Deep-learning model catalog creation

1. A method, comprising:

mining a plurality of deep-learning models from a plurality of input sources;

extracting information from each of the deep-learning models, by parsing at least one of (i) code corresponding to the deep-learning model and (ii) text corresponding to the deep-learning model;

identifying, for each of the deep-learning models, operators that perform operations within the deep-learning model;

producing, for each of the deep-learning models and from (i) the extracted information and (ii) the identified operators, an ontology comprising terms and features of the deep-learning model, wherein the producing comprises populating a pre-defined ontology format with features of each deep-learning model; and

generating a deep-learning model catalog comprising the plurality of deep-learning models, wherein the deep-learning model catalog comprises, for each of the deep-learning models, the ontology corresponding to the deep-learning model.

2. The method of claim 1 , wherein the extracting comprises defining and utilizing at least one annotator specific to a deep-learning model domain based upon (i) deep-learning model vocabulary and (ii) deep-learning model parsing rules.

3. The method of claim 1 , wherein the parsing the code comprises (i) identifying a code framework of the code and (ii) identifying concepts contained within the code using a semantic code parser specific to the identified code framework.

4. The method of claim 3 , wherein the parsing the code comprises tagging, using a knowledge base, the code using deep-learning model identifiers by mapping identified concepts to code lines.

5. The method of claim 1 , wherein the parsing the text comprises utilizing one or more text parsers unique to a deep-learning model domain.

6. The method of claim 1 , wherein the identifying operators comprises identifying sub-tasks of the deep-learning model by mapping a sequence of identified operators to a sub-task.

7. The method of claim 1 , wherein the identifying operators comprises generating an abstract syntax tree from the code using an abstract syntax tree builder; and

performing on the abstract syntax tree both (i) a data flow analysis and (ii) a control flow analysis.

8. The method of claim 7 , wherein the identifying operators comprises performing (i) a semantic code analysis and (ii) an application programming interface call analysis on information generated from both (i) the data flow analysis and (ii) the control flow analysis.

9. The method of claim 1 , wherein the deep-learning model is mined from an input source having no code corresponding to the deep-learning model; and

wherein the extracting information comprises (i) mapping images corresponding to the deep-learning model mined from the input source to code and (ii) utilizing a text annotator specific to a deep-learning model domain to annotate text corresponding to the deep-learning model mined from the input source.

10. The method of claim 1 , wherein the producing an ontology comprises evaluating the deep-learning model using at least one dataset to fill-in information missing from the ontology after populating the extracted information and the identified operators in the pre-defined ontology format.

11. An apparatus, comprising:

at least one processor; and

a computer readable storage medium having computer readable program code embodied therewith and executable by the at least one processor, the computer readable program code comprising:

computer readable program code configured to mine a plurality of deep-learning models from a plurality of input sources;

computer readable program code configured to extract information from each of the deep-learning models, by parsing at least one of (i) code corresponding to the deep-learning model and (ii) text corresponding to the deep-learning model;

computer readable program code configured to identify, for each of the deep-learning models, operators that perform operations within the deep-learning model;

computer readable program code configured to produce, for each of the deep-learning models and from (i) the extracted information and (ii) the identified operators, an ontology comprising terms and features of the deep-learning model, wherein the producing comprises populating a pre-defined ontology format with features of each deep-learning model; and

computer readable program code configured to generate a deep-learning model catalog comprising the plurality of deep-learning models, wherein the deep-learning model catalog comprises, for each of the deep-learning models, the ontology corresponding to the deep-learning model.

12. A computer program product, comprising:

a computer readable storage medium having computer readable program code embodied therewith, the computer readable program code executable by a processor and comprising:

computer readable program code configured to mine a plurality of deep-learning models from a plurality of input sources;

computer readable program code configured to extract information from each of the deep-learning models, by parsing at least one of (i) code corresponding to the deep-learning model and (ii) text corresponding to the deep-learning model;

computer readable program code configured to identify, for each of the deep-learning models, operators that perform operations within the deep-learning model;

computer readable program code configured to produce, for each of the deep-learning models and from (i) the extracted information and (ii) the identified operators, an ontology comprising terms and features of the deep-learning model, wherein the producing comprises populating a pre-defined ontology format with features of each deep-learning model; and

computer readable program code configured to generate a deep-learning model catalog comprising the plurality of deep-learning models, wherein the deep-learning model catalog comprises, for each of the deep-learning models, the ontology corresponding to the deep-learning model.

13. The computer program product of claim 12 , wherein the extracting comprises defining and utilizing at least one annotator specific to a deep-learning model domain based upon (i) deep-learning model vocabulary and (ii) deep-learning model parsing rules.

14. The computer program product of claim 12 , wherein the parsing the code comprises (i) identifying a code framework of the code and (ii) identifying concepts contained within the code using a semantic code parser specific to the identified code framework.

15. The computer program product of claim 12 , wherein the parsing the text comprises utilizing one or more text parsers unique to a deep-learning model domain.

16. The computer program product of claim 12 , wherein the identifying operators comprises identifying sub-tasks of the deep-learning model by mapping a sequence of identified operators to a sub-task.

17. The computer program product of claim 12 , wherein the identifying operators comprises generating an abstract syntax tree from the code using an abstract syntax tree builder; and

performing on the abstract syntax tree both (i) a data flow analysis and (ii) a control flow analysis.

18. The computer program product of claim 12 , wherein the deep-learning model is mined from an input source having no code corresponding to the deep-learning model; and

wherein the extracting information comprises (i) mapping images corresponding to the deep-learning model mined from the input source to code and (ii) utilizing a text annotator specific to a deep-learning model domain to annotate text corresponding to the deep-learning model mined from the input source.

19. The computer program product of claim 12 , wherein the producing an ontology comprises evaluating the deep-learning model using at least one dataset to fill-in information missing from the ontology after populating the extracted information and the identified operators in the pre-defined format.

20. A method, comprising:

generating a deep-learning model catalog, wherein the deep-learning model catalog comprises (i) a plurality of deep-learning models and (ii) an ontology, in a pre-defined format, for each of the plurality of deep-learning models, wherein the generating comprises:

obtaining, from at least one input source, a plurality of deep-learning models; and

creating an ontology for each of the plurality of deep-learning models, wherein the ontology identifies attributes and terms for the deep-learning model, wherein the creating comprises:

parsing, using a deep-learning model annotator, each of the deep-learning models to extract semantic information for each of the deep-learning models;

identifying, for each of the deep-learning models, layers of the deep learning model, wherein the identifying comprises (i) identifying granular operations and (ii) grouping the granular operators into a layer that performs a task of the deep-learning model; and

populating, for each of the deep-learning models, a pre-defined ontology format with (i) the extracted semantic information and (ii) the layers.