Building Bridges between Regression, Classification and Clustering by José Celestino

This presentation summarizes the paper “Building Bridges between Regression, Classification and Clustering”, which proposes improving supervised regression by reformulating it as a classification problem inspired by clustering approaches, and introduces a range of novel methods, including a unified end-to-end framework.

Presentation

Key Takeaways

• Reframing regression as a classification problem can improve performance on regression tasks.
• Soft-binning classification improves on baseline one-hot target labels
• End-to-end joint method. that bridges this tasks improves by results by 25% when compared to baseline least squares.

Reference

Stewart, L., Bach, F., & Berthet, Q. , 2025. Building Bridges between Regression, Clustering, and Classification.. https://arxiv.org/pdf/2502.02996