Expert-in-the-loop AI for materials generation
Candidate material for polymerization can be received. One or more desired features in the candidate material can be identified. A machine learning model can be trained to generate a new material having one or more of the desired features. Permissively, the candidate material can be determined from running a machine learning classification model that ranks a plurality of material as candidates. Permissively, the generated new material can be input to the machine learning classification model, for the machine learning classification model to include in ranking the plurality of material as candidates.
1. A computer-implemented method comprising:
obtaining a first set of candidate materials for polymerization;
inputting the first set of candidate materials into a machine learning classification model such that, in response, the machine learning classification model produces a ranking of the first set of candidate materials at least based on synthesizability;
receiving, from at least one subject matter expert, one or more selections of individual candidate materials from the ranking;
identifying at least one structural feature of the one or more selected individual candidate materials by at least feeding decomposed fragments of the one or more selected individuals candidate materials to a trained random forest model of decision trees trained to recognize patterns of the decomposed fragments;
training a generative machine learning model on the selected one or more individual candidate materials, wherein a generator of the generative machine learning model generates new examples that follow example patterns in the selected one or more individual candidate materials, and a discriminator of the generative machine learning model receives as input at least one of the selected one or more individual candidate materials and at least one of the new examples and identifies which of the input is generated by the generator;
using the trained generative machine learning model with a constraint that includes the identified at least one structural feature and with input of a first candidate material to generate data representative of a second candidate material that is different from the first candidate material; and
interacting with the at least one subject matter expert via a user interface that presents a visual representation of the identified at least one structural feature, the visual representation being an artificial intelligence explanation.
2. The method of claim 1 , further comprising generating the first set of candidate materials from a library of molecular fragments.
3. The method of claim 1 , further comprising retraining the machine learning classification model using the data representative of the second candidate material.
4. The method of claim 1 , further comprising training a random forest model of decision trees using structured fingerprint data representation of molecular graphs such that the random forest model becomes the trained random forest model of decision trees.
5. The method of claim 1 , further comprising presenting the second candidate material on the user interface of a computer.
6. The method of claim 1 , further comprising inputting the one or more selections of individual candidate materials into the machine learning classification model such that, in response, the machine learning classification model re-ranks the first set of candidate materials.
7. The method of claim 1 , further comprising further training the generative machine learning model using candidate monomers selected by the at least one subject matter expert.
8. A computer system comprising:
a hardware processor; and
a memory device coupled with the hardware processor, the hardware processor configured to at least:
obtain a first set of candidate materials for polymerization;
input the first set of candidate materials into a machine learning classification model such that, in response, the machine learning classification model produces a ranking of the first set of candidate materials at least based on synthesizability;
receive, from at least one subject matter expert, one or more selections of individual candidate materials from the ranking;
identify at least one structural feature of the one or more selected individual candidate materials by at least feeding decomposed fragments of the one or more selected individuals candidate materials to a trained random forest model of decision trees trained to recognize patterns of the decomposed fragments;
train a generative machine learning model on the selected one or more individual candidate materials, wherein a generator of the generative machine learning model generates new examples that follow example patterns in the selected one or more individual candidate materials, and a discriminator of the generative machine learning model receives as input at least one of the selected one or more individual candidate materials and at least one of the new examples and identifies which of the input is generated by the generator;
use the trained generative machine learning model with a constraint that includes the identified at least one structural feature and with input of a first candidate material to generate data representative of a second candidate material that is different from the first candidate material; and
interact with the at least one subject matter expert via a user interface that presents a visual representation of the identified at least one structural feature, the visual representation being an artificial intelligence explanation.
9. The computer system of claim 8 , wherein the hardware processor is further configured to generate the first set of candidate materials from a library of molecular fragments.
10. The computer system of claim 8 , wherein the hardware processor is further configured to retrain the machine learning classification model using the data representative of the second candidate material.
11. The computer system of claim 8 , wherein the hardware processor is further configured to train a random forest model of decision trees using structured fingerprint data representation of molecular graphs such that the random forest model becomes the trained random forest model of decision trees.
12. The computer system of claim 8 , wherein the hardware processor is further configured to present the second candidate material on the user interface of a computer.
13. The computer system of claim 8 , wherein the hardware processor is further configured to input the one or more selections of individual candidate materials into the machine learning classification model such that, in response, the machine learning classification model re-ranks the first set of candidate materials.
14. The computer system of claim 8 , wherein the hardware processor is further configured to train the generative machine learning model using candidate monomers selected by the at least one subject matter expert.
15. A computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to:
obtain a first set of candidate materials for polymerization;
input the first set of candidate materials into a machine learning classification model such that, in response, the machine learning classification model produces a ranking of the first set of candidate materials at least based on synthesizability;
receive, from at least one subject matter expert, one or more selections of individual candidate materials from the ranking;
identify at least one structural feature of the one or more selected individual candidate materials by at least feeding decomposed fragments of the one or more selected individuals candidate materials to a trained random forest model of decision trees trained to recognize patterns of the decomposed fragments;
train a generative machine learning model on the selected one or more individual candidate materials, wherein a generator of the generative machine learning model generates new examples that follow example patterns in the selected one or more individual candidate materials, and a discriminator of the generative machine learning model receives as input at least one of the selected one or more individual candidate materials and at least one of the new examples and identifies which of the input is generated by the generator;
use the trained generative machine learning model with a constraint that includes the identified at least one structural feature and with input of a first candidate material to generate data representative of a second candidate material that is different from the first candidate material; and
interact with the at least one subject matter expert via a user interface that presents a visual representation of the identified at least one structural feature, the visual representation being an artificial intelligence explanation.
16. The computer program product of claim 15 , wherein the program instructions are executable by the computer to further cause the computer to generate the first set of candidate materials from a library of molecular fragments.
17. The computer program product of claim 15 , wherein the program instructions are executable by the computer to further cause the computer to retrain the machine learning classification model using the data representative of the second candidate material.
18. The computer program product of claim 15 , wherein the program instructions are executable by the computer to further cause the computer to train a random forest model of decision trees using structured fingerprint data representation of molecular graphs such that the random forest model becomes the trained random forest model of decision trees.
19. The computer program product of claim 15 , wherein the program instructions are executable by the computer to further cause the computer to present the second candidate material on the user interface of a computer.
20. The computer program product of claim 15 , wherein the program instructions are executable by the computer to further cause the computer to input the one or more selections of individual candidate materials into the machine learning classification model such that, in response, the machine learning classification model re-ranks the first set of candidate materials.