Gauss: program synthesis by reasoning over graphs

While input-output examples are a natural form of specification for program synthesis engines, they can be imprecise domains such as table transformations. In this paper, we investigate how extracting readily-available information about the user intent behind these helps speed up and reduce overfitting. We present Gauss, algorithm transformations that accepts partial examples, along with graphs. Gauss includes novel conflict-resolution reasoning over graphs enables it to learn from mistakes made during search use knowledge explore space programs even faster. It also ensures final is consistent specification, reducing implement domain (supporting Pandas R), compare three state-of-the-art synthesizers accepting only examples. find able by 56×, 73× 664× on average, resulting in 7×, 26× 7× speedups times respectively.