We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: A surrogate loss function for optimization of $F_β$ score in binary classification with imbalanced data

Abstract: The $F_\beta$ score is a commonly used measure of classification performance, which plays crucial roles in classification tasks with imbalanced data sets. However, the $F_\beta$ score cannot be used as a loss function by gradient-based learning algorithms for optimizing neural network parameters due to its non-differentiability. On the other hand, commonly used loss functions such as the binary cross-entropy (BCE) loss are not directly related to performance measures such as the $F_\beta$ score, so that neural networks optimized by using the loss functions may not yield optimal performance measures. In this study, we investigate a relationship between classification performance measures and loss functions in terms of the gradients with respect to the model parameters. Then, we propose a differentiable surrogate loss function for the optimization of the $F_\beta$ score. We show that the gradient paths of the proposed surrogate $F_\beta$ loss function approximate the gradient paths of the large sample limit of the $F_\beta$ score. Through numerical experiments using ResNets and benchmark image data sets, it is demonstrated that the proposed surrogate $F_\beta$ loss function is effective for optimizing $F_\beta$ scores under class imbalances in binary classification tasks compared with other loss functions.
Comments: 17 pages
Subjects: Machine Learning (cs.LG); Machine Learning (stat.ML)
ACM classes: I.5.2
Cite as: arXiv:2104.01459 [cs.LG]
  (or arXiv:2104.01459v1 [cs.LG] for this version)

Submission history

From: Namgil Lee [view email]
[v1] Sat, 3 Apr 2021 18:36:23 GMT (1619kb,D)

Link back to: arXiv, form interface, contact.