Method for generating chemical structure, neural network device and non-transitory computer-readable recording medium

Abstract

The present invention relates to a method for generating a chemical structure, a neural network device and a non-transitory computer-readable recording medium. A new chemical structure is generated by using an expression region expressing a specific property in a descriptor or image of a reference chemical structure using a neural network. The new chemical structure can be generated by changing a local structure corresponding to the expression region in the reference chemical structure.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims the priority of Korean Patent Application No. 10-2018-0098373 filed on August 23, 2018 in the Korean Intellectual Property Office, the disclosure of which is incorporated herein by reference in its entirety. Technical Field

[0003] The present disclosure relates to methods and apparatus for generating chemical structures using neural networks. Background Art

[0004] Neural network refers to a computing architecture that imitates (is modeled after) the biological brain. Using advanced neural network technology, various types of electronic systems have been developed by using neural networks to analyze input data and generate optimized information.

[0005] In recent years, a lot of research has been conducted on methods for selecting chemical structures to be used in material development by evaluating the properties of chemical structures using neural network technology. In particular, there is a need to develop a method for generating new chemical structures that meet various requirements using neural network technology. Summary of the invention

[0006] Embodiments of the present disclosure relate to methods and apparatus for generating chemical structures using neural networks. In addition, a computer-readable recording medium is provided, which includes a program that implements the method when executed by a computer. The technical problems to be solved are not limited to those described, but other technical problems may exist.

[0007] Additional aspects will be set forth in part in the description which follows and, in part, will be apparent from the description, or may be learned by practice of the presented embodiments.

[0008] According to aspects of an embodiment, a method for generating a chemical structure by using a neural network device is provided, comprising: inputting a descriptor of a chemical structure into a trained neural network, the trained neural network generating a property value of a property of the chemical structure, wherein the descriptor of the chemical structure represents a structural characteristic of the chemical structure and the property of the chemical structure is a characteristic possessed by the chemical structure; determining an expression region in the descriptor for expressing the property, the expression region including a bit position in the descriptor; and generating a new chemical structure by modifying a local structure in the chemical structure, the local structure corresponding to the expression region.

[0009] Determining the expression region may include determining the expression region in the descriptor for expressing the property by performing an interpretation process by the trained neural network to determine whether the property value is expressed by the local structure in the chemical structure.

[0010] Determining the expression region may include determining the expression region in the descriptor for expressing the property by applying a layer-wise relevance propagation (LRP) technique to the trained neural network, wherein an activation function applied to a node of the trained neural network may be selected (specified) as a linear function to apply the LRP technique to the trained neural network, and a mean square error (MSE) may be selected for optimization.

[0011] Generating a new chemical structure may include: obtaining a bit value of the bit position of the expression region in the descriptor; and generating the new chemical structure by applying a genetic algorithm to the bit value of the bit position and modifying the local structure corresponding to the expression region.

[0012] Generating a new chemical structure may include: generating a new first chemical structure by modifying the local structure in the chemical structure, the local structure corresponding to the expression region; inputting a descriptor for the new first chemical structure into the trained neural network to output a property value for a specific property of the new first chemical structure; and when the property value for the specific property of the new first chemical structure is less than a preset value, generating a new second chemical structure by changing the local structure in the new first chemical structure, the local structure corresponding to the expression region, and when the property value for the specific property of the new first chemical structure is equal to or greater than the preset value, storing the new first chemical structure.

[0013] According to an aspect of an embodiment, a neural network device configured to generate a chemical structure is provided, comprising: a memory configured to store at least one (kind of) program; and a processor configured to drive the neural network by executing the at least one program, wherein the processor is configured to: input a descriptor of a chemical structure into a trained neural network, the trained neural network generating a property value of a property of the chemical structure, the descriptor of the chemical structure representing a structural characteristic of the chemical structure and the property of the chemical structure being a characteristic possessed by the chemical structure; determine an expression region in the descriptor for expressing the property, the expression region comprising a bit position in the descriptor; and generate a new chemical structure by modifying a local structure in the chemical structure, the local structure corresponding to the expression region.

[0014] According to aspects of an embodiment, a method for generating a chemical structure by using a neural network device is provided, comprising: inputting an image of a chemical structure into a trained neural network, the trained neural network generating a property value of a property of the chemical structure, the image of the chemical structure representing the structural characteristics of the chemical structure and the property of the chemical structure being a characteristic possessed by the chemical structure; determining an expression region in the image for expressing the property, the expression region comprising one or more pixels in the image; and generating a new chemical structure by modifying a local structure in the chemical structure, the local structure corresponding to the expression region.

[0015] Determining the expression region may include determining the expression region in the image for expressing the property by performing an interpretation process by the trained neural network to determine whether the property value is expressed by the local structure in the chemical structure.

[0016] Determining the expression area may include determining the expression area in the image for expressing the property by applying a layer-by-layer relevance propagation (LRP) technique to the trained neural network, wherein an activation function applied to a node of the trained neural network may be selected as a linear function to apply the LRP technique to the trained neural network, and a mean square error (MSE) may be selected for optimization.

[0017] Generating a new chemical structure may include: obtaining pixel values ​​of the one or more pixels in the expression area in the image; and generating the new chemical structure by applying Gaussian noise to the pixel values ​​of the one or more pixels and modifying the local structure corresponding to the expression area.

[0018] Generating a new chemical structure may include: when multiple expression regions expressing the property exist in the image, obtaining coordinate information corresponding to the multiple expression regions in the image; calculating the center points of the multiple expression regions in the image based on the coordinate information and obtaining the pixel values ​​of the center points; and generating the new chemical structure by applying Gaussian noise to the pixel values ​​and modifying the local structure corresponding to the center points.

[0019] Generating a new chemical structure may include: generating a new first chemical structure by modifying the local structure in the chemical structure, the local structure corresponding to the expression region; inputting an image of the new first chemical structure into the trained neural network to output a property value for a specific property of the new first chemical structure; and when the property value for the specific property of the new first chemical structure is less than a preset value, generating a new second chemical structure by changing the local structure in the new first chemical structure, the local structure corresponding to the expression region, and when the property value for the specific property of the new first chemical structure is equal to or greater than the preset value, storing the new first chemical structure.

[0020] According to aspects of the embodiments, a neural network device configured to generate a chemical structure is provided, comprising: a memory configured to store at least one program; and a processor configured to drive the neural network by executing the at least one program, wherein the processor is configured to: input an image of a chemical structure into a trained neural network, the trained neural network generating a property value of a property of the chemical structure, the image of the chemical structure representing the structural characteristics of the chemical structure and the property of the chemical structure being a characteristic possessed by the chemical structure; determine an expression region in the image for expressing the property, the expression region comprising one or more pixels in the image; and generate a new chemical structure by modifying a local structure in the chemical structure, the local structure corresponding to the expression region.

[0021] According to an aspect of the embodiment, there is provided a non-transitory computer-readable recording medium including a program which, when executed by a computer, carries out any one of the methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] These and / or other aspects will become apparent and more readily understood from the following description of the embodiments considered in conjunction with the accompanying drawings, in which:

[0023] Figure 1 A block diagram illustrating a hardware configuration of a neural network device according to an embodiment;

[0024] Figure 2 A diagram for illustrating calculations performed by a deep neural network (DNN) according to an embodiment;

[0025] Figure 3 A diagram illustrating calculations performed by a recurrent neural network (RNN) according to an embodiment;

[0026] Figure 4 A conceptual diagram for explaining a neural network system for generating chemical structures according to an embodiment;

[0027] Figure 5 A diagram illustrating a method of representing a chemical structure according to an embodiment;

[0028] Figure 6 A diagram for explaining a method of interpreting a neural network according to an embodiment;

[0029] Figure 7 A diagram illustrating an example of changing an expression region of a descriptor to generate a new chemical structure according to an embodiment;

[0030] Figure 8 A diagram for explaining an example of changing a local structure by changing a bit value of a descriptor according to an embodiment;

[0031] Fig. 9 A diagram for explaining an example of changing a local structure by changing a pixel value of an image according to an embodiment;

[0032] Fig.10 A diagram for explaining an example of changing a pixel value when a plurality of expression areas exist on an image according to an embodiment;

[0033] Fig.11 A flowchart of a method for generating a new chemical structure by changing a descriptor of the chemical structure in a neural network device according to an embodiment; and

[0034] Fig.12 Flow chart of a method for generating a new chemical structure by changing an image of the chemical structure in a neural network device according to an embodiment. DETAILED DESCRIPTION

[0035] Embodiments will now be described in detail, examples of which are shown in the accompanying drawings, wherein the same reference numerals always refer to the same elements. In this regard, the present embodiment may have different forms and should not be construed as being limited to the description set forth herein. Therefore, the following description of the embodiment is to illustrate aspects only by referring to the accompanying drawings. As used herein, the term "and / or" includes any and all combinations of one or more of the related enumerated items. Statements such as "at least one (kind) of ..." modify the entire list of elements when before or after the list of elements without modifying the individual elements of the list, so that the statement of "at least one (kind) of a, b, and c" or similar statements thereof include: only a, only b, only c, only a and b, only b and c, only a and c, and all of a, b, and c.

[0036] The terms "according to some embodiments" or "according to an embodiment" used throughout the specification do not necessarily refer to the same embodiment.

[0037] Some embodiments of the present disclosure can be described by function block (frame) construction and various processing operations. Some or all of these function blocks can be implemented using various numbers of hardware and / or software components (components, ingredients) that perform specific functions. For example, the function blocks of the present disclosure can be implemented using one or more microprocessors or circuits that execute instructions to implement a given function. In addition, for example, the function blocks of the present disclosure can be implemented in a variety of programming or scripting languages. The function blocks can be implemented using algorithms executed by one or more processors. The present disclosure can also use conventional techniques for electronic construction, signal processing, and / or data processing. The terms "mechanism", "element", "unit" and "construction" can be used in a broad sense and are not limited to mechanical and physical constructions.

[0038] In addition, the connecting lines or connecting members between components shown in the drawings only illustrate functional connections and / or physical or circuit connections. In actual equipment, the connections between components can be provided by various functional connections, physical connections, or circuit connections that can be replaced or added.

[0039] Meanwhile, regarding the terms used herein, the descriptor as data used in the neural network system refers to an indicative value for describing the structural characteristics (features) of a chemical structure and can be obtained by performing relatively simple calculations on a given chemical structure. According to an embodiment, the descriptor may include a molecular structure fingerprint (e.g., Morgan fingerprint and extended connectivity fingerprint (ECFP)) indicating whether a specific local structure is included. In addition, the descriptor may be a quantitative structure-property relationship (QSPR) model configured with a value that can be immediately calculated from a given chemical structure, such as a molecular weight or the number of local structures (e.g., rings) included in the molecular structure.

[0040] In addition, properties refer to characteristics possessed by a chemical structure and may be real values ​​measured by experiments or may be calculated by simulation. For example, when the chemical structure is used as a display (display) material, the property of the chemical structure may be a transmission wavelength, an emission wavelength, etc. for light. When the substance is used as a battery material, the property of the chemical structure may be voltage. Unlike descriptors, calculation of properties may require complex simulations, which require additional calculations and computations beyond similar simulations for descriptors.

[0041] In addition, structure refers to the atomic level structure of a chemical structure. In order to derive (derive) properties by performing first principle calculations, the structure needs to be expressed at the atomic level. Therefore, the atomic level structure needs to be derived to generate a new chemical structure. The structure can be a structural formula based on atomic bonding relationships or a string in a simple format (one-dimensional). The format of the string expressing the structure can be a simplified molecular linear input system (specification) (SMILES) code, a SmilesArbitrary Target Specification (SMARTS) code, an International Chemical Identification (InChi) code, etc.

[0042] In addition, factors refer to elements that define the relationship between descriptors, properties, and structures. Factors can be determined by machine learning based on descriptor-property-structure formulas stored in a database. Therefore, it can be determined how the relationship between factors, descriptors, properties, and structures is.

[0043] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings.

[0044] Figure 1 1 is a block diagram illustrating a hardware configuration of a neural network device 100 according to an embodiment.

[0045] The neural network device 100 may be implemented using various types of devices such as personal computers (PCs), servers, mobile devices, and embedded devices. Examples of the neural network device 100 may include, but are not limited to, smartphones, tablet devices, augmented reality (AR) devices, Internet of Things (IoT) devices, autonomous vehicles, robots, medical devices, etc. that use neural networks to perform speech recognition, image recognition, image classification, etc. In addition, the neural network device 100 may be a dedicated hardware (HW) accelerator disposed on, connected to, or installed in the devices described above. The neural network device 100 may be a hardware accelerator such as a neural processor (processing unit) (neural network processor) (NPU), a tensor processor (TPU), or a neural engine as a dedicated module (component) for driving a neural network, but is not limited thereto.

[0046] Reference Figure 1 , the neural network device 100 includes a processor 110 and a memory 120 . Figure 1 Only the components of the neural network device 100 related to the embodiments of the present disclosure are described. Therefore, it is clear to those skilled in the art that the neural network device 100 may further include Figure 1 Any other generic components other than those shown in .

[0047] The processor 110 controls the overall functions for driving the neural network device 100. For example, the processor 110 controls the overall operation of the neural network device 100 by executing a program stored in the memory 120 of the neural network device 100. The processor 110 may be implemented as a central processing unit (CPU), a graphics processing unit (GPU), an application processor (AP), etc. provided in the neural network device 100, but is not limited thereto.

[0048] Memory 120 is a hardware component that stores various data processed in neural network device 100. For example, memory 120 may store data processed by neural network device 100 and to be processed by neural network device 100. Memory 120 may also store applications, drivers, etc. to be executed by processor 110 of neural network device 100. Memory 120 may include random access memory (RAM) such as dynamic random access memory (DRAM) and static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), CD-ROM, Blu-ray or other optical disk storage, hard disk drive (HDD), solid state drive (SSD), or flash memory (flash memory).

[0049] The memory 120 may store the descriptors for the chemical structures and the property values ​​representing the properties of the chemical structures in numbers that match each other as a group or a pair. The neural network device 100 may read the descriptors and the property values ​​corresponding thereto from the memory 120 or write the descriptors and the property values ​​corresponding thereto into the memory 120. In an embodiment, the descriptors may include a plurality of bit values, and the property values ​​may be values ​​for a transmission wavelength, an emission wavelength, a voltage, etc.

[0050] Although not in Figure 1 , but the memory 120 may store the image of the chemical structure and the property value representing the property of the chemical structure in numbers in association with each other as a group or pair. In an embodiment, the image may include n×m pixels (where n and m are natural numbers). In the following, the description of the descriptor is also applicable to the case where the descriptor is replaced by the image.

[0051] The memory 120 may store the structural feature value representing the chemical structure and the descriptor and property value matching the structural feature value as a group or pair. The structural feature value may be a SMILES code or a SMARTS code which is a string format expressing the chemical structure.

[0052] The processor 110 may execute instructions to implement an artificial neural network (ANN), such as a deep neural network (DNN) and a recurrent neural network (RNN).

[0053] The processor 110 may allow the DNN to learn by using the descriptors and the property values ​​corresponding to the descriptors, and in the process may determine the factors defining the relationship between the descriptors and the property values. Then, the processor 110 may output the property values ​​corresponding to the new descriptors as output data by driving the trained DNN by using the new descriptors that have not been used in the learning process of the DNN as input data.

[0054] The processor 110 may allow the RNN to learn by using the descriptor and the structural feature value, and in the process may determine a factor defining the relationship between the descriptor and the structural feature value. Then, the processor 110 may output the structural feature value corresponding to the new descriptor as output data by driving the trained RNN by using the new descriptor that has not been used in the learning process of the RNN as input data.

[0055] The neural network device 100 may further include a user interface. The user interface refers to a device or software for inputting data to control the neural network device 100. Examples of the user interface may include, but are not limited to, a keypad, a dome switch, a touch pad (e.g., a capacitive superposition type, a resistive superposition type, an infrared beam type, a surface acoustic wave type, an integral strain gauge type, and a piezoelectric type), a jog wheel, and a jog switch, and a graphical user interface (GUI) that may be displayed to receive user input.

[0056] Figure 2 A diagram illustrating calculations performed by a DNN according to an embodiment.

[0057] Reference Figure 2 The DNN 20 may have a structure including an input layer, a hidden layer, and an output layer, and may generate a plurality of output layers based on the received input data (e.g., I 1 and I 2 ) performs calculations and generates output data (e.g., 1 and O 2 ).

[0058] For example, Figure 2 As shown in FIG. 1 , DNN 20 may include an input layer (layer 1), two hidden layers (layers 2 and 3), and an output layer (layer 4). Since DNN 20 may include many layers for processing effective information, DNN 20 may process complex data compared to a neural network including a single layer. At the same time, although Figure 2The DNN 20 shown in FIG. 1 includes 4 layers, but the DNN 20 is only an example and may also include more or fewer layers and more or fewer channels than those shown therein. That is, the DNN 20 may have Figure 2 The various layer structures shown in FIG.

[0059] Each layer included in the DNN 20 may have a plurality of channels. The channels may correspond to a plurality of artificial nodes, which are referred to as neurons, processing elements (PEs), units, or similar terms. For example, Figure 2 As shown in FIG, layer 1 may include two channels (nodes), and layers 2 and 3 may include three channels, respectively. However, the layers are merely examples, and each of the layers included in DNN 20 may have various numbers of channels (nodes) and interconnections with other nodes.

[0060] Channels included in each of the layers of the DNN 20 may be interconnected to process data. For example, a channel may perform calculations on data received from a channel of one layer and output calculation results to a channel of another layer.

[0061] The input and output of each channel may be referred to as input activation and output activation. That is, the activation may not only be the output of one channel, but also a parameter corresponding to the input of the channel included in the successive layer. At the same time, each of the channels may determine its activation based on the activation and weight received from the channel included in the previous layer. The weight is a parameter for calculating the output activation of each channel and may be a value assigned to the relationship between the channels.

[0062] Each of the channels can be processed by a computational unit or processing element that receives input and produces output activation. The input-output of each channel can be plotted. For example, when σ is an activation function, is the weight from the kth channel included in the (i-1)th layer to the jth channel included in the i-th layer, is the bias of the jth channel included in the i-th layer, and When the activation of the jth channel of the i-th layer is used, the activation can be calculated using the following equation 1

[0063] [Equation 1]

[0064]

[0065] like Figure 2 As shown in FIG. 1 , the activation of the first channel CH1 of the second layer 2 can be performed as In addition, according to Equation 1, Can have In Equation 1, σ represents an activation function such as Relu, sigmoid (S-shaped), and tanh (hyperbolic tangent). As a result, the activation of a specific channel in a specific layer can be represented by using The value of is obtained by passing the activation function.

[0066] However, the above Equation 1 is merely an example for describing activations and weights for processing data in the DNN 20, and embodiments are not limited thereto.

[0067] In an embodiment, the neural network device 100 may allow the DNN 20 to learn by using the descriptors (or images) and property values ​​stored in the memory. The DNN 20 may determine a factor defining a relationship between the descriptor (or image) and the property value during the learning process using the descriptors and the property values.

[0068] That is, among layers 1 to 4 constituting DNN 20, the descriptor (or image) may correspond to the values ​​of multiple channels (nodes) of the input layer (layer 1), the property value may correspond to the values ​​of multiple channels (nodes) of the output layer (layer 4), and the factor may correspond to the values ​​of multiple channels (nodes) of at least one hidden layer (layer 2 and / or 3).

[0069] The trained DNN 20 may then be driven by receiving a new descriptor (or a new image) as input data, and may thus output a property value corresponding to the received new descriptor (or new image) as output data.

[0070] Figure 3 A diagram illustrating calculations performed by an RNN according to an embodiment.

[0071] In the following, for the convenience of description, the above reference will not be repeated. Figure 2 Description given.

[0072] The RNN 30 is a neural network that analyzes data that changes over time, such as time series data, and is constructed by connecting networks between a reference time point t and a next time point t + 1. That is, the RNN 30 is a neural network in which a time aspect is taken into account and is capable of efficiently learning patterns from data that is input sequentially or data that is input in a feature sequence by modifying the model to allow recursive input to a hidden layer of the neural network.

[0073] Reference Figure 3 , showing the node s constituting the hidden layer of the RNN 30. The node s can perform calculations based on the input data x and generate output data o. The RNN 30 can iteratively apply the same task to all sequences, and the final output result of the node s is affected by the result of the previous calculation.

[0074] RNN 31 is an unfolded RNN 30 with loops. The term “unfolded” with respect to RNN 30 refers to expressing RNN 30 for the entire sequence. In RNN 31, x t is the input value at time step t, and s t is the hidden state at time step t. The term s t It can be expressed by the following equation 2. In equation 2, a tanh or Relu function can be used as the function f. The s used to calculate the first hidden state t-1 can usually be initialized to 0. In addition, in RNN 31, o t is the output value at time step t.

[0075] [Equation 2]

[0076]

[0077] Here, s t is the memory part of the network and stores information about events at previous time steps. Output value o t Depends only on the storage of the current time step t.

[0078] Meanwhile, unlike the existing neural network structure in which parameters are different from each other, RNN 31 shares parameters U, V, and W for all time steps. That is, since each step of RNN 31 performs almost the same calculation except for the input value, the number of parameters to be learned can be reduced.

[0079] In an embodiment, the neural network device 100 may allow the RNN 31 to learn by using descriptors (or images) and property values ​​stored in the memory. Alternatively, the neural network device 100 may allow the RNN 31 to learn by using factors and property values ​​determined during the learning process of the DNN 20.

[0080] For example, when W of RNN 31 is a factor determined in the learning process of DNN 20 and the structural feature value represented by the SMILES code is "ABCD", t-1 and x t Can be "ABC", and O t and x t+1 It can be "BCD". Then, the SMILES codes in each time step can be combined to output a SMILES code "ABCDEFG", that is, a structural feature value as output data.

[0081] Then, the trained RNN 31 may be driven by receiving a new descriptor (or a new image) as input data, and thus may output a structural feature value corresponding to the received new descriptor (or a new image) as output data. Alternatively, the trained RNN 31 may be driven by receiving a factor for a new descriptor (or a new image) as input data, and thus may output a structural feature value corresponding to the new descriptor (or a new image) as output data.

[0082] Figure 4 A conceptual diagram illustrating a neural network system for generating chemical structures according to an embodiment.

[0083] Reference Figure 4 , illustrating a neural network system configured to generate chemical structures by using a DNN 410 and a RNN 420.

[0084] The descriptor of data used in the neural network system may be represented by an ECFP which is an indication value for representing a structural property of a chemical structure. Property refers to a property possessed by a chemical structure and may be a real value indicating a transmission wavelength and an emission wavelength with respect to light. Structure refers to an atomic level structure of a chemical structure and may be represented by a SMILES code. For example, a structural formula may be expressed according to a SMILES code as shown in the following equation 3.

[0085] [Equation 3]

[0086] OCl=C(C=C2C=CNC2=Cl)Cl=C(C=CC=Cl)Cl=CC2=C(NC=C2)C=Cl

[0087] A factor is an element defining the relationship between a descriptor, a property, and a structure. A factor may be at least one hidden layer. When a factor includes a plurality of hidden layers, a factor defining the relationship between a descriptor and a property, a factor defining the relationship between a descriptor and a structure, etc. may be determined for each hidden layer.

[0088] The DNN 410 may be driven by receiving a descriptor as input data, and outputs a property value corresponding to the received descriptor as output data. In a learning process using the descriptor and the property value, the DNN 410 may determine a factor defining a relationship between the descriptor and the property value. The RNN 420 may be driven by receiving a factor or descriptor determined in a learning process of the DNN 410 as input data, and outputs a structural feature value as output data.

[0089] Figure 5 is a diagram illustrating a method of representing a chemical structure 510 according to an embodiment.

[0090] Reference Figure 5 , the chemical structure 510 represents the shape of a molecule formed by the combination of atoms. The chemical structure 510 can be represented by the positions of atoms, the distances between atoms, the strength of atomic bonds, etc.

[0091] In one embodiment, the chemical structure 510 may be represented by a descriptor 520 including a plurality of bit values ​​(1 or 0). Through the descriptor 520, it may be determined whether the chemical structure 510 includes a specific partial structure.

[0092] In another embodiment, the chemical structure 510 may be represented as an image 530 having a certain size. The image 530 may include three channels (red, green, and blue (RGB)) of n×m pixels (where n and m are natural numbers). In the image, 8 bits, i.e., values ​​from 0 (black) to 255 (white), may be assigned to each pixel of the image 530. For example, a bright red color may be synthesized with an R channel value of 246, a G channel value of 20, and a B channel value of 50, and when all channel values ​​are 255, white is synthesized.

[0093] Hereinafter, for convenience of description, a method in which the chemical structure 510 is displayed as the image 530 by using one channel will be described.

[0094] Atoms constituting the chemical structure 510 may be displayed in different colors from each other on the image 530. The chemical structure 510 may include carbon (C), nitrogen (N), and oxygen (O), and on the image 530, carbon (C) may be displayed in black, nitrogen (N) may be displayed in blue, and oxygen (O) may be displayed in red.

[0095] Reference Figure 5 , on the image 530 including 6×6 pixels, the value of the pixel where carbon (C) of the chemical structure 510 is located may be “0”, the value of the pixel where nitrogen (N) is located may be “50”, and the value of the pixel where oxygen (O) is located may be “186”. The value of the pixel where there is no atom may be “255”.

[0096] The type of color in which a certain atom is displayed on the image 530, the number of pixels constituting the image 530, and the like are not limited to the above examples.

[0097] The descriptor 520 or the image 530 for the chemical structure 510 may be used as input data for the neural network, and a specific property value for the chemical structure 510 may be output as output data of the neural network.

[0098] Figure 6 A diagram for explaining a method of interpreting a neural network according to an embodiment.

[0099] The neural network device 100 may obtain a descriptor or image for a reference chemical structure to output a specific property value for the reference chemical structure. In an embodiment, the descriptor may include multiple bit values, and the image may include n×m pixels (where n and m are natural numbers).

[0100] The neural network device 100 may input the descriptor or image of the reference chemical structure into a trained neural network as input data and drive the neural network through an inference process 610 to obtain a specific property value for the reference chemical structure as output data of the neural network.

[0101] In this case, neural network device 100 may perform interpretation process 620 to determine whether a particular property value is expressed by a local structure in the reference chemical structure.

[0102] Reference Figure 6 In an embodiment, the neural network device 100 may interpret the trained neural network by using a layer-by-layer relevance propagation (LRP) technique. The LRP technique is a method of propagating relevance in the reverse direction of the trained neural network (i.e., the direction from the output layer to the input layer). In the LRP technique, when the relevance propagates between layers, a node having the maximum relevance with the upper layer among a plurality of nodes in the lower layer obtains the maximum relevance from the corresponding node in the upper layer.

[0103] The method of calculating the correlation in the LRP technique can be expressed by Equation 4. In Equation 4, a i and a j are the output value to be determined in a specific node in the i-th layer and the output value to be determined in a specific node in the j-th layer, respectively. + ij is the weight value that links the specific node of the i-th layer with the specific node of the j-th layer. i and R j They respectively represent the relevance of the specific node in the i-th layer and the relevance of the specific node in the j-th layer.

[0104] [Equation 4]

[0105]

[0106] In an embodiment, for application of the LRP technique, the neural network device 100 may select an activation function applied to a node of a trained neural network as a linear function by using a regression analysis method, and may select a mean square error (MSE) for optimization. Specifically, in the regression analysis method, since the final output value may include several integer values, the neural network may be trained by selecting the activation function of the output node as a linear function. In order to implement the regression analysis method, the loss function may be selected as the MSE in the neural network learning process.

[0107] However, the techniques that may be used in interpretation process 620 to determine whether a particular property value is expressed by any local structure in the reference chemical structure are not limited to the examples described above.

[0108] When the input data of the neural network is a descriptor for the reference chemical structure, the plurality of nodes of the input layer may respectively correspond to the bit values ​​constituting the descriptor. The neural network device 100 may obtain the node of the input layer (i.e., the bit position of the descriptor) having the greatest correlation with the expression of the specific property value of the reference chemical structure through the interpretation process 620. Since the bit position of the descriptor corresponds to a specific local structure in the reference chemical structure, the neural network device 100 may determine the specific local structure having the greatest correlation with the expression of the specific property value of the reference chemical structure by obtaining the bit position of the descriptor through the interpretation process 620.

[0109] When the input data of the neural network is an image of the reference chemical structure, the plurality of nodes of the input layer may respectively correspond to pixel values ​​constituting the image. The neural network device 100 may obtain the node of the input layer (i.e., the pixel coordinates of the image) having the greatest correlation with the expression of the specific property value of the reference chemical structure through the interpretation process 620. Since the pixel coordinates of the image correspond to the specific local structure in the reference chemical structure, the neural network device 100 may determine the specific local structure having the greatest correlation with the expression of the specific property value of the reference chemical structure by obtaining the pixel coordinates of the image through the interpretation process 620.

[0110] In the following, the bit positions of the descriptors and the pixel coordinates of the image that have the greatest correlation with the expression of the specific property value of the reference chemical structure will be referred to as the expression region.

[0111] Figure 7 A diagram illustrating an example of changing an expression region of a descriptor to generate a new chemical structure according to an embodiment.

[0112] Reference Figure 7, the descriptor 712 of the reference chemical structure 710 may be “11100011011010110”. The neural network device 100 may sequentially input the bit values ​​constituting the descriptor 712 to the nodes of the input layer of the neural network (e.g., DNN), respectively, and output the property value for the reference chemical structure 710 (i.e., “, emission wavelength: 320 nm”).

[0113] The neural network device 100 may obtain a node of the input layer (i.e., an expression region 713 of the descriptor 712) having the greatest correlation with the expression of the wavelength value of the reference chemical structure 710. The expression region 713 of the descriptor 712 may correspond to a specific position 711 in the reference chemical structure 710. Figure 7 In the example, the expression region 713 corresponds to one bit value. However, the expression region 713 may correspond to multiple consecutive bit values, and multiple expression regions may be in the descriptor 712.

[0114] The neural network device 100 may change the bit value of the expression region 713 to improve the properties of the reference chemical structure 710. When the bit value of the expression region 713 is changed, the structure of the specific position 711 may be changed. As a method of changing the bit value of the expression region 713, a genetic algorithm may be used, and then the reference chemical structure 710 may be changed. Figure 8 The neural network device 100 may change the bit value of the expression region 713 and / or the bit value around the expression region 713 .

[0115] The neural network device 100 may change the bit value of the expression region 713 and output a new descriptor 722. Figure 7 , when the bit value “1” of the expression area 713 in the new descriptor 722 is changed to “110100”, the neural network device 100 may apply the new local structure 721 corresponding to the bit value “110100” to the specific position 711 and generate a new chemical structure 720, and the new local structure 721 is applied to the new chemical structure 720.

[0116] Regarding the method of generating the new chemical structure 720, the neural network device 100 may input the new descriptor 722 as input data of a neural network (eg, RNN) and output a structural feature value as output data, and may generate the new chemical structure 720 based on the output structural feature value.

[0117] The neural network device 100 may input the descriptor 722 of the new chemical structure 720 into the neural network and output a property value (i.e., "emission wavelength: 325 nm") corresponding to the descriptor 722 input into the neural network. That is, the neural network device 100 may improve the property by changing the local structure of the reference chemical structure 710 and generating the new chemical structure 720.

[0118] The neural network device 100 can repeatedly generate a chemical structure through the above process until a chemical structure having a property value close to a preset value (eg, "emission wavelength: 350 nm") is generated.

[0119] Specifically, the neural network device 100 may compare the property value for the new chemical structure 720 (e.g., "emission wavelength: 325nm") with a preset value (e.g., "emission wavelength: 350nm"), and when the property value for the new chemical structure 720 is less than the preset value, generate a new chemical structure by changing the bit value of the expression area 723 of the descriptor 722.

[0120] When the property value for the new chemical structure generated through the above process is equal to or greater than the preset value, the neural network device 100 may store the generated new chemical structure in a memory.

[0121] Figure 8 A diagram for explaining an example of changing a local structure by changing a bit value of a descriptor according to an embodiment.

[0122] In an embodiment, the neural network device 100 may apply a genetic algorithm to the bit values ​​constituting the descriptor of the reference chemical structure, and perform a selection, crossover (overlap), or mutation operation (run) on the bit values.

[0123] The neural network device 100 may change the descriptor of the reference chemical structure by applying a genetic algorithm to the bit values ​​constituting the descriptor of the reference chemical structure. When the descriptor of the reference chemical structure is changed or modified, a local structure in the reference chemical structure may be mutated, removed, or replaced, or a local structure may be added to the reference chemical structure.

[0124] Reference Figure 8 , the neural network device 100 may mutate a local structure in the reference chemical structure by applying a genetic algorithm to the bit values ​​of the descriptors constituting the reference chemical structure. For example, the neural network device 100 may change the carbon (C) in the first position 810 in the reference chemical structure to nitrogen (N). Alternatively, the neural network device 100 may change the adjacent atoms 811 and 812 combined with the atom in the first position 810 to other atoms.

[0125] In addition, the neural network device 100 may add a local structure to the reference chemical structure by applying a genetic algorithm to the bit values ​​of the descriptors constituting the reference chemical structure. For example, the neural network device 100 may add a local structure 823 to connect to the atom in the second position 820 in the reference chemical structure. Alternatively, the neural network device 100 may add a local structure to connect to adjacent atoms 821 and 822 combined with the atom in the second position 820. Alternatively, the neural network device 100 may add a local structure 824 in the form of a fused ring connected to both the atom in the second position 820 and the adjacent atom 821 combined with the atom in the second position 820.

[0126] In addition, the neural network device 100 may remove the local structure in the reference chemical structure by applying a genetic algorithm to the bit values ​​of the descriptors constituting the reference chemical structure. For example, the neural network device 100 may remove the local structure 831 in the reference chemical structure connected to the atom in the third position 830. Alternatively, the neural network device 100 may change the ring structure by removing the atom in the third position 830.

[0127] In addition, the neural network device 100 can replace the local structure in the reference chemical structure by applying a genetic algorithm to the bit values ​​of the descriptor constituting the reference chemical structure. For example, the neural network device 100 can change the ring structure of the fourth position 840 in the reference chemical structure into a new local structure 841 or 842.

[0128] However, the example of changing the local structure by changing the bit value of the descriptor is not limited to the above description.

[0129] Fig. 9 A diagram for explaining an example of changing a local structure by changing pixel values ​​of an image according to an embodiment.

[0130] Reference Fig. 9 , the image 912 of the reference chemical structure 910 may include 6×6 pixels. The atoms constituting the reference chemical structure 910 may be displayed on the image 912 in different colors from each other. The reference chemical structure 910 may include carbon (C), nitrogen (N), and oxygen (O), and on the image 912, carbon (C) may be displayed in black, nitrogen (N) may be displayed in blue, and oxygen (O) may be displayed in red. On the image 912, the value of the pixel where carbon (C) is located may be "0", the value of the pixel where nitrogen (N) is located may be "50", and the value of the pixel where oxygen (O) is located may be "186".

[0131] The neural network device 100 may sequentially input pixel values ​​constituting the image 912 to nodes of an input layer of a neural network (eg, DNN), respectively, and may thus output a property value for the reference chemical structure 910 (ie, “emission wavelength: 320 nm”).

[0132] The neural network device 100 may obtain a node of the input layer (i.e., an expression region 913 of the image 912) having the greatest correlation with the expression of the wavelength value of the reference chemical structure 910. The expression region 913 of the image 912 may correspond to a specific position 911 in the reference chemical structure 910. Fig. 9 In the image 912, the expression region 913 corresponds to one pixel value. However, the expression region 913 may correspond to a plurality of adjacent pixel values, and a plurality of expression regions may be provided in the image 912.

[0133] The neural network device 100 may change the pixel value of the expression region 913 and / or the pixel value around the expression region 913 to improve the property of the reference chemical structure 910. When the pixel value of the expression region 913 and / or the pixel value around the expression region 913 is changed, the structure of the specific position 911 may be changed. In an embodiment, the pixel value of the expression region 913 and / or the pixel value around the expression region 913 may be changed by using Gaussian noise. Gaussian noise refers to noise whose distribution function of any order is represented by a normal distribution.

[0134] The neural network device 100 may change the pixel values ​​of the expression region 913 and / or the pixel values ​​around the expression region 913, and thus output a new image 922. Fig. 9 When the pixel value of the expression region 913 and / or the pixel value around the expression region 913 is changed in the new image 922, the neural network device 100 may apply a new local structure 921 corresponding to the changed pixel value to the specific position 911 and generate a new chemical structure 920.

[0135] Regarding the method of generating the new chemical structure 920, the neural network device 100 may input the new image 922 as input data of a neural network (e.g., RNN) and output structural feature values ​​as output data, and may generate the new chemical structure 920 based on the output structural feature values.

[0136] The neural network device 100 may input the image 922 of the new chemical structure 920 into the neural network and output a property value (i.e., "emission wavelength: 325 nm") corresponding to the image 922 input into the neural network. That is, the neural network device 100 may improve the property by changing the local structure of the reference chemical structure 910 and generating the new chemical structure 920.

[0137] The neural network device 100 can repeatedly generate a chemical structure through the above process until a chemical structure having a property value close to a preset value (eg, "emission wavelength: 350 nm") is generated.

[0138] Specifically, the neural network device 100 may compare the property value for the new chemical structure 920 (e.g., “emission wavelength: 325nm”) with a preset value (e.g., “emission wavelength: 350nm”), and when the property value for the new chemical structure 920 is less than the preset value, generate a new chemical structure by changing the pixel value of the expression area 913 in the image 922 and / or the pixel value around the expression area 913.

[0139] When the property value for the new chemical structure generated through the above process is equal to or greater than the preset value, the neural network device 100 may store the generated new chemical structure in the memory.

[0140] Fig.10 A diagram for explaining an example of changing a pixel value when a plurality of expression areas exist on an image according to an embodiment.

[0141] Reference Fig.10 , the image 1012 of the reference chemical structure 1010 may include 6×6 pixels. The atoms constituting the reference chemical structure 1010 may be displayed in different colors from each other on the image 1012. For example, on the image 1012, the value of the pixel where carbon (C) is located may be "0", the value of the pixel where nitrogen (N) is located may be "50", and the value of the pixel where oxygen (O) is located may be "186".

[0142] The neural network device 100 may sequentially input pixel values ​​constituting the image 1012 and output property values ​​(ie, “emission wavelength: 320 nm”) for the reference chemical structure 1010 to nodes of an input layer of a neural network (eg, DNN).

[0143] In an embodiment, there may be a plurality of nodes of the input layer having the greatest correlation with the expression of the wavelength value of the reference chemical structure 1010, or a high correlation relative to other nodes. That is, there may be a plurality of expression regions on the image 1012, namely, a first expression region 1013a and a second expression region 1013b. The first expression region 1013a and the second expression region 1013b in the image 1012 may correspond to the first position 1011a and the second position 1011b in the reference chemical structure 1010, respectively. Fig.10 As shown in , the second position 1011b corresponding to the second expression region 1013b may be outside the reference chemical structure 1010.

[0144] When there are multiple expression regions, namely, a first expression region 1013a and a second expression region 1013b, on the image 1012, the neural network device 100 may obtain coordinate information corresponding to the multiple expression regions, namely, the first expression region 1013a and the second expression region 1013b on the image 1012. For example, based on the lower left corner of the image 1012 having the origin (0,0), the coordinate information of the first expression region 1013a may be (3,3) and the coordinate information of the second expression region 1013b may be (5,3).

[0145] The neural network device 100 may output the coordinate information (4, 3) of the center point 1014 based on the coordinate information corresponding to the plurality of expression regions, namely the first expression region 1013a and the second expression region 1013b. The neural network device 100 may change the pixel value of the center point 1014 and / or the pixel value around the center point 1014 to improve the property of the reference chemical structure 1010. When the pixel value of the center point 1014 and / or the pixel value around the center point 1014 is changed, the structure of a specific position 1015 corresponding to the center point 1014 on the reference chemical structure 1010 may be changed. In an embodiment, the pixel value of the center point 1014 and / or the pixel value around the center point 1014 may be changed by using Gaussian noise.

[0146] The neural network device 100 may change the pixel value of the center point 1014 and / or the pixel values ​​around the center point 1014 and output a new image 1022. Fig.10 When the pixel value of the center point 1014 and / or the pixel values ​​around the center point 1014 in the new image 1022 are changed, the neural network device 100 may apply a new local structure 1021 corresponding to the changed pixel value to the specific position 1015 and generate a new chemical structure 1020.

[0147] The neural network device 100 may input the image 1022 of the new chemical structure 1020 into the neural network and output a property value (i.e., "emission wavelength: 325 nm") corresponding to the image 1022 input into the neural network. That is, the neural network device 100 may improve the property by changing the local structure of the reference chemical structure 1010 and generating the new chemical structure 1020.

[0148] Fig.11 The flowchart is a method for generating a new chemical structure by changing a descriptor of the chemical structure in a neural network device according to an embodiment.

[0149] The method of generating a chemical structure in a neural network device relates to the embodiment described above with reference to the accompanying drawings, and therefore, although omitted in the following description, the description given above with reference to the accompanying drawings can also be applied to the embodiment described above with reference to the accompanying drawings. Fig.11The method shown in .

[0150] Reference Fig.11 , in operation 1110, the neural network device may obtain a descriptor for a reference chemical structure.

[0151] A descriptor is an indicator value used to represent a structural characteristic of a chemical structure. A descriptor can be obtained by performing relatively simple operations on a given chemical structure. In an embodiment, a descriptor can be represented by an ECFP and can include multiple bit values. However, the expression of a descriptor is not limited to this.

[0152] In the following, the descriptor for the reference chemical structure will be referred to as the reference descriptor.

[0153] In operation 1120, the neural network device may input the reference descriptor into the trained neural network and output a property value for a specific property of the reference chemical structure.

[0154] Properties refer to characteristics possessed by a chemical structure and may be real numerical values ​​indicating transmission and emission wavelengths with respect to light. Unlike the case of descriptors, calculation of properties may require complex simulations and be time consuming.

[0155] The memory of the neural network device may store descriptors for specific chemical structures and property values ​​numerically representing properties of the specific chemical structures that match each other as a group.

[0156] In an embodiment, the neural network device may allow a neural network (e.g., DNN) to learn by using descriptors and property values ​​stored in a memory. In the learning process using the descriptors and the property values, a factor defining a relationship between the descriptors and the property values ​​may be determined in the neural network.

[0157] The neural network device may output a property value corresponding to the reference descriptor as output data of the neural network by: inputting the reference descriptor as input data of a trained neural network, and driving the neural network.

[0158] In operation 1130 , the neural network device may determine an expression region expressing a specific property in the reference descriptor.

[0159] The neural network device may perform an interpretation process to determine whether a particular property value is expressed by any local structure in the reference chemical structure.

[0160] In an embodiment, the neural network device may interpret the trained neural network by using the LRP technique. The LRP technique is a method of propagating correlation in the reverse direction of the trained neural network (i.e., the direction from the output layer to the input layer). In the LRP technique, when propagating correlation between layers, a node having the maximum correlation with an upper layer among a plurality of nodes in a lower layer obtains the maximum correlation from a corresponding node in the upper layer.

[0161] For application of the LRP technique, the neural network device may optionally direct the activation function applied to the nodes of the trained neural network to be a linear function, and may select MSE for optimization.

[0162] The plurality of nodes of the input layer of the neural network may respectively correspond to the bit values ​​constituting the descriptor. The neural network device may obtain the node of the input layer having the greatest correlation in the expression of the specific property value of the reference chemical structure, that is, the bit position (or expression region) of the reference descriptor through an interpretation process. Since the expression region of the reference descriptor corresponds to a specific local structure in the reference chemical structure, the neural network device may determine the specific local structure having the greatest correlation in the expression of the specific property value of the reference chemical structure by obtaining the expression region of the reference descriptor through the interpretation process.

[0163] In operation 1140, the neural network device may generate a new chemical structure by changing a local structure corresponding to the expression region in the reference chemical structure.

[0164] The neural network device may receive a target property value as input. In an embodiment, the neural network device may include a user interface, which is a tool for inputting data for controlling the neural network device. For example, the user interface may be a keypad, a touchpad, etc., but is not limited thereto.

[0165] The target property value is a numerical value of a specific property of a chemical structure that will be ultimately generated in the neural network device. In an embodiment, the target property value may be a refractive index value, an elastic modulus, a melting point, a transmission wavelength, and / or an emission wavelength. For example, the neural network device may receive "emission wavelength: 350 nm" as a target property value. Alternatively, the target property value may be set in an increasing (+) direction or a decreasing (-) direction, rather than in a numerical value.

[0166] The neural network device can generate a new chemical structure having a property value close to the target property value by changing the local structure in the reference chemical structure.

[0167] In an embodiment, the neural network device may output a new descriptor by changing the bit value of the expression region of the reference descriptor. When the bit value of the expression region of the reference descriptor is changed, the local structure in the reference chemical structure may be changed. The method of changing the bit value of the expression region may use a genetic algorithm, but is not limited thereto.

[0168] The neural network device may output a structural feature value corresponding to a new descriptor as output data of the neural network by inputting a new descriptor in which the bit value of the expression region of the reference descriptor is changed as input data of a trained neural network (e.g., RNN), and driving the neural network. The neural network device may generate a new chemical structure based on the output structural feature value. Alternatively, the neural network device may use a factor for a new descriptor outputted during the learning process of the DNN as input data of a trained neural network (e.g., RNN).

[0169] The neural network device can repeatedly generate chemical structures through the above process until a chemical structure having a property value close to the target property value (eg, "emission wavelength: 350 nm") is generated.

[0170] Specifically, the neural network device may compare the property value for the new chemical structure with the target property value, and when the property value for the new chemical structure is less than the target property value, generate a new chemical structure again by changing the bit value of the expression area of ​​the reference descriptor.

[0171] When the property value of the new chemical structure generated through the above process is equal to or greater than the target property value, the neural network device may store the generated new chemical structure in a memory.

[0172] Fig.12 Flow chart of a method for generating a new chemical structure by changing an image of the chemical structure in a neural network device according to an embodiment.

[0173] In the following, the references to Fig.11 The same descriptions as those given.

[0174] Reference Fig.12 In operation 1210, the neural network device may obtain an image for a reference chemical structure.

[0175] In an embodiment, the image for the reference chemical structure may include n×m pixels (where n and m are natural numbers). For example, 8 bits, i.e., values ​​from 0 (black) to 255 (white), may be assigned to each pixel of the image.

[0176] Hereinafter, the image for the reference chemical structure will be referred to as a reference image.

[0177] In operation 1220, the neural network device may input the reference image into the trained neural network and output a property value for a specific property of the reference chemical structure.

[0178] The memory of the neural network device may store images for specific chemical structures that match each other and property values ​​that numerically represent properties of the specific chemical structure as a group.

[0179] In an embodiment, the neural network device may allow a neural network (e.g., DNN) to learn by using an image and a property value stored in a memory. In the learning process using the image and the property value, a factor defining a relationship between the image and the property value may be determined in the neural network.

[0180] The neural network device may output a property value corresponding to the reference image as output data of the neural network by: inputting the reference image as input data of a trained neural network, and driving the neural network.

[0181] In operation 1230, the neural network device may determine an expression region expressing a specific property in the image.

[0182] The plurality of nodes of the input layer of the neural network may respectively correspond to the pixel values ​​constituting the image. The neural network device may obtain the node of the input layer having the greatest correlation in the expression of the specific property value of the reference chemical structure through an interpretation process, that is, the pixel coordinates (or expression area) of the reference image. Since the expression area of ​​the reference image corresponds to a specific local structure in the reference chemical structure, the neural network device may determine the specific local structure having the greatest correlation in the expression of the specific property value of the reference chemical structure by obtaining the expression area of ​​the reference image through an interpretation process.

[0183] In operation 1240, the neural network device may generate a new chemical structure by changing a local structure corresponding to the expression region in the reference chemical structure.

[0184] In an embodiment, the neural network device may generate a new image by changing the pixel value of the expression region of the reference image and / or the pixel value around the expression region. When the pixel value of the expression region of the reference image and / or the pixel value around the expression region is changed, the local structure in the reference chemical structure may be changed. In an embodiment, the pixel value of the expression region of the reference image and / or the pixel value around the expression region may be changed by using Gaussian noise, but the method of changing the pixel value is not limited thereto.

[0185] The neural network device may output a structural feature value corresponding to a new image as output data of the neural network by: inputting a new image in which the pixel values ​​of the expression region of the reference image and / or the pixel values ​​around the expression region are changed as input data of a trained neural network (e.g., RNN), and driving the neural network. The neural network device may generate a new chemical structure based on the output structural feature value. Alternatively, the neural network device may use a factor for a new image output during the learning process of the DNN as input data of a trained neural network (e.g., RNN).

[0186] The neural network device can repeatedly generate chemical structures through the above process until a chemical structure having a property value close to the target property value (eg, "emission wavelength: 350 nm") is generated.

[0187] Specifically, the neural network device may compare the property value for the new chemical structure with the target property value, and when the property value for the new chemical structure is less than the target property value, generate the new chemical structure again by changing the pixel value of the expression area of ​​the reference image and / or the pixel value around the expression area.

[0188] When the property value of the new chemical structure generated through the above process is equal to or greater than the target property value, the neural network device may store the generated new chemical structure in a memory.

[0189] According to the above embodiments, the trained neural network can be interpreted to specify the local structure that expresses the properties of the chemical structure. In addition, by changing the specified local structure, a new chemical structure with improved properties can be generated.

[0190] In addition, the aforementioned embodiments can be embodied in the form of a recording medium storing instructions executable by a computer, such as a program module executed by a computer. A computer-readable medium can be any recording medium accessible by a computer, and can include volatile and non-volatile media and removable and non-removable media. In addition, the computer-readable medium can include computer storage media and communication media. The computer storage medium includes volatile and non-volatile and removable and non-removable media implemented using any method or technology for storing information such as computer-readable instructions, data structures, program modules, or other data. The communication medium includes computer-readable instructions, data structures, program modules, or other data of a modulated data signal or other transmission mechanism and can include any transmission medium.

[0191] In addition, throughout the specification, the term "unit" may be a hardware component such as a processor or a circuit and / or a software component executed by a hardware component such as a processor.

[0192] The above description of the present disclosure is provided for illustrative purposes, and those skilled in the art will appreciate that various changes and modifications may be made without changing the technical concept and essential features of the present disclosure. Therefore, it is clear that the above illustrative embodiments are illustrative and do not limit the present disclosure in all aspects. For example, each component described as a single type can be implemented in a distributed manner. Similarly, the components described as distributed can be implemented in a combined manner.

[0193] It should be understood that the embodiments described herein should be considered in a descriptive sense only and not for purposes of limitation. Descriptions of features or aspects in each embodiment should typically be considered as available for other similar features or aspects in other embodiments.

[0194] Although one or more embodiments have been described with reference to the drawings, it will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope as defined by the following claims.

Claims

1. A method for generating a chemical structure by using a neural network device, the method comprising: Inputting a descriptor of a chemical structure into a trained neural network, the trained neural network generating a property value of a property of the chemical structure, wherein the descriptor of the chemical structure represents a structural characteristic of the chemical structure and the property of the chemical structure is a characteristic possessed by the chemical structure; determining an expression region in the descriptor for expressing the property, the expression region comprising a first bit position in the descriptor, wherein the descriptor comprises the expression region and a second bit position outside the expression region; and generating a new chemical structure by modifying a local structure in the chemical structure, the local structure corresponding to the expression region, The determining includes determining an expression region in the descriptor for expressing the property by applying a layer-by-layer relevance propagation technique to a deep neural network in the trained neural network.

2. The method of claim 1, wherein the activation function applied to the nodes of the trained neural network is selected as a linear function to apply the layer-by-layer relevance propagation technique to the trained neural network, and mean squared error is selected for optimization.

3. The method of claim 1, wherein said generating comprises: obtaining a bit value of a bit position of the expression region in the descriptor; and The new chemical structure is generated by applying a genetic algorithm to the bit values ​​of the bit positions and modifying the local structure corresponding to the expression region.

4. The method of claim 1, wherein said generating comprises: generating a new first chemical structure by modifying the local structure in the chemical structure, the local structure corresponding to the expression region; inputting the descriptor for the new first chemical structure into the trained neural network to output a property value for a specific property of the new first chemical structure; as well as When a property value of a specific property of the new first chemical structure is less than a preset value, a new second chemical structure is generated by changing a local structure in the new first chemical structure, the local structure corresponding to the expression region, and when the property value of the specific property of the new first chemical structure is equal to or greater than the preset value, the new first chemical structure is stored.

5. A neural network device configured to generate a chemical structure, the neural network device comprising: a memory configured to store at least one program; and a processor configured to control the neural network device to implement a neural network by executing the at least one program, wherein when the at least one program is executed, the processor is configured to: Inputting a descriptor of a chemical structure into a trained neural network, the trained neural network generating a property value of a property of the chemical structure, wherein the descriptor of the chemical structure represents a structural characteristic of the chemical structure and the property of the chemical structure is a characteristic possessed by the chemical structure; determining an expression region in the descriptor for expressing the property, the expression region comprising a first bit position in the descriptor, wherein the descriptor comprises the expression region and a second bit position outside the expression region; and generating a new chemical structure by modifying a local structure in the chemical structure, the local structure corresponding to the expression region, The processor is further configured to determine an expression region in the descriptor for expressing the property by applying a layer-by-layer relevance propagation technique to a deep neural network in the trained neural network.

6. The neural network device of claim 5, wherein when the at least one program is executed, the processor is further configured to: An activation function applied to a node of the trained neural network is selected as a linear function to apply the layer-by-layer relevance propagation technique and a selected mean square error to the trained neural network for optimization.

7. The neural network device of claim 5, wherein when the at least one program is executed, the processor is further configured to obtain bit values ​​of bit positions of the expression region in the descriptor and generate the new chemical structure by applying a genetic algorithm to the bit values ​​of the bit positions and modifying the local structure corresponding to the expression region.

8. The neural network device of claim 5, wherein when the at least one program is executed, the processor is configured to: generating a new first chemical structure by modifying the local structure in the chemical structure, the local structure corresponding to the expression region; inputting the descriptor for the new first chemical structure into the trained neural network to output a property value for a specific property of the new first chemical structure; and When a property value of a specific property of the new first chemical structure is less than a preset value, a new second chemical structure is generated by changing a local structure in the new first chemical structure, the local structure corresponding to the expression region, and when the property value of the specific property of the new first chemical structure is equal to or greater than the preset value, the new first chemical structure is stored in the memory.

9. A method for generating a chemical structure by using a neural network device, the method comprising: Inputting an image of a chemical structure to a trained neural network, the trained neural network generating a property value of a property of the chemical structure, wherein the image of the chemical structure represents a structural characteristic of the chemical structure and the property of the chemical structure is a characteristic possessed by the chemical structure; determining an expression region in the image for expressing the property, the expression region comprising a first pixel in the image, wherein the image comprises the expression region and a second pixel outside the expression region; and generating a new chemical structure by modifying a local structure in the chemical structure, the local structure corresponding to the expression region, The determining comprises determining an expression region in the image for expressing the property by applying a layer-by-layer relevance propagation technique to a deep neural network in the trained neural network.

10. The method of claim 9, wherein the activation function applied to the nodes of the trained neural network is selected as a linear function to apply the layer-by-layer relevance propagation technique to the trained neural network, and mean squared error is selected for optimization.

11. The method of claim 9, wherein said generating comprises: obtaining pixel values ​​of pixels in the expression area in the image; and The new chemical structure is generated by applying Gaussian noise to the pixel values ​​of the pixels and modifying the local structure corresponding to the expression region.

12. The method of claim 9, wherein the expression region comprises a plurality of expression regions expressing the property, and the generating comprises: Obtaining coordinate information corresponding to the plurality of expression regions in the image; Calculate the center points of the multiple expression areas in the image based on the coordinate information and obtain the pixel value of the center point; and The new chemical structure is generated by applying Gaussian noise to the pixel values ​​and modifying the local structure corresponding to the center point.

13. The method of claim 9, wherein said generating comprises: generating a new first chemical structure by modifying the local structure in the chemical structure, the local structure corresponding to the expression region; inputting an image for the new first chemical structure into the trained neural network to output a property value for a specific property of the new first chemical structure; and When a property value of a specific property of the new first chemical structure is less than a preset value, a new second chemical structure is generated by changing a local structure in the new first chemical structure, the local structure corresponding to the expression region, and when the property value of the specific property of the new first chemical structure is equal to or greater than the preset value, the new first chemical structure is stored.

14. A neural network device configured to generate a chemical structure, the neural network device comprising: a memory configured to store at least one program; and a processor configured to control the neural network device to implement a neural network by executing the at least one program, wherein when the at least one program is executed, the processor is configured to: Inputting an image of a chemical structure to a trained neural network, the trained neural network generating a property value of a property of the chemical structure, wherein the image of the chemical structure represents a structural characteristic of the chemical structure and the property of the chemical structure is a characteristic possessed by the chemical structure; determining an expression region in the image for expressing the property, the expression region comprising a first pixel in the image, wherein the image comprises the expression region and a second pixel outside the expression region; and generating a new chemical structure by modifying a local structure in the chemical structure, the local structure corresponding to the expression region, The processor is further configured to determine an expression region in the image for expressing the property by applying a layer-by-layer relevance propagation technique to a deep neural network in the trained neural network.

15. The neural network device of claim 14, wherein when the at least one program is executed, the processor is further configured to: An activation function applied to a node of the trained neural network is selected as a linear function to apply the layer-by-layer relevance propagation technique and a selected mean square error to the trained neural network for optimization.

16. The neural network device of claim 14, wherein when the at least one program is executed, the processor is further configured to obtain pixel values ​​of pixels in the expression region in the image and generate the new chemical structure by applying Gaussian noise to the pixel values ​​of the pixels and modifying the local structure corresponding to the expression region.

17. The neural network device of claim 14, wherein the expression region comprises a plurality of expression regions expressing the property, and when the at least one program is executed, the processor is further configured to: Obtaining coordinate information corresponding to the plurality of expression regions in the image; Calculating the center points of the plurality of expression areas in the image based on the coordinate information and obtaining pixel values ​​of the center points; and The new chemical structure is generated by applying Gaussian noise to the pixel values ​​and modifying the local structure corresponding to the center point.

18. The neural network device of claim 14, wherein when the at least one program is executed, the processor is further configured to: generating a new first chemical structure by modifying the local structure in the chemical structure, the local structure corresponding to the expression region; inputting an image for the new first chemical structure to the trained neural network to output a property value for a specific property of the new first chemical structure; and When a property value of a specific property of the new first chemical structure is less than a preset value, a new second chemical structure is generated by changing a local structure in the new first chemical structure, the local structure corresponding to the expression region, and when the property value of the specific property of the new first chemical structure is equal to or greater than the preset value, the new first chemical structure is stored in the memory.

19. A non-transitory computer-readable recording medium comprising a program which, when executed by a computer, implements the method according to any one of claims 1 to 4 and 9 to 13.

Images