Helix: Holistic Optimization for Accelerating Iterative Machine Learning

TL;DR

Helix is a system that optimizes iterative machine learning workflows by intelligently caching and recomputing intermediates, significantly reducing runtime compared to existing systems across diverse applications.

Contribution

Helix introduces a holistic optimization approach for iterative ML workflows, combining a Scala DSL with algorithms for caching and reuse, addressing a gap in current ML systems.

Findings

Achieves up to 19x speedup over state-of-the-art systems

Handles diverse ML workflows within a unified framework

Effective heuristics for NP-hard caching problem

Abstract

Machine learning workflow development is a process of trial-and-error: developers iterate on workflows by testing out small modifications until the desired accuracy is achieved. Unfortunately, existing machine learning systems focus narrowly on model training---a small fraction of the overall development time---and neglect to address iterative development. We propose Helix, a machine learning system that optimizes the execution across iterations---intelligently caching and reusing, or recomputing intermediates as appropriate. Helix captures a wide variety of application needs within its Scala DSL, with succinct syntax defining unified processes for data preprocessing, model specification, and learning. We demonstrate that the reuse problem can be cast as a Max-Flow problem, while the caching problem is NP-Hard. We develop effective lightweight heuristics for the latter. Empirical evaluation shows that Helix is not only able to handle a wide variety of use cases in one unified workflow but also much faster, providing run time reductions of up to 19x over state-of-the-art systems, such as DeepDive or KeystoneML, on four real-world applications in natural language processing, computer vision, social and natural sciences.