Unsupervised Learning by Program Synthesis

@inproceedings{Ellis2015UnsupervisedLB,
  title={Unsupervised Learning by Program Synthesis},
  author={Kevin Ellis and Armando Solar-Lezama and Joshua B. Tenenbaum},
  booktitle={NIPS},
  year={2015}
}
We extend two ideas from the program synthesis community to make search over programs tractable: Sketching: Manually provide a sketch, or rough outline, of the program to be induced [3]. Our sketches are probabilistic context-free grammars. Symbolic search: We automatically translate our sketches into Satisfiability Modulo Theories (SMT) problems. SMT problems are intractable in general, but often solved efficiently in practice, much like SAT problems. 
Highly Cited
This paper has 26 citations. REVIEW CITATIONS

8 Figures & Tables

Topics

Statistics

01020201620172018
Citations per Year

Citation Velocity: 10

Averaging 10 citations per year over the last 3 years.

Learn more about how we calculate this metric in our FAQ.