Don’t Overweight Weights: Evaluation of Weighting Strategies for Multi-Task Bioactivity Classification Models

Machine learning models predicting the bioactivity of chemical compounds belong nowadays to the standard tools of cheminformaticians and computational medicinal chemists. Multi-task and federated learning are promising machine learning approaches that allow privacy-preserving usage of large amounts...

Descripción completa

Guardado en:
Detalles Bibliográficos
Autores principales: Lina Humbeck, Tobias Morawietz, Noe Sturm, Adam Zalewski, Simon Harnqvist, Wouter Heyndrickx, Matthew Holmes, Bernd Beck
Formato: article
Lenguaje:EN
Publicado: MDPI AG 2021
Materias:
Acceso en línea:https://doaj.org/article/1a348a51ff7646c193a44ace0101e956
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
id oai:doaj.org-article:1a348a51ff7646c193a44ace0101e956
record_format dspace
spelling oai:doaj.org-article:1a348a51ff7646c193a44ace0101e9562021-11-25T18:28:44ZDon’t Overweight Weights: Evaluation of Weighting Strategies for Multi-Task Bioactivity Classification Models10.3390/molecules262269591420-3049https://doaj.org/article/1a348a51ff7646c193a44ace0101e9562021-11-01T00:00:00Zhttps://www.mdpi.com/1420-3049/26/22/6959https://doaj.org/toc/1420-3049Machine learning models predicting the bioactivity of chemical compounds belong nowadays to the standard tools of cheminformaticians and computational medicinal chemists. Multi-task and federated learning are promising machine learning approaches that allow privacy-preserving usage of large amounts of data from diverse sources, which is crucial for achieving good generalization and high-performance results. Using large, real world data sets from six pharmaceutical companies, here we investigate different strategies for averaging weighted task loss functions to train multi-task bioactivity classification models. The weighting strategies shall be suitable for federated learning and ensure that learning efforts are well distributed even if data are diverse. Comparing several approaches using weights that depend on the number of sub-tasks per assay, task size, and class balance, respectively, we find that a simple sub-task weighting approach leads to robust model performance for all investigated data sets and is especially suited for federated learning.Lina HumbeckTobias MorawietzNoe SturmAdam ZalewskiSimon HarnqvistWouter HeyndrickxMatthew HolmesBernd BeckMDPI AGarticlemachine learningclassificationmulti-task learningfederatedweightingdrug designOrganic chemistryQD241-441ENMolecules, Vol 26, Iss 6959, p 6959 (2021)
institution DOAJ
collection DOAJ
language EN
topic machine learning
classification
multi-task learning
federated
weighting
drug design
Organic chemistry
QD241-441
spellingShingle machine learning
classification
multi-task learning
federated
weighting
drug design
Organic chemistry
QD241-441
Lina Humbeck
Tobias Morawietz
Noe Sturm
Adam Zalewski
Simon Harnqvist
Wouter Heyndrickx
Matthew Holmes
Bernd Beck
Don’t Overweight Weights: Evaluation of Weighting Strategies for Multi-Task Bioactivity Classification Models
description Machine learning models predicting the bioactivity of chemical compounds belong nowadays to the standard tools of cheminformaticians and computational medicinal chemists. Multi-task and federated learning are promising machine learning approaches that allow privacy-preserving usage of large amounts of data from diverse sources, which is crucial for achieving good generalization and high-performance results. Using large, real world data sets from six pharmaceutical companies, here we investigate different strategies for averaging weighted task loss functions to train multi-task bioactivity classification models. The weighting strategies shall be suitable for federated learning and ensure that learning efforts are well distributed even if data are diverse. Comparing several approaches using weights that depend on the number of sub-tasks per assay, task size, and class balance, respectively, we find that a simple sub-task weighting approach leads to robust model performance for all investigated data sets and is especially suited for federated learning.
format article
author Lina Humbeck
Tobias Morawietz
Noe Sturm
Adam Zalewski
Simon Harnqvist
Wouter Heyndrickx
Matthew Holmes
Bernd Beck
author_facet Lina Humbeck
Tobias Morawietz
Noe Sturm
Adam Zalewski
Simon Harnqvist
Wouter Heyndrickx
Matthew Holmes
Bernd Beck
author_sort Lina Humbeck
title Don’t Overweight Weights: Evaluation of Weighting Strategies for Multi-Task Bioactivity Classification Models
title_short Don’t Overweight Weights: Evaluation of Weighting Strategies for Multi-Task Bioactivity Classification Models
title_full Don’t Overweight Weights: Evaluation of Weighting Strategies for Multi-Task Bioactivity Classification Models
title_fullStr Don’t Overweight Weights: Evaluation of Weighting Strategies for Multi-Task Bioactivity Classification Models
title_full_unstemmed Don’t Overweight Weights: Evaluation of Weighting Strategies for Multi-Task Bioactivity Classification Models
title_sort don’t overweight weights: evaluation of weighting strategies for multi-task bioactivity classification models
publisher MDPI AG
publishDate 2021
url https://doaj.org/article/1a348a51ff7646c193a44ace0101e956
work_keys_str_mv AT linahumbeck dontoverweightweightsevaluationofweightingstrategiesformultitaskbioactivityclassificationmodels
AT tobiasmorawietz dontoverweightweightsevaluationofweightingstrategiesformultitaskbioactivityclassificationmodels
AT noesturm dontoverweightweightsevaluationofweightingstrategiesformultitaskbioactivityclassificationmodels
AT adamzalewski dontoverweightweightsevaluationofweightingstrategiesformultitaskbioactivityclassificationmodels
AT simonharnqvist dontoverweightweightsevaluationofweightingstrategiesformultitaskbioactivityclassificationmodels
AT wouterheyndrickx dontoverweightweightsevaluationofweightingstrategiesformultitaskbioactivityclassificationmodels
AT matthewholmes dontoverweightweightsevaluationofweightingstrategiesformultitaskbioactivityclassificationmodels
AT berndbeck dontoverweightweightsevaluationofweightingstrategiesformultitaskbioactivityclassificationmodels
_version_ 1718411060733542400