Systems and methods for source code understanding using spatial representations
Systems and methods for source code understanding using spatial representations are disclosed. In one embodiment, a method may include: (1) receiving, by a source code understanding computer program, a source code snippet; (2) converting, by the source code understanding computer program, the source code snippet to a two-dimensional image representation using an encoding technique; (3) mapping, by the source code understanding computer program, the two-dimensional image representation into a three-dimensional image representation; (4) determining, by the source code understanding computer program, a classification for the source code snippet using a deep learning network; and (5) providing, by the source code understanding computer program, the classification for the source code snippet to a downstream system.
1. A method for source code understanding using spatial representations, comprising:
receiving, by a source code understanding computer program, a source code snippet;
converting, by the source code understanding computer program, the source code snippet to a matrix representation using an encoding technique, wherein the matrix representation encodes structural and contextual information of the source code snippet;
encoding, by the source code understanding computer program, the matrix representation into a three-dimensional image representation using one hot encoding or trainable embeddings;
determining, by the source code understanding computer program, a discriminative task for the source code snippet using a deep learning network; and
providing, by the source code understanding computer program, the discriminative task for the source code snippet to a downstream system.
2. The method of claim 1 , wherein the source code snippet is in a text format.
3. The method of claim 1 , wherein the encoding technique comprises American Standard Code for Information Interchange (ASCII) encoding.
4. The method of claim 1 , wherein the encoding technique comprises red-green-blue (RGB) formatting.
5. The method of claim 1 , wherein the source code understanding computer program uses one-hot encoding to encode the matrix representation into the three-dimensional image representation.
6. The method of claim 1 , wherein the source code understanding computer program uses trainable embedding of arbitrary dimension to encode the matrix representation into the three-dimensional image representation.
7. The method of claim 1 , wherein the discriminative task is learning a problem that the source code snippet is trying to solve.
8. The method of claim 1 , wherein the discriminative task is learning whether the source code snippet compiles.
9. The method of claim 1 , wherein the discriminative task is learning a memory footprint for the source code snippet.
10. The method of claim 1 , wherein the deep learning network comprises a deep convolutional neural network or a vision transformer.
11. A non-transitory computer readable storage medium, including instructions stored thereon, which when read and executed by one or more computer processors, cause the one or more computer processors to perform steps comprising:
receiving a source code snippet;
converting the source code snippet to a matrix representation using an encoding technique;
encoding the matrix representation into a three-dimensional image representation;
determining a discriminative task for the source code snippet using a deep learning network; and
providing the discriminative task for the source code snippet to a downstream system.
12. The non-transitory computer readable storage medium of claim 11 , wherein the source code snippet is in a text format.
13. The non-transitory computer readable storage medium of claim 11 , wherein the encoding technique comprises American Standard Code for Information Interchange (ASCII) encoding.
14. The non-transitory computer readable storage medium of claim 11 , wherein the encoding technique comprises red-green-blue (RGB) formatting.
15. The non-transitory computer readable storage medium of claim 11 , wherein the matrix representation is encoded into the three-dimensional image representation using one-hot encoding.
16. The non-transitory computer readable storage medium of claim 11 , wherein the matrix representation is encoded into the three-dimensional image representation using trainable embedding of arbitrary dimension.
17. The non-transitory computer readable storage medium of claim 11 , wherein the discriminative task is learning a problem that the source code snippet is trying to solve.
18. The non-transitory computer readable storage medium of claim 11 , wherein the discriminative task is learning whether the source code snippet compiles.
19. The non-transitory computer readable storage medium of claim 11 , wherein the discriminative task is learning a memory footprint for the source code snippet.
20. The non-transitory computer readable storage medium of claim 11 , wherein the deep learning network comprises a deep convolutional neural network or a vision transformer.