1. Program synthesis algorithm based on context consistency heuristic.
- Author
-
Dou, Quansheng, Pan, Hao, Tang, Huanling, Jiang, Ping, and Wang, Huixian
- Abstract
Programming-by-example (PBE) is a program synthesis technique that automatically synthesizes computer programs from input–output (I/O) examples. There are many approaches to program synthesis, and one of the most commonly used solutions is to perform a search over the space of programs. In this paper, we present a novel program synthesis algorithm based on context consistency heuristic (PSACC) that differs from previous heuristic methods. We consider the input–output example as a state and develop a deep neural network-based predictor called DHFnet. DHFnet has the ability to map the current state to the first two instructions of the program which achieves the current state. By considering the current state and its subsequent state obtained by executing the corresponding instruction, we design a context consistency-based search algorithm. Our search algorithm enables a more reasonable and efficient selection of instructions in the program, by judging whether the instructions maintain the context consistency between the current state and its subsequent state. Furthermore, under the same experimental and testing environments, PSACC successfully enhances the success rate of program synthesis by 2–5%, and reduces the synthesis time compared to the baseline method. PSACC also exhibits better synthesis efficiency in terms of comprehensive performance compared to existing program synthesis methods. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF