Paper Info
Title
On the difficulty of benchmarking inductive program synthesis methods.
Abstract
A variety of inductive program synthesis (IPS) techniques have recently been developed, emerging from different areas of computer science. However, these techniques have not been adequately compared on general program synthesis problems. In this paper we compare several methods on problems requiring solution programs to handle various data types, control structures, and numbers of outputs. The problem set also spans levels of abstraction; some would ordinarily be approached using machine code or assembly language, while others would ordinarily be approached using high-level languages. The presented comparisons are focused on the possibility of success; that is, on whether the system can produce a program that passes all tests, for all training and unseen testing inputs. The compared systems are Flash Fill, MagicHaskeller, TerpreT, and two forms of genetic programming. The two genetic programming methods chosen were PushGP and Grammar Guided Genetic Programming. The results suggest that PushGP and, to an extent, TerpreT and Grammar Guided Genetic Programming are more capable of finding solutions than the others, albeit at a higher computational cost. A more salient observation is the difficulty of comparing these methods due to drastically different intended applications, despite the common goal of program synthesis.
Year
DOI
Venue
2017
10.1145/3067695.3082533
GECCO (Companion)
Keywords
Field
DocType
Genetic programming, Machine Learning, Inductive Program Synthesis, Benchmarking
Program synthesis,Computer science,Inductive programming,Problem set,Grammar,Assembly language,Genetic programming,Data type,Machine code,Artificial intelligence,Machine learning
Conference
Citations 
PageRank 
References 
2
0.41
13
Authors
4
Name
Order
Citations
PageRank
Edward R. Pantridge182.24
Thomas Helmuth215018.71
Nicholas F. McPhee3237.35
Lee Spector419517.32