Data Augmentation for Regression Machine Learning Problems in High Dimensions

• DOI: 10.20944/preprints202312.0092.v1
• Publication Date: 2023-12-04T00:49:36Z
• Authors: Clara Guilhaumon; Nicolas Hascoët; Francisco Chinesta; Marc Lavarde; Fatima Daim
• Keywords: High dimensional

Machine learning approaches are currently used to understand or model complex physical systems. In general, a substantial number of samples must be collected create with reliable results. However, collecting numerous data is often relatively time-consuming expensive. Moreover, the problems industrial interest tend more and depending on high parameters. High dimensional intrinsically involve need large amount through curse dimensionality. That why, new based smart sampling techniques investigated minimize given train model, such as Active Learning methods. Here, we propose technique combination Fisher information matrix Sparse Proper Generalized Decomposition that enables definition informativeness criterion in dimensions. We provide examples proving performances this theoretical 5D polynomial function an crash simulation application. The results prove proposed strategy over-perform usual ones.