Improving molecular machine learning through adaptive subsampling with active learning
Yujing Wen, Zhixiong Li, Yan Xiang, Daniel Reker
Abstract
Active machine learning can be used to sample training data in an autonomous manner to improve machine learning performance. This approach is competitive with state-of-the-art data sampling approaches, especially on erroneous data.
Topics & Concepts
Active learning (machine learning)Computer scienceMachine learningArtificial intelligenceOnline machine learningSemi-supervised learningCompetitive learningSample (material)Sampling (signal processing)Training setUnsupervised learningFilter (signal processing)Computer visionChemistryChromatographyMachine Learning and AlgorithmsMachine Learning in Materials ScienceImbalanced Data Classification Techniques