Machine learning framework for discovering materials with desired properties

Abstract

A computer-implemented method for discovering new material candidates from a chemical database is proposed. The method comprises: extracting a feature vector from a chemical formula; learning a prediction model for predicting a property value from the feature vector using a chemical database using a sparse kernel model, selecting an existing material from a list of existing materials sorted in descending order based on the property value predicted by the prediction model learned in the learning step, selecting a base material from a list of base materials sorted in descending order of absolute reaction magnitude to the selected existing material, and generating a new material candidate as a variant of the selected existing material taking into account the selected base material.

Description

Background Art Technical Field

[0001] The present invention relates generally to discovering new materials, and more particularly to a machine learning framework for discovering new materials with desired properties.

[0002] Related technical description

[0003] Given the availability of large-scale data, the practice of data mining has caused a paradigm shift in scientific discovery, from empirical science to theoretical, computational science, and now to data science. In particular, there is tremendous interest in applying advanced machine learning techniques to aid in materials discovery, which envisions the computationally assisted discovery, development, manufacture, and deployment of advanced materials at greatly accelerated speeds and reduced costs. Real-world applications encourage the development of methods that help objectively create new materials with desired properties for specific applications by linking what one sees (e.g., the structure of a crystal, the composition of a compound) to what one wants (e.g., a certain strength requirement for an alloy). Summary of the invention

[0004] According to an embodiment, a method for discovering new material candidates from a chemical database is provided. The method includes: extracting a feature vector from a chemical formula; learning a prediction model for predicting a property value from the feature vector using a chemical database using a sparse kernel model, selecting an existing material from a list of existing materials sorted in descending order based on the property value predicted by the prediction model learned in the learning step, selecting a base material from a list of base materials sorted in descending order of absolute reaction magnitude to the selected existing material, and generating a new material candidate as a variant of the selected existing material considering the selected base material.

[0005] According to another embodiment, a system for discovering new material candidates from a chemical database is provided. The system includes a memory and one or more processors in communication with the memory, the processors being configured to use a feature vector extracted from a chemical formula, use the chemical database to learn a prediction model for predicting a property value from the feature vector using a sparse kernel model, select an existing material from a list of existing materials sorted in descending order based on the property value predicted by the prediction model learned in the learning step, select a base material from a list of base materials sorted in descending order of absolute reaction magnitude to the selected existing material, and generate a new material candidate as a variant of the selected existing material taking into account the selected base material.

[0006] According to yet another embodiment, a non-transitory computer-readable storage medium is provided, the non-transitory computer-readable storage medium including a computer-readable program for discovering new material candidates from a chemical database. The non-transitory computer-readable storage medium performs the following steps: extracting a feature vector from a chemical formula, using a sparse kernel model to employ the chemical database to learn a prediction model for predicting a property value from the feature vector, selecting an existing material from a list of existing materials sorted in descending order based on the property value predicted by the prediction model learned in the learning step, selecting a base material from a list of base materials sorted in descending order of absolute reaction magnitude to the selected existing material, and generating a new material candidate as a variant of the selected existing material taking into account the selected base material.

[0007] It should be noted that exemplary embodiments are described with reference to different subject matters. In particular, some embodiments are described with reference to method type claims, while other embodiments are described with reference to apparatus type claims. However, a person skilled in the art will gather from the above and following description that, unless otherwise stated, any combination of features belonging to one type of subject matter, any combination of features of different subject matters, in particular features of method type claims, and any combination of features of apparatus type claims, in addition to any combination of features belonging to one type of subject matter, is considered to be described within this document.

[0008] These and other features and advantages will become apparent from the following detailed description of illustrative embodiments thereof, which is to be read in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The present invention will be provided in detail in the following description of preferred embodiments with reference to the following drawings, in which:

[0010] Figure 1 is an exemplary deterministic generative method for discovering material candidates;

[0011] Figure 2 is an exemplary diagram showing a sampling of training data according to an embodiment of the present invention;

[0012] Figure 3 is an exemplary diagram comparing the present method with the existing method according to an embodiment of the present invention;

[0013] Figure 4 is a block diagram / flow chart of an exemplary deterministic generative method for discovering material candidates by employing a sparse kernel model according to an embodiment of the present invention;

[0014] Figure 5 is a block diagram / flow chart of exemplary equations for implementing a deterministic generation method for discovering material candidates according to an embodiment of the present invention;

[0015] Figure 6 is an exemplary processing system according to an embodiment of the present invention;

[0016] Figure 7 is a block diagram / flow chart of an exemplary cloud computing environment according to an embodiment of the present invention; and

[0017] Figure 8 is a schematic diagram of an exemplary abstract model layer according to an embodiment of the present invention.

[0018] Throughout the drawings, the same or similar reference numerals refer to the same or similar elements. DETAILED DESCRIPTION

[0019] According to an embodiment of the present invention, a method and apparatus are provided for generating new material candidates by adopting a deterministic method without randomness using a sparse kernel model, wherein the method generates candidates as variations between existing materials with higher target property values ​​in the sparse kernel model and base materials with large reaction magnitudes to these existing materials.

[0020] The problem of materials discovery involves determining certain compositions, formulations, and processing steps for materials with desired target properties. Neural networks as tools have been used in materials science applications such as spectroscopy classification and structure recognition, characterizing composition relationships of alloys, evaluation of descriptors, etc. However, neural networks and deep learning have not been widely used in materials discovery applications. Exemplary embodiments of the present invention employ deep learning techniques to discover new materials with desired properties.

[0021] Machine learning techniques are more effective in detecting and analyzing time series data than statistical techniques. This is because machine learning has two important features, namely feature engineering and prediction. The feature engineering aspect is used to solve the trend and seasonality problems of time series data. It can also be used to solve the problem of fitting models to time series data.

[0022] Deep learning is used to combine feature extraction of time series with nonlinear autoregressive models for more advanced predictions. Deep learning is used to automatically extract useful information from features without using any human effort or complex statistical techniques.

[0023] There are two most effective machine learning techniques, i.e. supervised and unsupervised learning. Supervised learning is performed on the training data points so that they can be classified into different categories of data points. However, for supervised learning, it is necessary to label the data points. However, another method is unsupervised learning of unlabeled data points. Unsupervised learning can be applied to train the system so that the next data point in the series can be predicted. In order to achieve this, a confidence interval or prediction error is made. Therefore, for the data points, a test can be implemented to check which data points exist within or outside the confidence interval. The most common supervised learning algorithms are supervised neural networks, support vector machine learning, k nearest neighbors, Bayesian networks, and decision trees. The most common unsupervised algorithms are self-organizing maps (SOM), K means, C means, expectation maximization meta-algorithm (EM), adaptive resonance theory (ART), and a class of support vector machines.

[0024] It should be understood that the present invention will be described according to a given illustrative architecture; however, other architectures, structures, substrate materials and process features and steps / blocks may be varied within the scope of the present invention. It should be noted that certain features may not be shown in all figures for the sake of clarity. This is not intended to be interpreted as limiting any particular embodiment, or description, or the scope of the claims.

[0025] Figure 1 is an exemplary deterministic generative method for discovering material candidates.

[0026] Discovering new materials such as organic polymers (e.g., epoxy resins) with desired properties (e.g., high glass transition temperature, low viscosity, etc.) is an important task for the chemical industry. There are many similar problems and needs (inverse analysis) in different fields, such as generating processes and recipes (for chemical reactions, cooking, fragrances, etc.), optimizing physical structures, advertising, etc.

[0027] The problem is to find or discover materials with high property values ​​from a chemical database, e.g. Figure 1 100. Input 102 includes multiple pairs of existing materials (e.g., chemical formulas that can be described using the Simplified Molecular Input Line Entry System (SMILES chemical file format)) and corresponding properties of the chemical formula. Output 118 includes chemical formulas of new materials with desired properties, where the desired properties can be, for example, high values. However, the desired properties can be generalized to other settings, such as exploring materials with specific property values.

[0028] In a conventional method, the following steps can be performed in a batch mode or a sequential mode. In the first step, a prediction model 106 for a chemical property is constructed based on some feature vectors 104 extracted from a chemical formula 102. Feature vector 104 is a vector describing the characteristics of a chemical formula. In the second step, feature vector 104 is optimized (inversely analyzed) based on prediction model 106. In the third step, the vector is converted into a chemical formula. In an example, attribute value 108 can be selected from a training database 110. The selected attribute value can be, for example, attribute value 112. Attribute value 112 can be fed back into prediction model 114, and new feature vector 116 needs to be converted into new material 118 from the prediction model. The conversion from new feature vector 116 to new material 118 may be very difficult.

[0029] As a result, these methods require less experimental computational cost or require sufficient training data to learn a good response or result. In addition, extrapolation may be meaningless and it may be difficult to convert feature vectors into chemical formulas. Therefore, certain assumptions can be made that conventional methods cannot directly solve the problem of predicting chemical formulas from attributes (or attribute values) and that due to insufficient training data stored in the training database 110, conventional methods cannot directly employ 3D structures as input to prediction models for attributes (or attribute values).

[0030] Figure 2 is an exemplary graph showing sampling of training data according to an embodiment of the present invention, and Figure 3 is an exemplary graph comparing the present method with the existing method according to an embodiment of the present invention.

[0031] Instead of adopting Figure 1 100, an exemplary embodiment employs a predictive model for property values ​​from a feature vector of a material and directly explores the chemical formula of new material candidates in the interpolation region of the predictive model (without converting the feature vector into a new material, such as Figure 1 Thus, exemplary embodiments of the present invention employ sparse kernel models for predictive models that predict properties or property values ​​from materials. By employing sparse kernel models, these methods explore and find materials between training data with good properties and the most reactive base materials in the predictive model.

[0032] The material in the training data can come from two sources: empirical experiments and computational simulations. Experimental data refers to trial-and-error iterations of experimental observations, examples being microscopic images acquired directly from material samples. Such data are usually very limited in size and can include noise that affects data quality with non-stationary variance. On the other hand, simulated data is more abundant, available, stable, and less prone to human error.

[0033] About the term "kernel model", the kernel model can be derived from the kernel function. The kernel function takes the vector in the original space as its input and returns the dot product of the vector in the feature space. The kernel method is a class of algorithms for pattern analysis. The general task of pattern analysis is to find and study the general types of relationships in the data set (e.g., clustering, ranking, principal component, correlation, classification). In its simplest form, the kernel technique means transforming the data into another dimension with a clear segmentation margin between the categories of the data. For many algorithms to solve these tasks, the data in the original representation must be explicitly transformed into a feature vector representation via a feature map specified by the user. On the contrary, the kernel method only requires a kernel specified by the user, that is, a similarity function on the data point pairs in the original representation. Thus, the kernel function provides a way to manipulate data (as if it were projected into a higher dimensional space) by operating data in its original space.

[0034] With respect to the term "base material," a composite or compound material is a material made from two or more component materials (or base materials) with significantly different physical or chemical properties that, when combined, produce a material with characteristics different from those of the individual components. The individual components remain separate and distinct within the finished structure. A new material may be preferred for many reasons, such as being stronger, lighter, or less expensive when compared to a traditional material. Thus, a base material may be considered a constituent material of a composite material.

[0035] In the first step, the method takes feature vectors extracted from chemical formulas and learns a model for predicting properties from the feature vectors with a sparse kernel model using a chemical database.

[0036] The feature vector is arbitrary (e.g., substructure count, structural feature, etc.) The feature vector does not need to take into account the ease of converting the feature vector to a material.

[0037] Furthermore, the dimensionality reduction method of the feature vector can be used without affecting the algorithm.

[0038] The exemplary method may employ support vector machines (SVM), relevance vector machines (RVM), and other models for sparse kernel models, but Bayesian models (such as RVM) are more suitable because such exemplary methods may use the posteriori and confidence of the predictions and may also use a Bayesian optimization process in candidate generation.

[0039] SVM is a supervised machine learning technique that can be used for classification tasks. In its basic form, SVM represents data into instances in space, and SVM attempts to separate different classes by the widest possible gap (hyperplane) separating these classes. On the other hand, RVM uses probability measures to define the separation space. In one example, RVM uses Bayesian inference to obtain a solution. Therefore, RVM uses significantly fewer basis functions.

[0040] Furthermore, the exemplary method employs a kernel based on distances between data points in the feature space without requiring knowledge of the coordinates of the data points in the feature space.

[0041] In the second step, existing materials are selected in descending order based on the property values ​​predicted by the prediction model learned in the first step. In other words, a list of materials is created and sorted based on the property values ​​predicted by the prediction model. The descending order may be materials showing the closest desired property values ​​listed at the top and materials showing related desired property values ​​listed at the bottom. The descending order may be combined into absolutely closest, slightly closer, least closest, etc.

[0042] In some models (such as RVM), since predictions for the training data are required in the learning step, there is no additional computation for this prediction, only ranking.

[0043] The exemplary method can employ existing materials without attribute labels for this step and with additional prediction calculations. The lack of attribute labels makes these methods suitable for unsupervised learning techniques.

[0044] In the third step, base materials (materials from the training database) are selected in descending order of their absolute magnitude of reaction to the materials selected in the second step (distance between the selected material and the base x coefficient of the base). In other words, a list of materials is created and sorted based on the absolute magnitude of reaction to the selected existing materials.

[0045] From sparse modeling, activate a limited number of base materials.

[0046] In the fourth step, the exemplary method generates new material candidates as variations of the existing materials selected in the second step and the base materials selected in the third step as follows:

[0047] If the reaction in the third step is positive, the exemplary method "exchanges" the substructures between them as one or a combination of the following: adding a substructure of one material to the other material that is not included in the other material, or subtracting a substructure that is not shared between the materials.

[0048] If the reaction in the third step is negative, the exemplary method subtracts the substructure that shares the material.

[0049] Steps 2 to 4 are repeated until a predetermined number of candidate materials have been obtained. The predetermined number may be determined by the user or may be predefined by the software.

[0050] Since the exemplary method generates candidates as variations between existing data, the exemplary method can primarily generate candidates as interpolations of the prediction model. The exploration becomes wider depending on the amount and variation of the training data.

[0051] Exemplary embodiments of the present invention capture the property that chemical formulas can change structure discretely rather than continuously.

[0052] In sparse kernel models, the model has no information except for bases with non-zero weights learned through sparse learning. Typically, bases with non-zero weights will not be similar to each other in most cases, and materials and bases that are highly reactive with materials will be similar to each other due to the definition of the kernel (distance) in most cases.

[0053] Therefore, exchanging only materials with good properties will not be more efficient than the exemplary method.In addition, the exemplary method can be applied sequentially and does not require assumptions about the distinguishability of the prediction model.

[0054] about Figure 2 and 3 , an exemplary method samples training data from a function as shown in FIG200, learns a predictive model, and explores high values ​​of the function based on the predictive model.

[0055] The exemplary method is compared with conventional methods such as particle swarm optimization (mode 2) and genetic algorithm (mode 3).

[0056] The exemplary method further shows the found target values ​​and the average error of ten-fold cross validation for each experimental setting with different amounts of training data, as Figure 3 As shown in Figures 310 and 320.

[0057] The performance of the exemplary method (mode 1) is better than that of the other methods (modes 2 and 3), especially in settings where a small or insufficient amount of training data is available.

[0058] Therefore, the exemplary method is effective compared to other methods because the exemplary method has no randomness in its algorithm. In experiments on real chemical data, it has been determined that sparse kernel models (e.g., RVM) have better predictive performance than other methods (e.g., linear models, random forests, etc.). The exemplary method discovered new candidates with better performance than those found in the training database. In addition, in one instance, it was determined that 1 / 3 of the discovered candidates overlapped with new materials.

[0059] Figure 4 is a block diagram / flow chart of an exemplary deterministic generative method for discovering material candidates by employing a sparse kernel model according to an embodiment of the present invention.

[0060] At block 410 , a predictive model for a chemical property is constructed from feature vectors extracted from the chemical formula.

[0061] At block 412, existing materials are selected in descending order based on predicted property values ​​derived from the predictive model.

[0062] At block 414, base materials (selected from materials in the training data) are selected in descending order of absolute magnitude of response to the selected existing material.

[0063] At block 416 , new material candidates are generated as variations of the selected existing material and the selected base material. If the number of candidates is sufficient, the process ends and proceeds to block 418 .

[0064] At block 418 , the candidate list is output, for example, to a user interface of a computing device for evaluation and analysis.

[0065] Figure 5 is a block diagram / flow chart of exemplary equations for implementing a deterministic generative method for discovering material candidates according to an embodiment of the present invention.

[0066] The exemplary method can be approximated as a one-step up from existing material in the following optimization problem 510:

[0067]

[0068] where z is the chemical formula, x(z) is an eigenvector (arbitrary), and Φ(x(z)) is a basis function (distance-based, such as a radial basis function).

[0069] The gradient 520 can be expressed as:

[0070]

[0071] In summary, the properties of materials can be described at the macroscopic or microscopic level, such as hardness, melting point, ionic conductivity, glass transition temperature, molecular atomization energy, and lattice constant. There are two common methods for studying material properties: computational simulation and experimental measurement. These methods involve complex operations and experimental settings. Therefore, it is quite difficult to establish a computational simulation that fully captures the complex logical relationships between the properties of the material and its related factors, and some of these relationships may even be unknown. In addition, experiments conducted to measure the properties of compounds usually occur in the later stages of material selection. Therefore, if the results are not satisfactory, the large amount of time and experimental resources invested so far may prove to be invalid. In addition, in many cases, it is difficult or almost impossible to study the properties of materials even through a lot of computational or experimental efforts. Therefore, it is necessary to develop intelligent and high-performance prediction models that can correctly predict material properties at low computational costs. Machine learning involves the construction and research of algorithms that can learn patterns from data. The basic idea of ​​using machine learning methods for material property prediction is to analyze and map the relationship between the properties of materials and their related factors (in most cases nonlinear) by extracting knowledge from existing empirical data. However, empirical data may be insufficient. As a result, an exemplary embodiment of the present invention employs a machine learning framework to implement a randomness-free, deterministic method for generating new material candidates by using a sparse kernel model, wherein the method generates candidates that are variations between existing materials having higher target property values ​​and base materials having large response amplitudes to the existing materials in the sparse kernel model.

[0072] Figure 6 is an exemplary processing system according to an embodiment of the present invention.

[0073] See now Figure 6 , which shows the hardware configuration of a computing system 600 according to an embodiment of the present invention. As can be seen, this hardware configuration has at least one processor or central processing unit (CPU) 611. The CPU 611 is interconnected via a system bus 612 to a random access memory (RAM) 614, a read-only memory (ROM) 616, an input / output (I / O) adapter 618 (for connecting peripheral devices such as a disk unit 621 and a tape drive 640 to the bus 612), a user interface adapter 622 (for connecting a keyboard 624, a mouse 626, a speaker 628, a microphone 632 and / or other user interface devices to the bus 612), a communication adapter 634 for connecting the system 600 to a data processing network, the Internet, an intranet, a local area network (LAN), etc., and a display adapter 636 for connecting the bus 612 to a display device 638 and / or a printer 639 (e.g., a digital printer, etc.).

[0074] Figure 7is a block diagram / flow chart of an exemplary cloud computing environment according to an embodiment of the present invention.

[0075] It should be understood that although the present invention includes detailed descriptions about cloud computing, the implementation of the teachings described herein is not limited to cloud computing environments. Instead, embodiments of the present invention can be implemented in conjunction with any other type of computing environment now known or later developed.

[0076] Cloud computing is a service delivery model for convenient, on-demand network access to a shared pool of configurable computing resources. Configurable computing resources are resources that can be quickly deployed and released with minimal management cost or interaction with the service provider, such as networks, network bandwidth, servers, processing, memory, storage, applications, virtual machines, and services. This cloud model can include at least five characteristics, at least three service models, and at least four deployment models.

[0077] Features are as follows:

[0078] On-demand self-service: Cloud consumers can unilaterally and automatically deploy computing capabilities such as server time and network storage on demand without human interaction with the service provider.

[0079] Broad network access: Computing power can be accessed over the network through standard mechanisms that facilitate the use of the cloud through different types of thin-client or thick-client platforms (e.g., mobile phones, laptops, personal digital assistants (PDAs)).

[0080] Resource pool: The computing resources of the provider are grouped into resource pools and serve multiple consumers through a multi-tenant model, where different physical and virtual resources are dynamically allocated and reallocated on demand. Generally, consumers cannot control or even know the exact location of the provided resources, but can specify the location (such as country, state or data center) at a higher level of abstraction, so it is location-independent.

[0081] Rapid elasticity: The ability to quickly and elastically (sometimes automatically) deploy computing power to scale up quickly, and quickly release it to scale down quickly. To consumers, the computing power available for deployment often appears to be unlimited, and any amount of computing power can be accessed at any time.

[0082] Measurable services: Cloud systems automatically control and optimize resource usage by leveraging metering capabilities at a level of abstraction appropriate to the type of service (e.g., storage, processing, bandwidth, and active user accounts). Resource usage can be monitored, controlled, and reported, providing transparency to both service providers and consumers.

[0083] The service model is as follows:

[0084] Software as a Service (SaaS): The capability provided to consumers is to use the provider's applications running on a cloud infrastructure. Applications can be accessed from a variety of client devices through a thin client interface such as a web browser (e.g., web-based email). Aside from limited user-specific application configuration settings, consumers neither manage nor control the underlying cloud infrastructure including networks, servers, operating systems, storage, or even individual application capabilities.

[0085] Platform as a Service (PaaS): The capability provided to consumers is to deploy applications created or acquired by consumers on the cloud infrastructure, which are created using programming languages ​​and tools supported by the provider. Consumers neither manage nor control the underlying cloud infrastructure including networks, servers, operating systems or storage, but have control over the applications they deploy and may also have control over the configuration of the application hosting environment.

[0086] Infrastructure as a Service (IaaS): The capabilities provided to consumers are processing, storage, network, and other basic computing resources on which consumers can deploy and run arbitrary software including operating systems and applications. Consumers neither manage nor control the underlying cloud infrastructure, but have control over the operating system, storage, and applications they deploy, and may have limited control over selected network components (such as host firewalls).

[0087] The deployment model is as follows:

[0088] Private Cloud: Cloud infrastructure is run solely for an organization. The cloud infrastructure can be managed by the organization or a third party and can exist inside or outside the organization.

[0089] Community Cloud: A cloud infrastructure is shared by several organizations and supports a specific community with common interests (e.g., mission, security requirements, policies, and compliance considerations). A community cloud can be managed by multiple organizations within the community or by a third party and can exist inside or outside the community.

[0090] Public cloud: Cloud infrastructure is provided to the public or to large industry groups and is owned by an organization that sells cloud services.

[0091] Hybrid cloud: A cloud infrastructure consisting of two or more clouds (private, community or public) in a deployment model that remain distinct entities but are bound together by standardized or proprietary technologies (such as cloud burst sharing for load balancing between clouds) that enable data and application portability.

[0092] The cloud computing environment is service-oriented, with characteristics centered on statelessness, low coupling, modularity, and semantic interoperability. The core of cloud computing is the infrastructure that includes a network of interconnected nodes.

[0093] See now Figure 7 , depicts an illustrative cloud computing environment 750 for implementing the use case of the present invention. As shown, the cloud computing environment 750 includes one or more cloud computing nodes 710, and local computing devices used by cloud consumers (such as personal digital assistants (PDAs) or cellular phones 754A, desktop computers 754B, laptop computers 754C, and / or automotive computer systems 754N) can communicate with the cloud computing nodes 710. The nodes 710 can communicate with each other. They can be physically or virtually grouped (not shown) in one or more networks (such as the private clouds, community clouds, public clouds, or hybrid clouds described above, or a combination thereof). This allows the cloud computing environment 750 to provide infrastructure, platforms, and / or software as services without the cloud consumer having to maintain resources on local computing devices. It should be understood that Figure 7 The types of computing devices 754A-N shown in are intended to be illustrative only, and computing node 710 and cloud computing environment 750 may communicate with any type of computerized device over any type of network and / or network addressable connection (eg, using a web browser).

[0094] Figure 8 is a schematic diagram of an exemplary abstract model layer according to an embodiment of the present invention. It should be understood in advance that Figure 8 The components, layers, and functions shown in are intended to be illustrative only, and embodiments of the present invention are not limited thereto. As depicted, the following layers and corresponding functions are provided:

[0095] The hardware and software layer 860 includes hardware and software components. Examples of hardware components include: host 861; server 862 based on RISC (Reduced Instruction Set Computer) architecture; server 863; blade server 864; storage device 865; and network and networking components 866. In some embodiments, software components include network application server software 867 and database software 868.

[0096] Virtualization layer 870 provides an abstraction layer from which the following examples of virtual entities can be provided: virtual servers 871 ; virtual storage 872 ; virtual networks 873 , including virtual private networks; virtual applications and operating systems 874 ; and virtual clients 875 .

[0097] In one example, the management layer 880 may provide the functionality described below. Resource provisioning 881 provides dynamic acquisition of computing resources and other resources for performing tasks within a cloud computing environment. Metering and pricing 882 provides cost tracking when resources are utilized within a cloud computing environment, and bills or invoices for consumption of these resources. In one example, these resources may include application software licenses. Security provides authentication for cloud consumers and tasks, as well as protection of data and other resources. User portal 883 provides access to the cloud computing environment for consumers and system administrators. Service level management 884 provides cloud computing resource allocation and management so that the required service levels are met. Service level agreement (SLA) planning and fulfillment 885 provides pre-scheduling and procurement for cloud computing resources in anticipation of future requirements for the cloud computing resources based on the SLA.

[0098] The workload layer 890 provides examples of functions that can utilize a cloud computing environment. Examples of workloads and functions that can be provided from this layer include: mapping and navigation 891; software development and lifecycle management 892; virtual classroom education delivery 893; data analysis processing 894; transaction processing 895; and new material candidate generator 896.

[0099] As used herein, the terms "data," "content," "information," and similar terms may be used interchangeably to refer to data capable of being captured, transmitted, received, displayed, and / or stored in accordance with various example embodiments. Thus, the use of any such terms should not be taken to limit the spirit and scope of the present disclosure. Further, where a computing device is described herein as receiving data from another computing device, the data may be received directly from the other computing device or may be received indirectly via one or more intermediate computing devices (e.g., one or more servers, repeaters, routers, network access points, base stations, and / or the like).

[0100] To provide interaction with a user, embodiments of the subject matter described in this specification may be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display)) monitor for displaying information to the user, as well as a keyboard and a pointing device (e.g., a mouse or trackball through which a user can provide input to the computer). Other types of devices may also be used to provide interaction with a user; for example, feedback provided to a user may be any form of sensory feedback, such as visual feedback, auditory feedback, or tactile feedback; and input from a user may be received in any form, including acoustic, voice, or tactile input.

[0101] The present invention may be a system, method and / or computer program product. The computer program product may include a computer-readable storage medium (or multiple media) having computer-readable program instructions thereon for causing a processor to perform various aspects of the present invention.

[0102] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the foregoing. As used herein, a computer-readable storage medium is not to be interpreted as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0103] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0104] The computer program instructions for performing the operation of the present invention may be assembler instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, integrated circuit configuration data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed entirely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be personalized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present invention.

[0105] Various aspects of the present invention are described herein with reference to the flow charts and / or block diagrams of the methods, devices (systems) and computer program products according to embodiments of the present invention. It should be understood that each box of the flow chart and / or block diagram and the combination of each box in the flow chart and / or block diagram can be implemented by computer-readable program instructions.

[0106] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0107] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0108] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple embodiments of the present invention. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction includes one or more executable instructions for realizing the specified logical function. In some alternative implementations, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of special hardware and computer instructions.

[0109] References in the specification to "one embodiment" or "an embodiment" of the present principles and other variations thereof mean that a particular feature, structure, characteristic, etc. described in connection with the embodiment is included in at least one embodiment of the present principles. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" and any other variations appearing in various places throughout this specification are not necessarily all referring to the same embodiment.

[0110] It should be understood that the use of any of the following " / ", "and / or", and "at least one", for example, in the case of "A / B", "A and / or B", and "at least one of A and B", is intended to cover the selection of only the first listed option (A) or only the second listed option (B), or the selection of both options (A and B). As another example, in the case of "A, B, and / or C" and "at least one of A, B, and C", such language is intended to cover the selection of only the first listed option (A), or only the second listed option (B), or only the third listed option (C), or only the first and second listed options (A and B), or only the first and third listed options (A and C), or only the second and third listed options (B and C), or all three options (A and B and C). As is readily apparent to one of ordinary skill in the art, this can be extended for many of the items listed.

[0111] Having described preferred embodiments of systems and methods for discovering new material candidates from a chemical database (which are intended to be illustrative and not limiting), it should be noted that modifications and variations may be made by those skilled in the art based on the above teachings. Therefore, it should be understood that changes may be made to the specific embodiments described within the scope of the invention as outlined by the appended claims. Having therefore described various aspects of the invention with the details and particularity required by the patent laws, what is claimed and required is set forth in the appended claims.

Claims

1. A computer-implemented method executed on a processor for discovering new material candidates from a chemical database, the method comprising: Extract eigenvectors from chemical formulas; employing the chemical database to learn a prediction model for predicting a property value from the feature vector using a sparse kernel model; selecting an existing material from a list of existing materials sorted in descending order based on the property value predicted by the prediction model learned in the learning step; selecting a base material from a list of base materials sorted in descending order of absolute magnitude of response to said selected existing material; as well as New material candidates are generated as variations of the selected base material and the selected existing material.

2. The method of claim 1, wherein: The generating step further comprises: In response to a positive response, a substructure is exchanged between the selected existing material and the selected base material.

3. The method of claim 1, wherein: The generating step further comprises: In response to a negative response, a substructure shared between the selected existing material and the selected base material is subtracted from the selected existing material.

4. The method of claim 1, wherein: The base material is stored in a training database used to learn the predictive model.

5. The method of claim 1, wherein: The magnitude of the reaction is equal to the distance between the selected existing material and the base material x the coefficient of the prediction model.

6. The method of claim 1, wherein: The new material candidate is found in an interpolation region of the prediction model.

7. The method of claim 1, wherein the feature vector is arbitrary and the sparse kernel model is a relevance vector machine (RVM).

8. A non-transitory computer-readable storage medium comprising a computer-readable program executed on a processor in a data processing system for discovering new material candidates from a chemical database, wherein the computer-readable program, when executed on the processor, causes a computer to perform the following steps: Extract eigenvectors from chemical formulas; employing the chemical database to learn a prediction model for predicting a property value from the feature vector using a sparse kernel model; selecting an existing material from a list of existing materials sorted in descending order based on the property value predicted by the prediction model learned in the learning step; selecting a base material from a list of base materials sorted in descending order of absolute magnitude of response to said selected existing material; as well as New material candidates are generated as variations of the selected base material and the selected existing material.

9. The non-transitory computer-readable storage medium of claim 8, wherein: The generating step further comprises: In response to a positive response, a substructure is exchanged between the selected existing material and the selected base material.

10. The non-transitory computer-readable storage medium of claim 8, wherein: The generating step further comprises: In response to a negative response, a substructure shared between the selected existing material and the selected base material is subtracted from the selected existing material.

11. The non-transitory computer-readable storage medium of claim 8, wherein: The base material is stored in a training database used to learn the predictive model.

12. The non-transitory computer-readable storage medium of claim 8, wherein: The magnitude of the reaction is equal to the distance between the selected existing material and the base material x the coefficient of the prediction model.

13. The non-transitory computer-readable storage medium of claim 8, wherein: The new material candidate is found in an interpolation region of the predictive model.

14. The non-transitory computer-readable storage medium of claim 8, wherein the feature vector is arbitrary and the sparse kernel model is a relevance vector machine (RVM).

15. A system for discovering new material candidates from a chemical database, the system comprising: Memory; as well as One or more processors in communication with the memory, configured to: Extract eigenvectors from chemical formulas; employing the chemical database to learn a prediction model for predicting a property value from the feature vector using a sparse kernel model; selecting an existing material from a list of existing materials sorted in descending order based on the property value predicted by the prediction model learned in the learning step; selecting a base material from a list of base materials sorted in descending order of absolute magnitude of response to said selected existing material; as well as New material candidates are generated as variations of the selected base material and the selected existing material.

16. The system of claim 15, wherein: The generating of the new material candidate comprises: In response to a positive response, a substructure is exchanged between the selected existing material and the selected base material.

17. The system of claim 15, wherein: The generating of the new material candidate comprises: In response to a negative response, a substructure shared between the selected existing material and the selected base material is subtracted from the selected existing material.

18. The system of claim 15, wherein: The base material is stored in a training database used to learn the predictive model.

19. The system of claim 15, wherein: The magnitude of the reaction is equal to the distance between the selected existing material and the base material x the coefficient of the prediction model.

20. The system of claim 15, wherein: The new material candidate is found in an interpolation region of the predictive model.

21. The system of claim 15, wherein the feature vector is arbitrary and the sparse kernel model is a relevance vector machine (RVM).

Images