Machine learning framework for finding materials with desired properties

1. A computer-implemented method executed on a processor for discovering new material candidates from a chemical database, the method comprising:

performing deep learning using training data derived from material samples and material simulations that includes:

extracting a feature vector from a chemical formula;

training a prediction model using the feature vector with a sparse kernel model employing the training data derived from material samples and material simulations;

predicting property values from the feature vector using the trained prediction model;

selecting an existing material from a list of existing materials sorted in descending order based on the predicted property values by the trained prediction model;

selecting a basis material from a list of basis materials sorted in descending order of absolute reaction magnitudes to the selected existing material;

generating the new material candidates as variants of the selected existing material with consideration of the selected basis material; and

outputting the new material candidates on a user interface of a computing device to allow a person to analyze and evaluate the new material candidates to match the new material candidates quickly and efficiently with chemical tasks assigned to the person with less experimental efforts and lower computational costs.

2. The method of claim 1 , wherein the generating step further includes:

in response to a positive reaction, swapping a substructure between the selected existing material and the selected basis material.

3. The method of claim 1 , wherein the generating step further includes:

in response to a negative reaction, subtracting substructures shared between the selected existing material and the selected basis material from the selected existing material.

4. The method of claim 1 , wherein the basis material is stored in a training database used to train the prediction model.

5. The method of claim 1 , wherein the reaction magnitude is a distance between the selected existing material and the basis material×coefficient of the trained prediction model.

6. The method of claim 1 , wherein the new material candidates are discovered in an interpolation area of the trained prediction model.

7. The method of claim 1 , wherein the chemical tasks include at least high glass transition temperature tasks, low viscosity tasks, and chemical reaction tasks.

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 steps of:

performing deep learning, using training, data derived from material samples and material simulations that includes:

extracting a feature vector from a chemical formula;

training a prediction model using the feature vector with a sparse kernel model employing the training data derived from material samples and material simulations;

predicting property values from the feature vector using the trained prediction model;

selecting an existing material from a list of existing materials sorted in descending order based on the predicted property values by the trained prediction model;

selecting a basis material from a list of basis materials sorted in descending order of absolute reaction magnitudes to the selected existing material;

generating the new material candidates as variants of the selected existing material with consideration of the selected basis material; and

outputting the new material candidates on a user interface of a computing device to allow a person to analyze and evaluate the new material candidates to match the new material candidates quickly and efficiently with chemical tasks assigned to the person with less experimental efforts and lower computational costs.

9. The non-transitory computer-readable storage medium of claim 8 , wherein the generating step further includes:

in response to a positive reaction, swapping a substructure between the selected existing material and the selected basis material.

10. The non-transitory computer-readable storage medium of claim 8 , wherein the generating step further includes:

in response to a negative reaction, subtracting substructures shared between the selected existing material and the selected basis material from the selected existing material.

11. The non-transitory computer-readable storage medium of claim 8 , wherein the basis material is stored in a training database used to train the prediction model.

12. The non-transitory computer-readable storage medium of claim 8 , wherein the reaction magnitude is a distance between the selected existing material and the basis material×coefficient of the trained prediction model.

13. The non-transitory computer-readable storage medium of claim 8 , wherein the new material candidates are discovered in an interpolation area of the trained prediction model.

14. The non-transitory computer-readable storage medium of claim 8 , wherein the chemical tasks include at least high glass transition temperature tasks, low viscosity tasks, and chemical reaction tasks.

15. A system for discovering new material candidates from a chemical database, the system comprising:

a memory; and

one or more processors in communication with the memory configured to:

perform deep learning using training data derived from material samples and material simulations that includes:

extract a feature vector from a chemical formula;

train a prediction model the feature vector with a sparse kernel model employing the training data derived from material samples and material simulations;

predict property values from the feature vector using the trained prediction model:

select an existing material from a list of existing materials sorted in descending order based on the predicted property values by the trained prediction model;

select a basis material from a list of basis materials sorted in descending order of absolute reaction magnitudes to the selected existing material;

generate the new material candidates as variants of the selected existing material with consideration of the selected basis material; and

output the new material candidates on a user interface of a computing device to allow a person to analyze and evaluate the new material candidates to match the new material candidates quickly and efficiently with chemical tasks assigned to the person with less experimental efforts and lower computational costs.

16. The system of claim 15 , wherein the generation of the new material candidates includes;

in response to a positive reaction, swapping a substructure between the selected existing material and the selected basis material.

17. The system of claim 15 , wherein the generation of the new material candidates includes:

in response to a negative reaction, subtracting substructures shared between the selected existing material and the selected basis material from the selected existing material.

18. The system of claim 15 , wherein the basis material is stored in a training database used to train the prediction model.

19. The system of claim 15 , wherein the reaction magnitude is a distance between the selected existing, material and the basis material×coefficient of the trained prediction model.

20. The system of claim 15 , wherein the chemical tasks include at least high glass transition temperature tasks, low viscosity tasks, and chemical reaction tasks.