Abstract: In the real world, the problem of weakly supervised learning often arises: this happens in cases where only part of the data is well labeled, and the other part is not labeled at all, or the labeling contains noise or errors. In this article, we consider optimization approaches to solving the problem of weakly supervised regression using manifold regularization in combination with a cluster ensemble, which is needed to obtain a co-association matrix. To reduce the training time and increase the quality of the resulting co-association matrix, a probabilistic approach is applied using fuzzy clustering. We also evaluated the improvement in the quality and stability of the solution from the use of each of the proposed techniques, and also compared the final algorithm with classical machine learning algorithms that are used in supervised learning problems.

Cite: Kalmutskiy K. , Berikov V.
Manifold Regularization and a Probabilistic Approach in the Problem of Weakly Supervised Regression
In compilation Mathematical Optimization Theory and Operations Research: Recent Trends. 21st International Conference, MOTOR 2022 Petrozavodsk, Russia, July 2–6, 2022 Revised Selected Papers. – Springer Nature., 2022. – ISBN 9783031162237.

Dates:

Accepted:

Jun 15, 2022

Identifiers: No identifiers

Citing: Пока нет цитирований