Many geophysical problems have an abundance of unlabeled data and a paucity of labeled data, and lithology classification of wire-line data reflects this situation. Training supervised algorithms on small labeled data sets can cause over-fitting, and subsequent predictions for the numerous unlabeled data may be unstable. However, semi-supervised algorithms are designed for classification problems with limited amounts of labeled data and are theoretically able to achieve better accuracies than supervised algorithms in these situations. We explore this hypothesis by applying two semi-supervised techniques to a well log dataset and compare their performance to three supervised algorithms. Our findings suggest that the semi-supervised methods we considered can match or outperform supervised methods if the model assumptions are met.
Presentation Date: Wednesday, September 18, 2019
Session Start Time: 1:50 PM
Presentation Time: 3:55 PM
Presentation Type: Oral