between all atom pairs in the unit cell can be an attention bias. That is, the preprocessing module can perform steps of parsing, characterization, and sequence generation to convert the input data for the transformer. Parsing is the process of extracting crystal structure information from the CIF file, which is the process of reading the CIF file and extracting the atom type (element symbol), atomic coordinates (3D coordinates), and lattice parameters (size and angle of the unit cell). Characterization is the process of converting the extracted information into numbers that the transformer can understand, which is the process of viewing each atom as a token and generating a vector representing the token. The preprocessing module can generate vectors for key atomic features, such as atomic number, periodic table, electronegativity, ionization energy, atomic radius, and covalent radius, and calculate the Euclidean distance between all atom pairs to produce a relative distance matrix. This can generate three-dimensional spatial information about the atoms. The preprocessing module can transform the transformer input data by combining the characterized information. The preprocessing module can list the atomic feature vectors of each atom in the crystal structure in order to create an input sequence, and add the relative distance matrix, which is the three-dimensional positional information of the atoms, as a bias to the attention matrix.

[0128] The preprocessing module (210) can convert and generate input data of a transformer including a relative distance matrix with an atomic feature vector tensor and attention bias as main inputs through the process described above.

[0129] The local interaction calculation unit (220) is for calculating local interaction information at the atomic level, and can calculate direct interactions between adjacent atoms, such as chemical bonds between atoms and van der Waals forces, as direct interactions between a specific atom and its surrounding neighboring atoms within a material. That is, the local interaction calculation unit (220) can calculate interaction information between an atom and its surrounding neighboring atoms within a specific radius (cutoff radius), and in the embodiment of the present invention, a graph neural network (GNN) can be applied to the local interaction calculation unit (220).

[0130] A graph neural network (GNN), particularly a message passing neural network (MPNN) family model, repeats the process in which each node (atom) propagates its feature vector to neighboring nodes in the form of a "message" and simultaneously receives messages from neighbors to update its state. Through this process, the chemical and structural environmental information of surrounding atoms can be gradually encoded in the feature vector of each atom. In one embodiment of the present invention, the GNN can update the feature vector by repeatedly performing message passing operations on graph data in which atoms transmitted from the preprocessing module (210) are nodes and atom pairs within a certain distance (cutoff radius) are edges. That is, in each operation step of the GNN, an atom (node) collects information on neighboring atoms (neighboring nodes) and bonds (edges) and updates its feature vector. Through this process, local interaction features that precisely embed the local chemical environment and bonding characteristics at the atomic level can be extracted. By repeating message-passing operations across multiple layers, GNNs can allow atoms to gradually integrate information from a wider range of neighboring atoms. Specifically, by combining the feature vectors of an atom and those of its neighboring atoms and repeating this across multiple layers, GNNs can represent the local chemical environment based on geometric relationships, such as interatomic distances and angles, as an embedding vector. The features derived through this local interaction operation (220) can contain the most fundamental information that determines material properties.

[0131] Referring to FIG. 4, the local interaction between atoms calculated in the local interaction calculation unit (220) of one embodiment of the present invention is, for example, a Na atom and a cleavage radius (r max) can mean interactions within the local atomic environment of a unit cell centered on a specific atom, such as interactions between Cl atoms (Cl1-1, Cl1-2, etc.) within the unit cell. That is, local interaction operations between atoms in such a local atomic environment can derive properties that are directly related to the formation of chemical bonds between atoms.

[0132] The quasi-global interaction calculation unit (230) is intended to calculate structural feature information at the unit cell level by comprehensively considering the interactions between all atomic pairs existing within a unit cell. That is, the quasi-global interaction calculation unit (230) can calculate structural feature information of the entire unit cell beyond the local environment of individual atoms of the material by modeling the atomic arrangement and interactions of the entire unit cell, which is the basic structural unit of the material. Through this, the quasi-global interaction calculation unit (230) can calculate characteristics such as symmetry or density of the entire unit cell, including interactions between relatively distant atoms that do not have direct bonds, that is, interactions between distant atoms beyond the truncation radius of FIG. 4 (the truncation radius of GNN). In one embodiment of the present invention, the quasi-global interaction calculation unit (230) may apply a transformer architecture to comprehensively consider interactions between all atoms within a unit cell.

[0133] As a quasi-global interaction operation unit (230) according to one embodiment of the present invention, a transformer can set all atoms in a unit cell as a single sequence and apply a self-attention mechanism to this atomic sequence. The quasi-global interaction operation unit (230) can calculate an attention weight (attention score) through the self-attention mechanism of the transformer, which indicates how related one atom in the sequence, that is, one atom in the unit cell, is to all other atoms. As this relationship weight is calculated, the interaction relationship between all atoms in the unit cell can be captured. In other words, the self-attention mechanism, which is the core of the transformer, can generate "Query", "Key", and "Value" vectors for all input atoms (tokens), and calculate the similarity of all query-key pairs to calculate the attention weight. These attention weights indicate how much each atom should pay attention to all the other atoms in the unit cell, and by applying them to the value vector to produce a weighted sum, the feature vector of each atom can be updated.

[0134] In this way, the quasi-global interaction operation unit (230) of one embodiment of the present invention can learn the characteristics of the unit cell, such as symmetry, orientation (coordination environment), and electrical interaction of the unit cell, which are difficult to capture by GNN. That is, unlike GNN, the quasi-global interaction operation unit (230) of one embodiment directly models the relationship between all atoms without distance constraints, and thus can effectively learn the interaction across the entire unit cell.

[0135] The global interaction calculation unit (240) is for calculating long-range interactions between unit cells that are periodically repeated within the material crystal structure. That is, the global interaction calculation unit (240) of one embodiment of the present invention can calculate information on interactions between a reference unit cell and neighboring unit cells or long-range interactions within an extended supercell. The global interaction calculation unit (240) of one embodiment of the present invention can capture interactions that occur when unit cells are repeated in three-dimensional space by modeling long-range interactions that occur due to the periodicity of the crystal structure beyond the unit cell.

[0136] Referring to FIG. 4, the global interaction captured through the global interaction operation unit (240) according to one embodiment refers to a long-range interaction occurring between a reference unit cell and neighboring unit cells repeatedly arranged by periodic boundary conditions, which can play a very important role in determining the bulk properties of a material.

[0137] In one embodiment of the present invention, the global interaction operation unit (240) may utilize GNN, but in another embodiment, the global interaction operation unit (240) may apply a transformer.

[0138] In one embodiment, when a GNN is used as the global interaction operation unit (240), the GNN can regard a 'supercell' composed of multiple copies of a unit cell as a single large graph, and apply the GNN to this to model long-distance interactions. That is, the GNN-based global interaction operation unit (240) can perform operations by constructing a larger-scale graph in which the unit cell itself is regarded as a node of the graph, or by applying a periodic boundary condition to expand the graph so that atoms within the original unit cell interact with atoms in other adjacent unit cells. When a GNN is applied to this expanded graph, interactions between lattices can be naturally learned as message passing propagates beyond the boundaries of the unit cells.

[0139] In another embodiment of the present invention, when a transformer is used as the global interaction calculation unit (240), it can be designed to intensively learn global characteristics of the entire crystal structure by introducing a special token such as a 'class token'. That is, in another embodiment, a 'class token' is added to a set of atoms in a unit cell input to the transformer, and this class token can be learned to intensively learn global information of the entire lattice while passing through the transformer layer. After introducing this class token, the global interaction calculation unit (240) can optionally reflect the periodicity of the unit cells in the positional encoding. The global interaction calculation unit (240) can calculate the most comprehensive range of interactions by applying self-attention to all sets of atoms in the extended large lattice (supercell). Through this process, the global interaction operation unit (240) can extract global interaction features related to macroscopic scale physical phenomena such as lattice strain and phonon vibration of the entire crystal.

[0140] As illustrated in FIG. 2, the material property prediction model (200) of one embodiment of the present invention may further include a material property prediction unit (250) for combining feature vectors (embeddings) output from each of the operation units into a single vector and predicting a target material property. In one embodiment of the present invention, the material property prediction unit (250) may combine local / semi-global / global interaction feature vectors output from the local interaction operation unit (220), the quasi-global interaction operation unit (230), and the global interaction operation unit (240) into a single integrated feature vector through a concatenation or element-by-element summation operation. This integrated feature vector includes all of the microscopic / macroscopic / multi-layer interaction information of the material. The material property prediction model (200) can map the integrated feature vector to a neural network such as a multi-layer perceptron (MLP) including at least one fully-connected layer, thereby mapping it to final material property values ​​such as formation energy, band gap, elastic modulus, etc. Through this process, the material property prediction model (200) of one embodiment of the present invention can perform the task of predicting target material properties of a material using a two-layer MLP.

[0141] As illustrated in FIG. 3, a method for predicting properties of an embodiment of the present invention is executed by a computer and may include an input data generation step (S110) of receiving atomic structure data of a material and generating input data for local interaction calculations, quasi-global interaction calculations, and global interaction calculations, respectively; a local interaction feature extraction step (S120) of calculating local interactions between specific atoms and their surrounding neighboring atoms based on preprocessed data to extract local interaction features; a quasi-global interaction feature extraction step (S130) of calculating interactions between all atoms in a unit cell of a crystal based on preprocessed data to extract quasi-global interaction features; a global interaction feature extraction step (S140) of calculating long-range interactions considering periodicity between unit cells based on preprocessed data to extract global interaction features; and a property prediction step (S150) of integrating each feature extracted at each step and predicting target properties of the material based on the same.

[0142] The input data generation step (S110) can convert the atomic structure data of a material so that it can be processed by a material property prediction model. In one embodiment of the present invention, data in a standard format such as a crystallographic information file (CIF) can be input. In this step, information such as the type of each atom, three-dimensional coordinates, and unit cell parameters are extracted from the input CIF data, and then three types of input data can be generated according to the characteristics of subsequent operation steps (local, semi-global, and global).

[0143] To generate graph data for local interaction operations, a graph structure can be created, with atoms represented as nodes and pairs of atoms within a specific cutoff radius represented as edges. Each node has an initial feature vector representing the element's unique properties, such as atomic number and electronegativity, while edges contain geometric information, such as the distance between two atoms. This input data (graph) can be used for GNN-based local interaction operations.

[0144] To generate sequence data for quasi-global interaction operations, an input sequence listing the feature vectors of all atoms within a unit cell in order and a relative distance matrix calculated by calculating the Euclidean distance between all pairs of atoms can be generated. This relative distance matrix can be used as an attention bias in the transformer's self-attention operation, providing three-dimensional spatial information.

[0145] To generate extended graph data for global interaction operations, a larger truncation radius than the local interaction graph can be applied to model the periodicity of the crystal structure, and periodic boundary conditions (PBCs) can be applied to generate an extended graph that extends beyond the unit cell boundary to neighboring atoms. This graph can be used for GNN-based global interaction operations.

[0146] The local interaction feature extraction step (S120) receives local graph data generated in the preprocessing step and extracts atomic-level local interaction features. For this purpose, a graph neural network (GNN), such as a message passing neural network (MPNN), can be used.

[0147] During the computation process, each atom (node) transmits its feature vector to its connected neighbors in the form of a "message," and simultaneously receives messages from neighbors to update its own state. By repeating this message-passing process across multiple layers, each atom's final feature vector can embed sophisticated local chemical environment information, such as chemical bonds, distances, and angles between neighboring atoms. These extracted local interaction features can provide the most fundamental information for determining material properties.

[0148] The semi-global interaction feature extraction step (S130) can use the sequence data generated in the preprocessing step to extract structural features at the unit cell level. A transformer architecture can be applied to comprehensively consider all interatomic interactions within the unit cell.

[0149] All atoms within a unit cell can be considered as a single sequence and a self-attention mechanism can be applied. Self-attention calculates an attention weight that indicates how closely an atom within a sequence is related to all other atoms. In this process, the relative distance matrix generated in the preprocessing step can be utilized as an attention bias to provide three-dimensional distance information between atoms. This allows for modeling interactions between atoms at long distances without direct bonding, without the constraints of the GNN's truncation radius, effectively learning semi-global features such as symmetry, density, and coordination environment of the entire unit cell.

[0150] The global interaction feature extraction step (S140) can extract long-range interaction features between unit cells that arise due to the periodicity of the crystal structure. For this purpose, the extended graph data generated in the preprocessing step can be used. In one embodiment, a GNN can be used to extract global interaction features. When a message passing operation is performed on an extended graph with periodic boundary conditions (PBC), the message is propagated beyond the boundaries of the unit cell to atoms in adjacent unit cells. This process allows for the extraction of global features related to macroscopic physical phenomena, such as the interaction between the reference unit cell and neighboring unit cells, i.e., lattice strain or phonon vibrations of the entire crystal. These global features provide important information for determining the bulk properties of the material. Meanwhile, in another embodiment of the present invention, a transformer can be applied to extract global interaction features. For example, by adding a special token, such as a 'class token', to the atomic sequence, the transformer can be trained to integrate information about the entire crystal structure into this token.

[0151] The material property prediction step (S150) combines the local, quasi-global, and global interaction feature vectors extracted from the three previous steps into a single integrated feature vector. The combining method can be either concatenation, which simply connects each vector, or element-wise summation, which adds each element of the vector. This integrated feature vector contains multi-layered information, from microscopic to macroscopic interactions of the material. This integrated feature vector can be input to a fully connected neural network, such as a multilayer perceptron (MLP), and ultimately mapped to the target material property values ​​to be predicted (e.g., formation energy, band gap, elastic modulus, etc.).

[0152] As illustrated in FIG. 5, a material property prediction service system (300) utilizing a material property prediction model (200) according to one embodiment of the present invention includes a user computing device (310) and a server computing system (320), and these components are communicatively connected through a service environment (e.g., an Internet site) to provide services such as property prediction for input molecules to the user computing device (310).

[0153] The user computing device (310) is a client terminal for requesting an analysis service by uploading an atomic structure data file of a material to be analyzed for its physical properties, such as a crystallographic information file (CIF), and receiving the results. The user computing device (310) may include any type of computing device capable of Internet access, such as a smartphone, tablet PC, or desktop computer.

[0154] The user computing device (310) may include a user input unit (311) (e.g., a touchscreen, a keyboard) for receiving user input and a display (312) for outputting analysis results received from the server. The user may access the service via a web browser or a dedicated application and input information about the molecule to be analyzed, for example, 3D molecule data, in the form of a prompt. The input request data may be transmitted to the server computing system (320) via a communication unit.

[0155] The server computing system (320) includes a high-performance processor such as a central processing unit (CPU) and a graphics processing unit (GPU) and a large-capacity memory, and can load and execute a property prediction model (200) according to an embodiment of the present invention. The property prediction model (200) according to an embodiment may include a preprocessing module (210), a multi-scale interaction operation unit, and a property prediction unit (250). The multi-scale interaction operation unit extracts local / quasi-global / global characteristics of a material from molecular structure data input by a user, and may be configured to include a local interaction operation unit (220), a quasi-global interaction operation unit (230), and a global interaction operation unit (240).

[0156] The preprocessing module (210) converts atomic structure data in the same format as the crystallographic information file (CIF) input by the user into three types of data according to the characteristics of subsequent operation units and generates them. The first input data is for calculating local characteristics, and can generate a graph structure in which atoms are nodes and atom pairs within a specific distance are edges. The second input data is for calculating quasi-global characteristics, and can generate an input sequence that sequentially lists the feature vectors of atoms in a unit cell and a matrix of relative distances between atoms. The third input data is for calculating global characteristics, and can generate an extended graph structure connected to atoms beyond the unit cell boundary by considering the periodicity of the unit cell. The user input data converted in this way are input to the local interaction operation unit (220), the quasi-global interaction operation unit (230), and the global interaction operation unit (240) of the multi-scale interaction operation unit, respectively, so that the interaction characteristics of the molecular structure input by the user can be extracted.

[0157] The local interaction operation unit (220) can extract local interaction features, such as chemical bonds between atoms, through message passing operations by applying a graph neural network (GNN) to the first input data (graph). The quasi-global interaction operation unit (230) can extract quasi-global features by capturing the interaction relationships between all atoms within a unit cell through a self-attention mechanism by applying a transformer architecture to the second input data (sequence). At this time, three-dimensional spatial information can be reflected by using the relative distance matrix between atoms as an attention bias. The global interaction operation unit (240) can extract global features on a macroscopic scale, such as lattice deformation, through message passing that crosses the unit cell boundary by applying another graph neural network (GNN) to the third input data (extended graph).

[0158] The material property prediction unit (250) can combine the local, semi-global, and global feature vectors derived from the multi-scale interaction operation unit into a single integrated feature vector. This integrated feature vector can include everything from microscopic information to macroscopic information of the material. Finally, the material property prediction unit (250) can input this integrated feature vector into a fully connected neural network such as a multilayer perceptron (MLP) that has completed training to predict target material property values ​​requested by the user, such as formation energy, band gap, and elastic modulus. That is, the MLP can receive the integrated feature vector as input, calculate the optimal material property value based on the previously trained pattern, and output it. Through this series of processes, the final predicted material property value can be transmitted back to the user's computing device (310) through the communication unit and provided to the user.

[0159] Meanwhile, one embodiment of the present invention may be implemented as an application-specific integrated circuit (ASIC) manufactured to suit the special functions of a specific application field and device.

[0160] An application-specific integrated circuit is also called an application-specific semiconductor. Unlike standard semiconductors that have set specifications and can be applied to any electronic product or application as long as certain requirements are met, an application-specific semiconductor is an integrated circuit that a semiconductor manufacturer manufactures according to a specific order for a specific product or function. In other words, an application-specific semiconductor is designed and manufactured to perform only the functions required for a specific device or specific function. Depending on the design method, application-specific semiconductors are largely divided into full custom ICs, which design and manufacture the circuit from scratch according to the user's needs, and semi-custom ICs, which design and manufacture the circuit using some standardized designs.

[0161] Application-specific semiconductors are primarily used in communications systems, high-performance computing systems, consumer electronics, automobiles, industrial automation, medical devices, military, and aerospace industries. Recently, they are being applied to AI semiconductors that perform large-scale calculations required for AI implementation with high performance and power efficiency.

[0162] Application-specific integrated circuits (ASICs) are core components of network routers, switches, and modems in communication systems, performing data packet processing, protocol conversion, and signal processing to deliver high throughput and low latency. In high-performance computing systems, ASICs are key components for high-speed and parallel processing. In consumer electronics such as digital cameras, smartphones, tablets, and game consoles, ASICs provide high-performance and low-power solutions required to perform specific functions. In the automotive industry, ASICs control various electronic systems within vehicles, and in industrial automation systems, ASICs provide solutions for high-precision control and high-performance processing.

[0163] An application-specific integrated circuit to which one embodiment of the present invention is applied includes a memory in which an individual memory interface (I / F) is implemented, and may include a plurality of functional blocks that request memory access. Each functional block may be a direct memory access (DMA) functional block, a processor, a video processor, a cache controller, a decompression block, or a data path block. The basic configuration of the application-specific integrated circuit may include a transistor that amplifies or switches an electrical signal, a logic gate that is a circuit that performs a logical function by combining transistors, a memory cell that stores data, an analog circuit that is a circuit that processes a continuous voltage or current by combining transistors, and an IP core (Intellectual Property Core) such as a microprocessor, DSP, or graphic core that is pre-designed to perform a specific function.

[0164] The ASIC may also include a separate memory I / F interfacing with individual memories and an embedded memory I / F interfacing with embedded memories. The separate memory I / F is connected to each functional block, receives memory access signals (e.g., control signals, address signals, and data signals), and generates signals for controlling the individual memories based on these input signals. The embedded memory I / F is connected to each functional block, receives memory access signals (e.g., control signals, address signals, and data signals), and generates modified memory access signals for controlling the embedded memories based on these input signals. The separate memory I / F and the embedded memory I / F may be designed within the memory control block of the ASIC to provide a memory control structure that can be flexibly applied to both the individual memories and the embedded memories.

[0165] Additionally, an application-specific integrated circuit (ASIC) for an artificial neural network (ANN) may be configured to include a plurality of neurons arranged in an array and a plurality of synaptic circuits, each neuron including a register, a microprocessor, and at least one input, and each synaptic circuit including a memory for storing synaptic weights. Each neuron of the ASIC may be connected to at least one other neuron through one of the plurality of synaptic circuits.

[0166] Although the present disclosure has been described above as being generally implemented by a computing device, those skilled in the art will appreciate that the present disclosure may also be implemented in combination with computer-executable instructions and / or other program modules that may be executed on one or more computers and / or as a combination of hardware and software.

[0167] Those skilled in the art will appreciate that information and signals may be represented using any of a variety of different technologies and techniques. For example, the data, instructions, commands, information, signals, bits, symbols, and chips referenced in the above description may be represented by voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields or particles, or any combination thereof.

[0168] Those skilled in the art will appreciate that the various illustrative logical blocks, modules, processors, means, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, various forms of programs or design code (referred to herein, for convenience, as software), or a combination of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Those skilled in the art may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.

[0169] The various embodiments presented herein can be implemented as a method, apparatus, or article of manufacture using standard programming and / or engineering techniques. The term article of manufacture includes a computer program, carrier, or media accessible from any computer-readable storage device. For example, computer-readable storage media include, but are not limited to, magnetic storage devices (e.g., hard disks, floppy disks, magnetic strips, etc.), optical disks (e.g., CDs, DVDs, etc.), smart cards, and flash memory devices (e.g., EEPROMs, cards, sticks, key drives, etc.). Furthermore, various storage media presented herein include one or more devices and / or other machine-readable media for storing information.

[0170] It should be understood that the specific order or hierarchy of steps in the presented processes is merely an example of exemplary approaches. It should be understood that the specific order or hierarchy of steps in the processes may be rearranged within the scope of the present disclosure based on design priorities. The appended method claims provide elements of various steps in a sample order, but are not intended to be limited to the specific order or hierarchy presented.

[0171] The description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments without departing from the scope of the present disclosure. Therefore, the present disclosure is not intended to be limited to the embodiments disclosed herein, but is to be construed in the broadest scope consistent with the principles and novel features disclosed herein.

Claims

1. As a system for predicting the properties of materials, at least one processor; and At least one memory storing instructions or information or artificial intelligence models executed by at least one processor; The instructions or information or artificial intelligence model executed by at least one processor, A preprocessing module that receives atomic structure data of a material and generates first input data for local interaction operations, second input data for quasi-global interaction operations, and third input data for global interaction operations; A local interaction operation unit that extracts local interaction features by calculating local interactions between a specific atom and its surrounding neighboring atoms based on the first input data; A quasi-global interaction calculation unit that calculates the interaction between all atoms within the unit cell of the decision based on the second input data and extracts quasi-global interaction features; and A global interaction operation unit that extracts global interaction features by calculating long-distance interactions considering periodicity between unit cells based on the third input data. Material property prediction system.

2. In claim 1, The above local interaction operation unit includes a graph neural network, The first input data input to the graph neural network is a graph structure in which atoms of the material are nodes and pairs of atoms located within a specific cutting radius of the material are edges. Material property prediction system.

3. In claim 1, The above quasi-global interaction operation unit includes a transformer, The second input data input to the transformer includes an input sequence that sequentially lists the feature vectors of atoms in the unit cell. Material property prediction system, 4. In claim 3, The above quasi-global interaction operation unit extracts the quasi-global interaction features by using the relative distance matrix between all atom pairs in the unit cell as an attention bias of the self-attention operation. Material property prediction system.

5. In claim 1, The above global interaction operation unit includes another graph neural network, The third input data input to the above graph neural network is an extended graph structure that connects to neighboring atoms beyond the boundary of the unit cell. Material property prediction system.

6. In claim 1, It further includes a property prediction section, The above property prediction unit generates an integrated feature vector by integrating the respective feature vectors of the local interaction feature, the semi-global interaction feature, and the global interaction feature. Material property prediction system.

7. In claim 6, The above property prediction unit inputs the integrated feature vector into a multilayer perceptron to predict the target property. Material property prediction system.

8. In claim 1, The above preprocessing module receives a crystallographic information file (CIF) as input and obtains the atomic structure data. Material property prediction system.

9. A computerized method for predicting the properties of materials, A step of receiving a crystallographic information file (CIF) as atomic structure data of a material and generating input data including first input data for local interaction calculation, second input data for quasi-global interaction calculation, and third input data for global interaction calculation; A step of extracting local interaction features by calculating local interactions between a specific atom and its surrounding neighboring atoms based on the first input data; A step of extracting quasi-global interaction features by calculating interactions between all atoms within a unit cell of a decision based on the second input data; A step of extracting global interaction features by calculating long-range interactions considering periodicity between unit cells based on the third input data; and A step of predicting the target properties of a material by integrating the local interaction features, the semi-global interaction features, and the global interaction features; Method for predicting physical properties.

10. In claim 9, The above input data generation step is: Generating the first input data in a graph structure with atoms of the material as nodes and pairs of atoms within a specific cutting radius as edges through a preprocessing module. Method for predicting physical properties.

11. In claim 10, The above local interaction feature extraction step is, Extracting the local interaction features through a message passing operation of a graph neural network for the first input data. Method for predicting physical properties.

12. In claim 9, The above input data generation step is: Generating the second input data including an input sequence listing the feature vectors of atoms in the unit cell in order and a relative distance matrix calculating the distance between all pairs of atoms, Method for predicting physical properties.

13. In claim 12, The above semi-global interaction feature extraction step is: The semi-global interaction features are extracted by applying the self-attention mechanism of the transformer architecture to the second input data, and the relative distance matrix is ​​used as an attention bias. Method for predicting physical properties.

14. In claim 9, The above input data generation step is: Generating the third input data of an extended graph structure connected to neighboring atoms beyond the boundary of the unit cell, Method for predicting physical properties.

15. In claim 14, The above global interaction feature extraction step is: Applying a graph neural network to the extended graph structure of the third input data to extract the global interaction features including the interactions between unit cells. Method for predicting physical properties.

16. In claim 9, The above target property prediction step is: By combining vectors representing the local interaction feature, the semi-global interaction feature, and the global interaction feature respectively, an integrated feature vector is generated, and the integrated feature vector is passed through a fully connected neural network to predict the target property. Method for predicting physical properties.

17. In a service system that predicts the properties of materials, A user computing device that transmits an analysis request for atomic structure data of a material and receives predicted material property values; and A server computing system comprising: a preprocessing module that is communicatively connected to the user computing device and receives the atomic structure data to generate first input data for a local interaction operation, second input data for a quasi-global interaction operation, and third input data for a global interaction operation; a multi-scale interaction operation unit that extracts local interaction features based on the first input data, quasi-global interaction features based on the second input data, and global interaction features based on the third input data; and a property prediction unit that predicts target properties of a material by integrating each of the extracted interaction features; Material properties prediction service system.

18. In claim 17, The above multi-scale interaction operation unit is, A local interaction operation unit that extracts the local interaction features by applying a message passing operation of a graph neural network to the first input data; A quasi-global interaction operation unit that extracts the quasi-global interaction features by applying the self-attention of the transformer to the second input data; and A global interaction operation unit that extracts the global interaction features by applying a message passing operation of a graph neural network to the third input data; including, Material properties prediction service system.

19. In claim 17, The above property prediction unit is, The integrated feature vector is generated by combining the respective feature vectors representing the local, semi-global, and global interaction features, and the integrated feature vector is input into a multilayer perceptron to predict the target property. Material properties prediction service system.

20. In an application-specific integrated circuit for predicting the properties of a material, the application comprises a memory storing information, commands and artificial intelligence models and a functional block including at least one processor requesting access to the memory, The above memory receives a crystallographic information file as atomic structure data of a material, generates input data including first input data for local interaction calculation, second input data for quasi-global interaction calculation, and third input data for global interaction calculation, calculates local interactions between specific atoms and their surrounding neighboring atoms based on the first input data to extract local interaction features, calculates interactions between all atoms in a unit cell of a crystal based on the second input data to extract quasi-global interaction features, calculates long-range interactions considering periodicity between unit cells based on the third input data to extract global interaction features, and stores an artificial intelligence model, command, or information that predicts target properties of a material by integrating the local interaction features, the quasi-global interaction features, and the global interaction features. Custom integrated circuits.

Images