Semi-Supervised Learning in Industrial Quality Control
M. Wieler, Jens Roeder, A. Walstra, F. A. Hamprecht
AimHeaps of unlabeled data are often recorded in industrial processes, but labeling all is impractical (due to the sheer amount of data) or expensive (because user interaction is needed), only few labels can be generated. Therefore concepts are sought, which make optimal use of both data types.
Therefore we are developing concepts in graph-based semi-supervised and active learning, which use both labeled and unlabeled data, which can be used in industrial quality control.
Last update: 06.10.2010, 12:28