mlf-core: a framework for deterministic machine learning

Abstract Motivation 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. Solely fixing all random seeds is not sufficient for deterministic...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Bioinformatics (Oxford, England) Jg. 39; H. 4
Hauptverfasser: Heumos, Lukas, Ehmele, Philipp, Kuhn Cuellar, Luis, Menden, Kevin, Miller, Edmund, Lemke, Steffen, Gabernet, Gisela, Nahnsen, Sven
Format: Journal Article
Sprache:Englisch
Veröffentlicht: England Oxford University Press 03.04.2023
Schlagworte:
Algorithms
Ecosystem
Machine Learning
Original Paper
Software
Tomography, X-Ray Computed
ISSN:1367-4811, 1367-4803, 1367-4811
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Abstract Motivation 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. Solely fixing all random seeds is not sufficient for deterministic machine learning, as major machine learning libraries default to the usage of nondeterministic algorithms based on atomic operations. Results Various machine learning libraries released deterministic counterparts to the nondeterministic 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 computed tomography scans, and a liver cancer classifier based on gene expression profiles with XGBoost. Availability and implementation The complete data together with the implementations of the mlf-core ecosystem and use case models are available at https://github.com/mlf-core.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
Gisela Gabernet and Sven Nahnsen contributed equally and share the senior authorship.
ISSN:1367-4811
1367-4803
1367-4811
DOI:10.1093/bioinformatics/btad164