mlf-core: a framework for deterministic machine learning

TL;DR

mlf-core is a new framework that ensures deterministic machine learning models, addressing the limitations of existing libraries and enabling reliable deployment in sensitive applications.

Contribution

The paper introduces mlf-core, a comprehensive ecosystem that guarantees deterministic behavior in machine learning workflows across multiple libraries and applications.

Findings

mlf-core effectively enforces determinism in various ML models

Deterministic algorithms impact runtime and reproducibility

Application of mlf-core in biomedical models demonstrates practical utility

Abstract

Machine learning has shown extensive growth in recent years and is now routinely applied to sensitive areas. To allow appropriate verification of predictive models before deployment, models must be deterministic. However, major machine learning libraries default to the usage of non-deterministic algorithms based on atomic operations. Solely fixing all random seeds is not sufficient for deterministic machine learning. To overcome this shortcoming, various machine learning libraries released deterministic counterparts to the non-deterministic algorithms. We evaluated the effect of these algorithms on determinism and runtime. Based on these results, we formulated a set of requirements for deterministic machine learning and developed a new software solution, the mlf-core ecosystem, which aids machine learning projects to meet and keep these requirements. We applied mlf-core to develop deterministic models in various biomedical fields including a single cell autoencoder with TensorFlow, a PyTorch-based U-Net model for liver-tumor segmentation in CT scans, and a liver cancer classifier based on gene expression profiles with XGBoost.