Adaptive Neural Compilation

TL;DR

This paper introduces an adaptive neural compilation framework that optimizes programs for efficiency on specific input distributions using differentiable program representations, enabling targeted algorithm learning.

Contribution

It presents a novel method to compile low-level programs into differentiable forms and optimize them for efficiency on specific data distributions, diverging from traditional static compiler transformations.

Findings

Successfully compiles low-level code into differentiable representations

Optimizes programs for efficiency on target input distributions

Achieves high success rate in learning distribution-specific algorithms

Abstract

This paper proposes an adaptive neural-compilation framework to address the problem of efficient program learning. Traditional code optimisation strategies used in compilers are based on applying pre-specified set of transformations that make the code faster to execute without changing its semantics. In contrast, our work involves adapting programs to make them more efficient while considering correctness only on a target input distribution. Our approach is inspired by the recent works on differentiable representations of programs. We show that it is possible to compile programs written in a low-level language to a differentiable representation. We also show how programs in this representation can be optimised to make them efficient on a target distribution of inputs. Experimental results demonstrate that our approach enables learning specifically-tuned algorithms for given data distributions with a high success rate.