2021
DOI: 10.1145/3434335
|View full text |Cite
|
Sign up to set email alerts
|

Combining the top-down propagation and bottom-up enumeration for inductive program synthesis

Abstract: We present an effective method for scalable and general-purpose inductive program synthesis. There have been two main approaches for inductive synthesis: enumerative search, which repeatedly enumerates possible candidate programs, and the top-down propagation (TDP), which recursively decomposes a given large synthesis problem into smaller subproblems. Enumerative search is generally applicable but limited in scalability, and the TDP is efficient but only works for special grammars or applications. In this pape… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
9
0

Year Published

2021
2021
2024
2024

Publication Types

Select...
4
2
1

Relationship

0
7

Authors

Journals

citations
Cited by 23 publications
(9 citation statements)
references
References 44 publications
0
9
0
Order By: Relevance
“…DryadSynth [Huang et al 2020] and Duet [Lee 2021] have shown that integrating bottom-up and top-down approaches results in synthesizers greater than the sum of their parts, and we believe these findings would generalize to problems involving recursion. In particular, we think there is promising future work in integrating Burst with a top-down recursive synthesizer like SMyth.…”
Section: Future Workmentioning
confidence: 72%
“…DryadSynth [Huang et al 2020] and Duet [Lee 2021] have shown that integrating bottom-up and top-down approaches results in synthesizers greater than the sum of their parts, and we believe these findings would generalize to problems involving recursion. In particular, we think there is promising future work in integrating Burst with a top-down recursive synthesizer like SMyth.…”
Section: Future Workmentioning
confidence: 72%
“…The solution quality can be roughly measured by the size of a solution program [50]. According to average and median sizes, solutions found by IGI-SBS and IGI-LGP have overall better quality than those found by GP and SA.…”
Section: B Results For List Manipulationmentioning
confidence: 99%
“…Program synthesis is an active research topic including a large and diverse body of work, such as enumerative program synthesis [5,40,50,51], constraint-based program synthesis [41,71,73] and neural program synthesis [17,22,38,61]. In what follows, we review some prior work on stochastic program synthesis that is most closely related to our proposed IGI.…”
Section: Related Workmentioning
confidence: 99%
“…The advantage of such a logic-based formalism is that it achieves a separation from solver and specification, which allows SyGuS to be solver-agnostic. Several different SyGuS solvers have been developed (e.g., [4,7,21,26]), many of which use drastically different internal algorithms that have different strengths for solving different kinds of problems. Moreover, a user of SyGuS need not consider the differing input languages or characteristics of these solvers, and instead can encode their problem just once in the SyGuS format to have access to all the different solvers.…”
Section: Why Isn't Existing Work In Synthesis Programmable?mentioning
confidence: 99%
“…For example, consider the problem of building an efficient enumeration algorithm for SemGuS, an algorithmic technique that is now successfully employed in most SyGuS solvers [2,4,21]. The success of enumeration has been driven by a number of clever ideas for efficiently pruning the search space of relevant programs.…”
Section: What Are We Working On Next?mentioning
confidence: 99%