Automatically improving accuracy for floating point expressions

By Pavel Panchekha, Alex Sanchez-Stern, James R. Wilcox, Zachary Tatlock. 2015

269 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computational Theory and Mathematics, Statistical and Nonlinear Physics, Optimization & Training

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computational Theory and Mathematics

In Statistical and Nonlinear Physics

In Optimization & Training

Citations in the corpus

CodeGen: An Open Large Language Model for Code with Multi-Turn Program Synthesis (2022, background): “Previous work (Joshi et al., 2002; Panchekha et al., 2015; Cheung et al., 2013) leverages domain-specific language to restrict the search space; however, this limits the applicability of synthesized programs.”
Learning programs from noisy data (2016, background): “Note that our approach is very different from recent work [5] which considers a more restricted case where the program is already fully available before the process of anomaly detection starts.”
Program Synthesis (2017, methodology): “Modern successful applications of similar techniques employ clever domain-specific heuristics for cutting down the derivation tree (see, for example, [63, 104]).”
FlashMeta: a framework for inductive program synthesis (2015, methodology): “Deductive synthesis has been successfully used for well-defined application domains such as numeric computations [29, 30].”
egg: Fast and extensible equality saturation (2020, methodology): “The “purely syntactic” restriction has forced existing equality saturation applications [Panchekha et al. 2015; Stepp et al. 2011; Tate et al. 2009] to resort to ad hoc passes over the e-graph to implement analyses like constant folding.”
Relay: a new IR for machine learning frameworks (2018, methodology): “[15]. Given this tolerance for low-accuracy arithmetic, we are eager to adapt recent techniques for automatically rewriting numerical code to improve accuracy at the cost of performance (e.g., Herbie [27]), to instead trade off accuracy for improved compute. By adapting tools like Herbie and STOKE [”