Learning to Compile Programs to Neural Networks

TL;DR

This paper introduces neural surrogate compilation, a method to generate neural network models directly from program text, improving efficiency and accuracy over traditional training methods.

Contribution

It presents a novel neural surrogate compiler that produces more data-efficient and accurate neural surrogates directly from program text, bypassing traditional training.

Findings

Neural surrogate compilers are 1.9-9.5x more data-efficient.

Produced surrogates are 1.0-1.3x more similar to ground truth.

Training requires 4.3-7.3x fewer epochs.

Abstract

A $\textit{neural surrogate of a program}$ is a neural network that mimics the behavior of a program. Researchers have used these neural surrogates to automatically tune program inputs, adapt programs to new settings, and accelerate computations. Researchers traditionally develop neural surrogates by training on input-output examples from a single program. Alternatively, language models trained on a large dataset including many programs can consume program text, to act as a neural surrogate. Using a language model to both generate a surrogate and act as a surrogate, however, leading to a trade-off between resource consumption and accuracy. We present $\textit{neural surrogate compilation}$, a technique for producing neural surrogates directly from program text without coupling neural surrogate generation and execution. We implement neural surrogate compilers using hypernetworks trained on a dataset of C programs and find that they produce neural surrogates that are $1.9$-$9.5\times$ as data-efficient, produce visual results that are $1.0$-$1.3\times$ more similar to ground truth, and train in $4.3$-$7.3\times$ fewer epochs than neural surrogates trained from scratch.